Portfolio Short Run Marginal Cost of

Electricity Supply in Half Hour

Trading Intervals
Technical Paper

Author: Adam McHugh

11 January 2008
 Economic Regulation Authority

Important Notice

This document has been made available in good faith by the Economic Regulation Authority
(Authority). This document is not a substitute for professional advice. No person or organisation
should act on the basis of any matter contained in this document without obtaining appropriate
professional advice.

The Authority and its staff members make no representation or warranty, expressed or implied, as
to the accuracy, completeness, reasonableness or reliability of the information contained in this
document, and accept no liability, jointly or severally, for any loss or expense of any nature
whatsoever (including consequential loss) arising directly or indirectly from any making available of
this document, or the inclusion in it or omission from it of any material, or anything done or not
done in reliance on it, including in all cases, without limitation, loss due in whole or part to the
negligence of the Authority and its employees. This notice has effect subject to the Trade
Practices Act 1974 (Cwlth) and the Fair Trading Act 1987 (WA), if applicable, and to the fullest
extent permitted by law.

Any summaries of legislation, regulations or codes in this document do not contain all material
terms of those laws or obligations. No attempt has been made in the summaries, definitions or
other material to exhaustively identify and describe the rights, obligations and liabilities of any
person under those laws or provisions.

A full copy of this document is available from the Economic Regulation Authority web site at
www.era.wa.gov.au.

For further information, contact:

Economic Regulation Authority

Perth, Western Australia
Phone: (08) 9213 1900

Economic Regulation Authority 2008

The copying of this document in whole or part for non-commercial purposes is permitted
provided that appropriate acknowledgment is made of the author, the Economic Regulation
Authority and the State of Western Australia. Any other copying of this document is not
permitted without the express written consent of the Authority
Economic Regulation Authority

Contents
1 Introduction 3
1.1 Economic efficiency 4
1.2 The definition of short run marginal cost (SRMC) 5
2 Short run cost concepts 7
2.1 The distinction between the short run and the long run 7
2.2 Defining economic cost: The opportunity cost concept 8
2.3 Sunk costs 9
2.4 Avoidable costs 10
3 Unit commitment 14
3.1 Short run time frames: The 24 hour trading day and the half hour trading
interval 14
3.2 Linkages between unit commitment and short run economic cost 14
3.3 The mix of generation technologies 15
3.4 Shutdown costs vs. mingen costs 16
3.5 Time averaging of cost estimations 17
4 Short run economic costs for individual generating facilities over a half hour
trading interval 19
4.1 Short run total cost for a plant with sunk startup costs and small shutdown
costs occurring below mingen 19
4.2 SRMC for an individual plant with sunk startup costs and small shutdown
costs occurring below mingen 20
4.3 Short run total cost for a plant with sunk startup costs and large shutdown
costs occurring below mingen 21
4.4 SRMC for a plant with sunk startup costs and large shutdown costs occurring
below mingen 22
4.5 Short run total cost for a plant with avoidable fixed costs 23
4.6 SRMC for a plant with avoidable fixed costs 23
4.7 Short run total cost for a plant with contracted commitments to output 24
4.8 SRMC for a plant with contracted commitments to output 25
5 Cost measurement and estimation methods 27
5.1 Determination of variable cost components 27
5.1.1 Fuel input 27
5.1.2 Operational and maintenance costs: Planned outages 30
5.1.3 Operational and maintenance costs: Unplanned outages 31
5.1.4 Water input 32
5.1.5 Other variable costs 33
5.2 Determination of non-variable cost components 33
5.2.1 Startup costs 33
5.2.2 Shutdown costs 34
5.3 Derivation of a plants short run total cost curve 35
5.3.1 Plants with sunk startup costs 35
5.3.2 Plants requiring startup to produce in a half hour trading interval 36

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 1
Technical Paper
 Economic Regulation Authority

5.4 Derivation of a plants SRMC curve 37

6 Line losses 38
6.1 Line losses and the SRMC optimisation problem 38
6.2 Line loss factors in the WEM 40
7 Portfolio short run cost over a half hour time frame 43
7.1 Short run total cost for a portfolio of plants with sunk startup costs 43
7.1.1 Example 1: Linearly increasing total cost technology, portfolio short
run total cost determined by a SRMC order of merit 43
7.1.2 Optimal portfolio SRMC 46
7.1.3 Example 2: Cubic based short run total cost technology, the (sub-
optimal) SRMC order of merit approach 47
7.1.4 Optimal short run total cost curve: The lower envelope of all feasible
cost paths 49
7.1.5 Optimal portfolio SRMC 50
7.1.6 Example 3: The effect of differing shutdown and mingen costs
between plants 51
7.1.7 Optimal portfolio SRMC 54
7.2 The equimarginal principle 55
7.3 Short run total cost for a portfolio comprising one or more plants with
avoidable fixed costs 56
7.3.1 Example 4: Linearly increasing total cost technology with differing
shutdown costs and startup costs between plants 56
7.3.2 Optimal portfolio SRMC 60
7.3.3 Example 5: Portfolio of plants with cubic based total cost functions
and with differing shutdown costs and startup costs between plants 60
7.3.4 Optimal portfolio SRMC 61
8 Discussion 63
8.1 Estimation of plant level functional form and its effect on portfolio SRMC 63
8.2 Basic efficiency criteria with respect to a saw tooth SRMC function 64
8.3 Empirical issues 67
8.4 Real competition vs. replicated competition 68
9 Conclusions 71
References 72
Appendices 74
Appendix 1. Optimisation examples 75
A1.01 Individual plant models 75
A1.02 Model of a portfolio comprising plants with sunk costs 79
A1.03 Portfolio model with a mix of sunk cost and avoidable fixed cost
plants 81

2 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

1 Introduction
This paper has been prepared by Adam McHugh, Manager of Projects, in the Economic
Regulation Authoritys Competition Markets and Electricity Division, in the interests of
advancing the debate in economic regulation. The paper seeks to assist market
participants in the understanding of short run marginal cost (SRMC). In doing so, it sets
out to identify what costs may be included in a firms short run portfolio supply curve
calculation and how SRMC may be estimated. The views presented herein are those of
the author and should not be taken as reflecting the views of the Authority, individual
members of the Authority, the Authoritys Secretariat, or members of other organisations. 1

Rules pertaining to the Wholesale Electricity Market (WEM) in Western Australia (WA)
are, in part, designed to replicate the outcomes of a competitive market. Specifically,
generators are required to offer electricity at the SRMC of production. Clause 6.6.3 of the
Wholesale Electricity Market Amending Rules (December 2006) (referred to in this
document as the market rules) states:
6.6.3. A Market Generator must not, for any Trading Interval, offer prices within its Portfolio
Supply Curve that do not reflect the Market Generators reasonable expectation of the
short run marginal cost of generating the relevant electricity when such behaviour relates
to market power.

The roles of the Economic Regulation Authority (Authority) and the Independent Market
Operator (IMO) with respect to clause 6.6.3. are outlined in clause 2.16.9. and its various
subclauses as follows:

2.16.9. The Economic Regulation Authority is responsible for monitoring the effectiveness
of the market in meeting the Wholesale Market Objectives and must investigate any
market behaviour if it considers that the behaviour has resulted in the market not
functioning effectively. The Economic Regulation Authority, with the assistance of the IMO,
must monitor:

(b) inappropriate and anomalous market behaviour, including behaviour related to market
power and the exploitation of shortcomings in the Market Rules or Market Procedures
by Rule Participants including, but not limited to:
i. prices offered by a Market Generator in its Portfolio Supply Curve that do not
reflect the Market Generators reasonable expectation of the short run marginal
cost of generating the relevant electricity;

2.16.9A. The IMO must assist the monitoring activities identified in clause 2.16.9(b)(i) by
examining prices in STEM Submissions, including Standing STEM Submissions, used in
forming STEM Bids and STEM Offers against information collected from Rule Participants
in accordance with clauses 2.16.6 and 2.16.7.
2.16.9B. Where the IMO concludes that prices offered by a Market Generator in its
Portfolio Supply Curve may not reflect the Market Generators reasonable expectation of
the short run marginal cost of generating the relevant electricity and the IMO considers that
the behaviour relates to market power the IMO must:
(a) as soon as practicable, request an explanation from the Market Participant which has
made the relevant STEM Submission; and

1
Comments and/or feedback in the development of this paper came from Peter Kolf and Robert Pullella of
the Economic Regulation Authority, Dora Guzeleva of the Independent Market Operator, and Duncan
Farrow of Murdoch University, and are gratefully acknowledged. The author takes sole responsibility for
any errors that may remain.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 3
Technical Paper
 Economic Regulation Authority

(b) advise the Economic Regulation Authority of its conclusions. The IMO advice must
outline the reasons for the IMOs conclusions.

2.16.9G. Where the Economic Regulation Authority determines that prices in the Portfolio
Supply Curve, subject to the investigation, did not reflect the Market Generators
reasonable expectation of the short run marginal cost of generating the relevant electricity,
the Economic Regulation Authority must request that the IMO applies to the Energy
Review Board for an order for contravention of clause 6.6.3.
2.16.9H. Where the IMO receives a request under clause 2.16.9G the IMO must refer the
relevant matter to the Energy Review Board requesting that a civil penalty be imposed on
the relevant Market Participant.

In summary, the Authority, with the assistance of the IMO, is responsible for monitoring
the effectiveness of the market in meeting the Wholesale Market Objectives and must
investigate any market behaviour if it considers that the behaviour has resulted in the
market not functioning effectively. If a firm with market power submits a portfolio supply
curve that does not reflect that firms reasonable expectation of SRMC for any given
trading interval and the Authority determines that to be the case, the matter must be
referred by the IMO to the Energy Review Board requesting that a civil penalty be
imposed on the relevant market participant. 2

A non-technical version of this paper entitled Short Run Marginal Cost (Economic
Regulation Authority, 2008) invites public submissions on the matters raised. A copy of
both versions are available on the Authoritys web site.

1.1 Economic efficiency

Economic theory labels a market as efficient if the net benefits (i.e. the total benefits
minus the total costs) created in that market reach a notional maximum. An inefficient
market wastes resources, either because:

a) more goods could have been created with the same amount of resources, and/or;
b) goods are not allocated in accordance with the tastes and preferences of
consumers.

That is, in the short run, economic efficiency requires:

a) technical efficiency;
b) allocative efficiency, and;
c) in the long run, dynamic efficiency.

Under theoretical assumptions that describe a competitive market, efficiency will occur
naturally. These assumptions are:

1) No market participant has the ability to independently influence the clearing price
by virtue of market share or control (i.e. market power);
2) There is no strategic behaviour in the market;
3) All participants are able to make fully informed decisions;

2
A firms market power generally corresponds to its ability to influence the clearing price. Future work will
focus on this concept more thoroughly.

4 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

4) There are no costs or benefits that accrue to third parties (i.e. externalities).

If these conditions are met, economic theory suggests that the price of electricity, or any
other good or service, will fall to SRMC. That is to say, price will be set by the cost of the
most expensive MWh of electricity produced during the relevant period of generation.

However, electricity markets are rarely, if ever, fully competitive. In the absence of
sufficient competition, an alternative is to replicate the competitive (and therefore efficient)
outcome through economic regulation and good market management.

Condition 3 above, for example, can be met by timely and centralised release of market
information by a market operator (the IMO in the case of WA). Condition 4 can be met by
taxes, subsidies or other forms of regulation that adjust the private costs of the firm so as
to internalise any external effects. This paper however, is concerned with conditions 1
and 2, which may, in theory, be met at relatively low regulatory cost by putting in place
legislation that requires firms to offer their output at SRMC. Provided firms act in
accordance with this rule (and provided conditions 3 and 4 are also appropriately
managed), the resulting price will be one that equals SRMC at the efficient equilibrium
between supply and demand.

1.2 The definition of short run marginal cost (SRMC)

The marginal concept in economics refers to the rate at which one quantity changes with
respect to extremely small increases in another quantity. It follows therefore that SRMC is
simply defined as the change in short run total cost for an extremely small change in
output. 3,4

3
Equally, SRMC can be defined as the change in short run variable cost for a small change in output.
4
The smaller the change in output used as a reference, the more accurate will be the measure of SRMC.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 5
Technical Paper
 Economic Regulation Authority

Box 1.1 Mathematical definition of SRMC

Consider the following generic function for an individual plant or a portfolio:

C = C (Q) , Q 0,

where short run total cost (denoted C) for a plant or a portfolio of plants is a function of its
output Q.

The most precise mathematical definition of SRMC is given by the first derivative of the
short run total cost function:

dC
C (Q) = , Q 0,
dQ

where SRMC (denoted C ) is a function of the plants output. That is, C is the
instantaneous rate of change of C with respect to Q as a continuous function.

This function can be very closely approximated by means of the following difference
equation:

C
C& = , Q 0,
Q

where SRMC (denoted C& ) is given as the change in cost over the change in quantity.
Use of this discrete method, however, is only appropriate for very small changes in Q such
that C& comes very close to C - i.e. the smaller the change in Q used, the closer the
result will be to that of the first derivative method.

6 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

2 Short run cost concepts

Since SRMC relates to the rate of change of short run total cost, an understanding of what
constitutes short run total cost is a pre-requisite for determining SRMC. Once such an
understanding is developed, SRMC can be derived mathematically. Statistical methods
can then be employed enabling data to be entered into an optimisation model (see
Section 5 and Appendix 1). Unfortunately, a disparity between accounting concepts of
cost and economic concepts of cost can often lead to confusion at the conceptual level.
Economic regulators expect participants to use the economic concept of cost in their
interpretation of the market rules. Part of the scope of this paper therefore is to state
clearly the economic concept of cost in a way that corresponds to these expectations.

The economic understanding of cost incorporates a number of fundamental concepts that

are not immediately intuitive. These are:

the distinction between the short run and the long run;
the distinction between sunk and avoidable costs;
the distinction between fixed, variable, shutdown, and total costs, and;
the concept of opportunity cost.
This section has two purposes:

1) to describe the above concepts; and

2) to provide clarification as to what should be incorporated in the economic
determination of SRMC.

2.1 The distinction between the short run and the long
run
The distinction between the short run and the long run in economics is only superficially
about any notion of duration or time. Rather, the distinction has more to do with whether
or not there is full scope to substitute a more cost effective input for a less cost effective
input.

An electricity generator that is able to choose any technique within budget for combining
all the various inputs such as plant and machinery, fuel, labour, land, etc. has much more
flexibility to minimise cost than a firm that is committed to a certain quantity of one or more
inputs. If the price of one type of input changes relative to another, a manager with the full
flexibility to utilise any affordable technique can simply choose one that employs less of
the more expensive input and more of the cheaper one. If fuel becomes cheap, for
example, it may become economic to sell off some expensive assets to provide the
necessary cash to purchase more fuel: this is fuel input substituting for capital input.

Alternatively, with rising fuel costs, the installation of fuel saving equipment may become
economic: this is capital input substituting for fuel input. When a firm is able to optimise its
expenditure in any way it pleases, it is able to choose that particular mix of inputs that
produce output at the minimum technologically feasible cost. This can only occur over the
long run, when the firm is free of any irrevocable commitments to input. In other words,
the long run is how long it takes for the technologically feasible mix of inputs to be able to
be (re)arranged in any combination a manager pleases.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 7
Technical Paper
 Economic Regulation Authority

A firm that is irrevocably committed to quantities of one or more input type does not have
the same flexibility. In this case the firm is said to face a short run decision. In the short
run, the firm must optimise as best it can given that it cannot vary certain quantities of
input. In other words, the short run means that a manager is stuck with some amount of
input that cannot be swapped for a different type of input. A firms portfolio of generating
facilities, for example, cannot be varied in the short run. Similarly, sunk labour contracts
constitute a short run constraint.

2.2 Defining economic cost: The opportunity cost

concept
An economic decision is one in which the decision maker acknowledges all the alternative
opportunities that are available to create benefits. The question that is always asked in a
rational economic decision is therefore what do I have to give up by taking this course of
action? which is equivalent to asking is this course of action going to create more net
benefits (profit) for myself (my firm) than any other possible course of action?. If the
answer to this second question is yes then the choice is an economic one. If the answer
is no, then the choice is uneconomic because an opportunity to create greater net
benefits has been forfeited. All economic costs incorporate this notion of opportunity cost.

Since an economic decision takes into account the benefits that have to be given up to
achieve greater benefits, a further, economically important, question must be asked: what
net benefits (profit) would I have created for myself (my firm) if I had taken the next best
course of action?. This question defines the concept of opportunity cost. An opportunity
cost is a potential benefit that is forgone to create greater benefits. The opportunity cost
of an input decision is therefore equal to the benefits that could have been derived from
employing resources in their next best use.

To clarify the economic concept of cost, consider the following example. A generating
plant, if started, will bring in revenue of $1.2M over a period. If the operating costs over
that period were to include past fuel expenditures of $0.6M plus other expenditures of
$0.4M, it could be said that the plant would make a $0.2M accounting profit.

However, before committing to operating the plant, it is appropriate to ask: what

alternatives are there?. If the value of the fuel required to operate the plant rose on the
current spot market to $0.9M, assuming the firm has the opportunity to sell the fuel, the
choice to produce electricity would be uneconomic. In this case, the firm could make an
accounting profit of $0.3M by selling the fuel as opposed to a $0.2M accounting profit by
selling electricity. In other words, the firm would profit by $0.1M more by selling the fuel
and not operating the plant. In economic terms, if the plant were to operate it could make
an economic loss of $0.1M.

As the above example suggests, the past purchase price or past contract price of fuel
does not represent a short run economic cost. Rather, the current price that can be
obtained by on-selling a quantity of fuel represents what is given up by using it to produce
electricity. It follows that if the fuel cannot be immediately on-sold then the opportunity
cost of that fuel for a firm in a competitive market may be zero unless there exists a real
prospect to either: (a) sell the fuel at some point in the future; or (b) use the fuel in the
future production of electricity at higher than current market values. 5 In either case, the
5
It is noted that if a plant is withdrawn as a result of fuel being withheld then penalties, including withdrawal
of capacity credits under the Reserve Capacity Mechanisms, may apply and would need to be taken into
consideration. Note, however, that from the regulators point of view it would be inappropriate to include
regulatory penalties in the calculation of SRMC under the legislated SRMC approach as this would result in
increased prices, thus reducing the incentives the penalties are intended to instill in producers.

8 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

current spot price, or alternatively the current market price for fuel contracts, still provides
the best basis for the opportunity cost of fuel if the view is taken that the current price
reflects the rational expectations (based on the latest information available to the market)
of all future prices for the resource. 6 However, this forgone benefit would need to be
adjusted downwards by storage costs, inflation, and the interest that could have been
earned while waiting for a buyer if the fuel had been converted to cash immediately.

2.3 Sunk costs

As stated in Section 2.1, firms in the short run face either technological or contractual
commitments to input. In economics, the costs associated with these commitments are
considered sunk. They cannot be avoided or reversed in the short run. 7

Importantly, the economic cost of a sunk commitment is zero. As already stated,

opportunity cost describes the potential benefits that are forgone by employing a resource
in its next best alternative use. By definition, expenditures that are sunk have no
alternative use since they cannot be avoided in the short run. It follows that sunk costs
have no influence on economic decision making in the short run. Rather, the short run
decision is entirely about current expenditure on marginal operating costs compared to
market price.

Take, for instance, the decision to build a gas fired power station as opposed to the
decision to sell electricity from a gas fired power station that has already been built. The
former is a long run decision: the firm must decide whether it believes the investment will
be worthwhile based on the present value of the expected flow of future revenues and
costs, with the certainty of revenue and cost flow typically guaranteed by long run
contracts with take-or-pay provisions (so as to ensure the recovery of sunk costs).

Once the plant is built however, the decision as to how much electricity should be
produced becomes a short run decision and the plants construction costs become
relegated to the past. In a competitive market, the firm would face a market price that it
cannot control: a take it or leave it proposition. Therefore, the optimal economic decision
for determining how much electricity should be produced, over and above that for which
take-or-pay contracts may exist, is determined simply by whether supply of an additional
MWh increases the firms bank balance or decreases it. This depends solely on whether
or not the price that is available for incremental amounts of electricity exceeds the
immediate costs of production.

Figure 2.1 provides an example. A firm operating in a competitive market is offered a

price of $30 per MWh to supply electricity. If the various current expenditures that the firm
must outlay to produce incremental MWhs of electricity are lower than $30, the sale of
that electricity will see the firms profits going up. If, however, an incremental MWh of
supplied electricity entails more than $30 of current expenditure, the firms profits will get
progressively smaller for each additional MWh produced. Thus, while the value of the firm
before the short run decision is made will have been affected by past decisions on capital

6
According to rational expectations theory, the market price in the tth period, designated Pt , is given

by: Pt = Pt e + u t , where Pt e is the expected price in the tth period, and where ut is a random error term
that is independent of the current price and has an expected value of zero (see: Muth, J. (1961), Rational
Expectations and the Theory of Price Movements, Econometrica, Vol. 29, No. 3, pp. 315-335).
7
In the long run there are no sunk costs because, by definition, there are no irrevocable commitments to
input. There are, however, redundant investments, the scrap value of which may be equal to or close to
zero.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 9
Technical Paper
 Economic Regulation Authority

expenditure, the amount previously invested will have absolutely no bearing on the best
decision that the firm can make today to see the value of the firm head in the right
direction. If, for example, the firm seeks to recover all of its capital costs by asking for a
higher price, the buyers in a competitive market will simply get their electricity elsewhere.

$ Short Run
Marginal Cost

Decreasing Profit*
30 Market Price =
Increasing Profit* Marginal Revenue

Max Profit MWh

Figure 2.1 Profit maximising short run supply decision in a perfectly competitive market
* Note: the short run marginal cost curve does not account for avoidable fixed costs such as those associated
with the startup of a plant during a trading interval. These must be subtracted from any profit or added to loss
for the trading interval. Avoidable fixed costs are independent of output and incurred as a lump sum. They
differ from sunk costs in that the latter are incurred prior to the current trading interval and so are unavoidable
in the current trading interval.

To define in summary:
Sunk costs. A sunk cost is a fixed cost that cannot be avoided in any given (short
run) time period.

2.4 Avoidable costs

An avoidable cost is an expenditure to which the firm is not irrevocably committed or that
can be reversed over the time period in question. Therefore, unlike a sunk cost, the
economic (opportunity) cost of an avoidable cost will be of positive amount because the
money or the resources saved by avoiding the cost could potentially be used to generate
benefits elsewhere. Avoidable costs can be broken down as follows:
Avoidable fixed costs. An avoidable fixed cost is an expenditure that must be
borne by the firm if it chooses to produce any amount of output in a given time
period. That is, an avoidable fixed cost is an expenditure that remains constant for
any level of output above zero but that can be avoided by producing zero output.
An avoidable fixed cost should not be confused with a sunk cost, which cannot be
avoided in the short run even if the firm were to choose not to produce. 8 The most
relevant avoidable fixed cost in the context of electricity production is the cost of
starting a plant.
Shutdown costs. A shutdown cost is an expenditure that must be borne by the
firm if it has invested in plant, is operating that plant and chooses to produce zero
8
While all sunk costs can be considered fixed costs, not all fixed costs are sunk (see Wang, H. & Yang, B
(2001), Fixed and Sunk Costs Revisited, Journal of Economic Education, Vol. 31, No. 2, Spring, pp. 178-
185).

10 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

output in a given time period. A shutdown cost is therefore the opposite of an

avoidable fixed cost in that it can only be avoided by producing non-zero output.
In the context of electricity production, a plant already operating in a half hour
trading interval will suffer a short run shutdown cost if demand for its output does
not equal or exceed its minimum technical output (mingen). 9 The shutdown cost
corresponds to the maximum amount a firm would be willing to pay to avoid the
shutdown. This will include any loss in net revenue that could be avoided if the
plant remained operational.
Variable costs. A variable cost is a cost that varies with the level of output
(beyond mingen) that the firm produces over the relevant time period. Prominent
variable costs in the context of electricity production are fuel costs, operating
costs, costs associated with wear and tear on plant and equipment. These costs
may either vary directly in proportion with output or otherwise change in steps as
output increases (e.g. labour costs). Most variable costs increase with output but
there are some that decrease with output (i.e. some variable costs are avoidable in
the short run by producing higher levels of output). If, for example, an electricity
generator has commitments to provide steam to nearby industries, any per unit
amount of compensation that they must pay to that firm if they do not provide the
steam constitutes a decreasing variable economic cost.
Total cost. Total economic cost is the variable cost for a given level of output plus
any avoidable fixed costs or shutdown costs. Sunk expenditures are also a
component of total cost, but because they have a zero opportunity value they do
not affect its calculation.
Figure 2.2 describes a taxonomy of cost categories consistent with the discussion above.
Figure 2.3 provides sequential questions that enable a clear categorisation of the short
run total cost components outlined in this section.

Variable Not Variable

Avoidable Fixed Costs

Avoidable Variable Costs
Shutdown Costs

Not Avoidable Sunk Costs

Figure 2.2 Categorisation of short run economic costs associated with electricity
generation with given examples.

9
Mingen is the minimum amount of output (and hence input) required to keep a plant running.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 11
Technical Paper
 Economic Regulation Authority

Start

It is a sunk cost.

No The cost is now irrelevant to the short

Can this cost be run economic decision. Therefore, the
avoided? cost is considered zero value in
economic (opportunity) terms.

Yes

It is a variable cost.
Yes
Does the cost vary with The cost has an economic value equal to
output? the opportunity cost of input associated
with incremental levels of output.

No

It is an avoidable fixed cost.

No
The cost has an economic value equal to
Does the cost only occur if the opportunity cost of the inputs that
output falls to zero? could be avoided if output was zero for
the half hour.

Yes
It is a shutdown cost

The cost has an economic value equal

to the opportunity cost of resources
required to shut the plant down and any
lost net revenue that could be avoided if
the output was non-zero for the half
hour.

Figure 2.3 Sequential question method of categorising the components of short run total
economic cost for a half hour trading interval.

Below are examples of the costs included in each of the categories identified in Figure 2.2
above.

Variable costs

Value of saleable fuel (at the current market rate) used to produce electricity
output.
Wear and tear on plant and equipment directly attributable to the production of
output.
The expected costs of plant failure - i.e. the probability of plant failure multiplied by
the cost of plant failure, where the probability of plant failure increases with the
level of output. 10
Value of water and other inputs used to produce electricity output.
10
Such costs are associated with unplanned maintenance and include not only repair costs, but also lost net
revenues as a result of plant unavailability, as well as financial penalties imposed for unscheduled outage.
It should be noted that if the probability of unplanned outage is not correlated with high levels of output, the
risk of equipment failure becomes an avoidable fixed cost as opposed to a variable cost.

12 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Costs as a result of per unit financial penalties imposed for not meeting contracted
output commitments. 11
Cost of pollution control permits, per unit taxes and other output linked regulatory
instruments.

Avoidable Fixed Costs

Fuel, water and other operating costs of starting a plant within the current trading
interval.

Shutdown costs

Fuel, water, and operating and maintenance costs of shutting a plant down within
the current trading interval.
Lost net revenue in future trading intervals as a result of the plant lying idle while
awaiting future startup (as opposed to incurring costs resulting from the plant
operating at mingen).
Includes costs associated with a future startup minus mingen costs over the
intervening period (adjusted for any avoided cost of base-load).

Sunk costs

Capital cost of plant and equipment.

Cost of starting a plant prior to the current trading interval.
Cost of fuel purchased on long term contracts that cannot be on-sold to third
parties.
Labour costs (apart from avoidable labour costs, such as overtime). 12

11
Note: where such costs relate to regulatory penalties intended to influence a firms private decision (e.g.
penalties for not meeting obligations under the reserve capacity mechanism or for renewable energy
certificates) they are not a cost a regulator would allow to be included in SRMC calculations.
12
Because the hiring of labour occurs over periods long than a half hour or even a day, such costs are sunk
in the short run.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 13
Technical Paper
 Economic Regulation Authority

3 Unit commitment
The time it takes to start plants up and shut them down again constrains the availability of
various plants for any given trading interval and therefore affects the day ahead estimation
of portfolio SRMC. It is reasonable to assume that this unit commitment decision is
motivated by the dual incentives of reliability and profit. Therefore, the unit commitment
plan, under the assumption of profit maximising behaviour, becomes the foundation upon
which a reasonable expectation of portfolio SRMC is built. What follows in this section is
a brief discussion of unit commitment considerations and how these impact upon the
reasonable, day ahead, expectations of half hourly economic cost.

3.1 Short run time frames: The 24 hour trading day and
the half hour trading interval
By the very nature of the product, the scheduling of electricity supply must be considered
with reference to very small time intervals. This is because electricity dispatch requires a
near perfect instantaneous balancing between supply and demand otherwise the system
will move outside of a narrow technical envelope and collapse at great cost. 13
Consequently, market equilibrium is an engineered solution in the first instance and the
economically efficient price the retrospective determination of a market operator.

Typically, an electricity market is cleared on a half hourly basis (or more frequently),
involving both planned and real time events. In Western Australia, market participants
submit portfolio supply and demand schedules on a day ahead basis to the IMO.
Electricity generators, therefore, make economic short run decisions that correspond to
strictly predefined temporal bounds. The predefined bounds in the Western Australian
WEM are a 24 hour trading day (8am to 8am) divided up into 48 half hourly trading
intervals. To the extent that real time events differ from the submitted day ahead portfolio
supply and demand schedules, a balancing mechanism settles any differences.

3.2 Linkages between unit commitment and short run

economic cost
In the States wholesale electricity market there are essentially two relevant short run
questions facing electricity generators:

1) How should resources be allocated over the trading day?

2) What is the profit maximising output for each half hour trading interval?

These two questions are inextricably linked. If the primary goal of the firm is to maximise
profit over the trading day, this can only be achieved under competitive conditions if the
SRMC of electricity supply is set at marginal revenue in each half hourly period. However,
the optimal allocation of resources through the trading day will affect the SRMC of supply
in each trading interval. The profit motive will cause a firm to attend to this allocation of
resources, known as unit commitment, as efficiently as possible. Hence, the assumption
of profit maximising behaviour by a firm simplifies analysis.

13
Electricity supply tolerances are defined and regulated in technical rules, which are very narrow to maintain
the integrity of the system. Electricity market equilibrium is therefore quite artificial and heavily dependent
on engineering and information technology for its existence.

14 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

The profit motive, however, can be a two edged sword. While it provides firms with the
incentive to minimise their costs of production, a profit maximising firm with the ability to
raise price by virtue of market power can extract excessive profits at the expense of
consumers, resulting in less net benefits for society as a whole. This is to say that in the
presence of market power the profit maximising level of output will not be economically
efficient.

A firm operating in a competitive (efficient) market, however, would find the profit
maximising level of output occurs where price equals SRMC in each half hour trading
interval. This would produce a socially optimal outcome. 14 The approach taken in the
WEM is to attempt to replicate this outcome by requiring generators to submit their
electricity supply offers so as to reflect SRMC in each trading interval. To the extent this
can be implemented, enforced, and provided the economic definition of cost is used, the
level of profit will reflect an efficient market outcome thereby arresting the problem of
market power.

3.3 The mix of generation technologies

Figure 3.1a shows a typical diurnal pattern of electricity output. Given the periodic pattern
of demand a mix of technologies is required to continuously balance system output with
load over a trading day. Generation technologies can be categorised as either
intermittent, base-load, mid-merit or peaking plants. The following characteristics
generally apply:

a) Currently, plants that use intermittent resources such as wind and solar provide a
relatively uncertain (stochastic) level of output in a trading interval, but have zero
fuel costs and so are typically the lowest SRMC plants. This is particularly likely to
be the case where environmental and technological regulations such as renewable
energy certificates, tradable pollution permits, and/or carbon taxes are significant.

b) Base-load plants have low variable costs but high startup and shutdown costs, and
have a relatively limited capability to ramp output up or down to follow load. Base-
load generation in WA often uses coal to produce steam to be converted into
electricity.

c) Mid-merit or load following units are normally medium cost plants both in terms of
variable cost, startup cost and shutdown cost, and are quite effective at adjusting
(ramping) output up or down to cover reasonably rapid changes in demand. In
WA mid-merit plants are usually natural gas fuelled turbines or co-generation
plants.

d) Peaking plants in WA are typically based upon gas turbine technology. These
units are very effective at following load and can be started up and shutdown
quickly and cheaply. However, they are usually the highest variable cost plants in
a portfolio, particularly when distillate is used as the fuel input.

14
Assuming any externalities are internalised through supplementary government policy.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 15
Technical Paper
 Economic Regulation Authority

MW Trading day 1 Trading day 2

Peaking

Base-load
variable Base-load
output unit forced
Mid-merit shutdown

Base-load
mingen
output

Intermittent Sources
0 4 8 48 96
Trading Intervals
Figure 3.1a Electricity output over two trading days

MW Mid-merit
mingen
output Mid-merit
variable output

Base-load
variable output

>0 4 5 6 7 8
Trading Intervals
Figure 3.1 Figure 3.1

Figure 3.1b Electricity output over five trading intervals

3.4 Shutdown costs vs. mingen costs

In maximising profits over a trading day, a critical consideration is the comparison
between the cost of a future startup for a plant (which can be large) and the cost of
keeping the plant running at mingen over the intervening period. The primary example is
base-load generation. Whenever possible, coal fired base-load facilities will continue to
operate even when electricity demand is low. This is because the costs associated with
shutting a coal fired base-load plant down are large.

The opportunity costs of forcing such a plant below mingen will include not only the
immediate costs associated with taking the plant offline but also the cost of starting the

16 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

plant up again when it is required. 15 The time that it takes to have such a plant come back
into operation can be considerable. If this causes the plant to be unavailable when it is
needed there will be an additional opportunity cost associated with lost revenue in future
trading intervals while the plant is lying idle. In other words, while within the trading
interval it may be cheaper to shut a plant down than to run the plant, it may not be the
best decision over the trading day. Therefore, the impact on cost in future trading
intervals must be considered in the current decision. For this reason, and for reasons of
security and reliability, coal fired plants are, ideally, only shut down for scheduled
maintenance. 16

The previous startup of a plant (i.e. in a prior half hour trading interval) constitutes a sunk
expenditure over the short run time period of half hourly electricity supply. That is, once a
plant is started, the associated expenditure is committed and can no longer be avoided.
Conversely, the costs of operating a plant at mingen or above are not sunk because they
can be avoided by shutting the plant down for the relevant half hour trading interval. Note,
however, that the shutdown cost can be either greater than or less than the cost of
operating at mingen, depending on the particular technology involved. Where the cost of
operating a plant at mingen is substantially lower than the cost of shutting the plant down,
a shutdown decision over a temporary period of low demand may prove uneconomic. In
any case, the standard opportunity cost criterion applies: the value of resources in their
alternative use should be fully accounted for. For example, the decision as to whether the
mid-merit plant in Figure 3.1b should be shutdown prior to trading interval 7, or
alternatively, left operating at mingen, would have to account for the avoided cost of
variable base-load (that is, the cost of base-load output the mid-merit plant will offset if it is
not shutdown).

3.5 Time averaging of cost estimations

Technological (metering) constraints typically require electricity output to be measured in
units of power. Conversely, market measures of electricity supply require a comparison to
be made between price and energy. As such, a suitable method must be adopted to
bridge between inter-interval representations such as Figure 3.1a and Figure 3.1b (which
give electricity output in units of power on their vertical axes against time on the horizontal
axis) to intra-interval representations such as Figure 2.1 (which gives electricity output in
units of energy on the horizontal axis against monetary units on the vertical axis).

Because electricity supply systems require output to be continuously balanced against

load, it is extremely unlikely, given the volatility of electricity demand, that the systems
output will be constant with time, even over small trading intervals. As a result some
degree of time-averaging will be required to convert dynamic electricity supply data (i.e.
power with respect to time over a trading day), to static, economic, representations of
electricity supply (i.e. energy with respect to price over a trading interval). This is likely to
produce a level of inaccuracy in cost estimations whenever mid-merit plants and, in
particular, peaking plants, which have the ability to ramp output up or down relatively

15
To avoid costly damage to steam turbines associated with expansion and contraction, venting steam at low
demand is not an option for most base-load plants. If this were technologically feasible, it would be
cheaper to operate a coal fired base-load plant at mingen and vent steam during periods of low demand for
electricity rather than shut it down.
16
The costs alluded to here can be avoided if a firm commits itself to a low (possibly negative) bidding
strategy that ensures dispatch. This can be thought of as a payment to the market that ensures against
another producer dispatching energy into the system. This discourages alternative sources of supply
enough to enable the base-load plant to be kept above mingen. A profit maximising firm would be willing to
pay the market up to the opportunity cost of shutting the plant down not to supply electricity.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 17
Technical Paper
 Economic Regulation Authority

quickly to follow load, are required to be operated to a high degree of intra-interval

variability in output.

A related constraint arises because of the relative inability of some power stations to ramp
output up or down quickly enough to respond to rapid fluctuations in electricity demand.
As already mentioned in Sections 3.3 and 3.4 above, base-load plants are designed to
operate continuously, typically at their rated capacity, and are only shut down for planned
or unplanned maintenance. Moreover, while some base-load power stations are able to
be run at reduced output, others cannot ramp output up or down at all, and those that can
vary output will do so at a much slower rate than mid-merit and peaking plants.
Therefore, there is likely to be a limit to a base-load plants feasible variability in output
from one trading interval to the next. In this way, optimisation of plant and equipment over
the trading day may impose limits to the range of output that some plants can offer for
particular trading intervals.

Finally, it should be noted that demand cannot be perfectly predicted a day ahead, and
the optimal allocation of generation resources is required to be reactive to market
conditions in real time. Therefore, plant availability, and the range of output deemed to be
available from particular plants, may need to be adjusted to reflect response time
considerations.

18 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

4 Short run economic costs for individual

generating facilities over a half hour trading
interval
This section will show how the concepts of short run total cost and its SRMC derivative
apply to individual generating facilities within an economic price-quantity framework. Two
broad categories of plant are considered: those operating with sunk costs and those for
which operations comprise an avoidable fixed cost component. Geometric interpretations
are supported by mathematical models (see Appendix 1). Cost information at the plant
level becomes the input for an optimisation process (outlined in Section 7 below) enabling
the construction of a portfolio SRMC curve, which in turn becomes the basis for the
portfolio supply curve as required by clause 6.6.3 of the market rules.

At least three general categories of plant-specific short run total cost curve have been
postulated, as an a priori set of hypotheses, for statistical based economic analysis : those
based on cubic functions, those based on quadratic functions, and those based on linearly
increasing functions. 17 SRMC is the rate of change (i.e. the first derivative) of the short
run total cost function. This suggests, as a starting point, three corresponding categories
of plant level SRMC curve to consider: quadratic functions, linearly increasing functions,
and linear constants. In other words, in a cost-output relationship, SRMC curves can be
considered either U shaped, increasing with output, or flat. The immediate discussion
(below) compares each of these postulated functions side by side. Appendix 1 applies the
optimisation process with reference to a hypothetical portfolio of plants, each with U
shaped SRMC functions.

It should be understood that the exact nature of SRMC functions can only be determined
by observation. Evidence collected by the author pertaining to Western Australian and
Californian plants suggests typically U shaped plant-specific SRMC functions. 18
Therefore, pending statistical analysis of data pertaining to Western Australian generating
facilities, each of the three categories of function will be considered in the following
discussion.

4.1 Short run total cost for a plant with sunk startup
costs and small shutdown costs occurring below
mingen
By definition, a plant can only face shutdown costs if it is operating during the trading
interval with the cost of prior startup considered sunk. Figure 4.1 depicts three possible
versions of the short run total cost curve over a half hour trading interval for a single

17
Johnston, J. (1952), Statistical Cost Functions in Electricity Supply, Oxford Economic Papers, New Series,
Vol. 4, No.1, pp. 68-105. This study considers the cost-quantity relationship of steam turbine plants in the
UK in both short run and long run analyses. It finds statistical evidence for a reasonably constant short run
marginal cost function. Note, however, that Johnston defines the short run purely in terms of a fixed capital
stock, as opposed to the predefined bounds of a half hour trading interval, and relies solely on aggregated
output and total cost data as opposed to technologically detailed modelling of electricity supply.
18
A U shaped SRMC curve is consistent with the law of diminishing returns and is therefore one that is
commonly proposed in economic theory. Moreover, U shaped SRMC functions are consistent with the
engineering studies of incremental heat rate for fossil fuel plants, thus incorporating a major
physical/technological characteristic relevant to electricity generation cost in the short run. See, for
example, Schweppe, F., Caramanis, M., Tabors, R. & Bohn, R. (1988), Spot Pricing of Electricity, Kluwer
Academic Publishers, Boston, pp. 283-285.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 19
Technical Paper
 Economic Regulation Authority

generation plant with sunk startup costs and small shutdown costs relative to the total cost
of operating the plant at mingen. The important features of Figure 4.1 are:

1) The curves are non-declining with output.

2) The curves get very steep at high levels of output as the thermodynamic limits of
the capital in place are approached at K. 19 This will occur to the right of some
manufacturer recommended output rating for the plant (denoted R).
3) The tangents of the short run total cost curves represent their rate of change. The
slope of each tangent represents the SRMC of the plant at that level of output.
4) The steeper the slope of a tangent line, the higher the rate of change and the
higher the SRMC.
5) Below mingen output inevitably falls to zero thereby incurring a shutdown cost. In
Figure 4.1 this effect occurs between zero output and output M. The slope of the
tangent over this section will be zero. 20
(a) Cubic (b) Quadratic (c) Linear

Short Run Total Cost $ Short Run Total Cost $ Short Run Total Cost
$

Slope of tangents = SRMC

Small shutdown costs

0 M R K 0 M R K 0 M R K
MWh MWh MWh
Figure 4.1 Alternative short run total cost curves for a plant with sunk startup costs and
small shutdown costs occurring below mingen

4.2 SRMC for an individual plant with sunk startup

costs and small shutdown costs occurring below
mingen
Once the short run total cost curve for a generating facility is determined for a half hourly
trading interval, it is then a matter of finding the rate of change of that curve to determine
the SRMC of the plant for the trading interval. This was shown in Figure 4.1 as the slope
of the tangent of the short run total cost curve. Figure 4.2 converts this detail into SRMC
curves, with diagrams (a), (b) and (c) derived from the correspondingly labelled diagrams
in Figure 4.1.

19
The economic cost of operating at output levels close to a plants thermodynamic limit can be thought of in
a probabilistic sense: as the plant is pushed outside its normal range of operation (e.g. beyond R in Figure
4.1) the likelihood of very costly failure of equipment becomes increasingly significant, thus resulting in the
exponential increase in the expected costs of generating the relevant electricity (see Section 5). There will,
therefore, be some upper bound to short run total cost that represents the worst possible failure of
equipment. This cost will include any lost revenues that result from plant outages.
20
The plant cannot operate in the range of output below mingen, as indicated by the dashed curve.

20 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

1) The SRMC curves do not exist up to the mingen quantity of output (M in Figure
4.2).
2) There is a sharp spike in marginal cost when output is reduced below mingen
(output M in Figure 4.2). 21
3) The SRMC curves can be increasing or decreasing over a range of output, but
eventually get very steep with increasing output as thermodynamic limits are
approached at K (and beyond the manufacturers recommended output rating at
R).
4) Because the short run total costs in Figure 4.1 are non-declining, the slope of the
total cost curve must also be non-declining. Therefore, SRMC in Figure 4.2 must
be of non-negative value.
(a) Quadratic (b) Linear (b) Constant
SRMC SRMC
$ SRMC $ $

0 M R K 0 M R K 0 M R K
MWh MWh MWh
Figure 4.2 Alternative SRMC curves for a plant with sunk startup costs and small
shutdown costs occurring below mingen

4.3 Short run total cost for a plant with sunk startup
costs and large shutdown costs occurring below
mingen
Three theoretical versions of the short run total cost curve for a plant with substantial
shutdown costs relative to the total cost of operating the plant at mingen is depicted in
Figure 4.3. If the facility is forced below mingen during the relevant half hour trading
interval it will be forced to shutdown incurring large costs. Beyond mingen, the plant faces
normal variable costs. The important features of Figure 4.3 are:

1) Because production is not possible below mingen quantity M shutdown costs will
occur if production falls below this level of output.
2) Greater than mingen quantity M the curves are non-declining, eventually becoming
very steep at high levels of output as the thermodynamic limits of the capital in
place are approached at K (and beyond the manufacturers rating for the plant at
R).

21
The first derivative of a vertical increase or decrease in total cost is mathematically undefined. However,
using a difference equation will give the spike in total cost a SRMC value. For example, with C = 1MWh,
where C denotes the change in total cost, SRMC (at just below mingen in Figure 4.1) will equal the
difference between the cost of shutting the plant down and the total cost of operating the plant at mingen. If
the average SRMC up to mingen is then taken, the value over this range will be equal to C / Q , where
Q denotes the change in quantity between zero output and mingen output.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 21
Technical Paper
 Economic Regulation Authority

3) The tangents of the short run total cost curves represent their rate of change and
therefore their SRMC.

(a) Cubic (b) Quadratic (c) Linear

$ Short Run Total Cost $ Short Run Total Cost $ Short Run Total Cost
Slope of tangents = SRMC
Large shutdown costs

0 M R K 0 M R K 0 M R K
MWh MWh MWh

Figure 4.3 Alternative short run total cost curves for a plant with large shutdown costs
occurring below mingen

4.4 SRMC for a plant with sunk startup costs and large
shutdown costs occurring below mingen
As always, the SRMC of a plant is determined by finding the slope of the tangents to the
short run total cost curve for the trading interval. Diagrams (a), (b) and (c) in Figure 4.4
are derived by this process from the correspondingly labelled diagrams in Figure 4.3.
However, because the short run total cost curves are vertical at mingen, SRMC as a first
derivative will be undefined at this quantity: i.e. a vertical line is infinitely steep.21 This can
be thought of as the narrow downward spike in SRMC as illustrated in each of the
diagrams in Figure 4.4. Because this occurs at negative prices, it will only intercept a
downward sloping demand curve at output below M. The important features of Figure 4.4
are:
1) Production is not possible up to mingen output M.
2) Just below mingen output M, SRMC is of negative value due to the large shutdown
cost and is highly inelastic.
3) Beyond mingen output M, a SRMC curve can be increasing or decreasing with
output but eventually gets very steep as thermodynamic limits are approached at
K (in particular, beyond the manufacturers rating for the plant at R).

(a) Quadratic (b) Linear (b) Constant

SRMC SRMC SRMC
$ $ $

0 0 0
R K R K R K
M M M
MWh MWh MWh

Figure 4.4 Alternative SRMC curves for a plant with sunk startup costs and large
shutdown costs occurring below mingen

22 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

4.5 Short run total cost for a plant with avoidable fixed
costs
The theoretical short run total cost curve for a plant with some level of avoidable fixed cost
is depicted in Figure 4.5. As stated above, in the case of electricity generation, the
avoidable fixed cost component for a plant not already operating prior to the trading
interval is essentially a startup cost. Here, particular attention should be paid to avoiding
any double counting of variable costs. For example, only that amount of fuel expended in
startup which is additional to variable fuel use should be counted in the startup costs. The
important features of Figure 4.5 are:
1) If the plant is to produce output in the half hour, it must produce beyond the
mingen level M.
2) The curves are non-declining with output.
3) Beyond the manufacturers rating for the plant at output R, the curve gets very
steep as the thermodynamic limits of the capital in place are approached at output
K.
4) The short run total cost curves comprise variable costs plus the avoidable fixed
costs of startup. 22
5) The tangents of the short run total cost curves represent their rate of change. The
slope of each tangent represents the SRMC of the plant at that level of output.
6) The higher the slope of a tangent line, the higher the rate of change and the higher
the SRMC.

(a) Cubic (b) Quadratic (c) Linear

Short Run Total Cost $ Short Run Total Cost $ Short Run Total Cost
$
Slope of tangents = SRMC

Start up costs

0 M R K 0 M R K 0 9 R K
MWh MWh MWh
Figure 4.5 Alternative short run total cost curves for a plant with avoidable fixed costs

4.6 SRMC for a plant with avoidable fixed costs

The derivation of the SRMC curves for a plant requiring startup in a half hour trading
interval is depicted in Figure 4.6. The important features of Figure 4.6 are:

22
Note: shutdown costs should also be added to short run total costs in the event where a peaking plant is
required to both startup and shutdown within the same trading interval.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 23
Technical Paper
 Economic Regulation Authority

1) The SRMC curves can be increasing or decreasing over a range of output, but
they eventually get very steep with increasing output beyond the manufacturers
rating R as thermodynamic limits are approached at K.
2) Because short run total and short run variable costs are non-declining, the slope of
the short run total cost curve must also be non-declining. Therefore, SRMC must
be of non-negative value.
3) The very first unit of output, produced at mingen (denoted M in Figure 4.6), incurs
large fixed costs as well as variable costs thus producing a sharp spike in marginal
cost.
(a) Quadratic (b) Linear (b) Constant
SRMC SRMC
$ SRMC $ $

0 M R K 0 M R K 0 M R K
MWh MWh MWh

Figure 4.6 Alternative SRMC curves for a plant with avoidable fixed costs

4.7 Short run total cost for a plant with contracted

commitments to output
In some cases a firm may be pre-committed to a contracted level of output over a half
hour interval from a certain plant. The case of an electricity generator that has contractual
commitments to provide steam to nearby industries provides an example. The breach of
such a commitment will incur costs, one form of which may be a financial penalty
(stipulated in the contractual arrangements) that is levied per unit of output dishonoured.
This would result in a cost situation which can be represented by a generating facility
specific, downward sloping, penalty curve.

A theoretical short run total cost curve for a plant with a contracted commitment to output
is depicted in Figure 4.7a as the vertical summation of the plants normal short run total
operating cost curve with the penalty curve that applies for not meeting contracted output.
As stated above, this can result in a downward sloping cost-quantity relationship. Figure
4.7b is similar to Figure 4.7a, but also comprises an avoidable fixed cost component for
plant startup. In both figures, output S represents the contractual commitment to output
for the half hour interval. Note that, while the diagrams apply to the case of a plant with a
U shaped SRMC curve, the vertical summation method can be similarly applied to other
functional forms. The important features of Figure 4.7a and Figure 4.7b are:

1) The short run total cost curve is equal to the total short run operating costs plus
any short run cost of penalties for not meeting output S.
2) Up to quantity S the short run total cost curve decreases with output, beyond
quantity S the short run total cost curve increases with output.
3) The curve gets very steep at high levels of output as the thermodynamic limits of
the capital in place are approached at K.

24 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

4) The tangents of the short run total cost curve represent its rate of change and
therefore its SRMC.

$ Total short run $ $ Short Run

operating costs Total Cost
Total short run penalty
for not meeting output S

Slope of tangents = SRMC

Shutdown costs
0 M R K 0 M S 0 M S RK
MWh MWh MWh
Figure 4.7a Short run total cost curve based on a cubic function for a plant with sunk
startup costs that is subject to penalties for not meeting contracted output

Short Run
$ $ Total Cost
$
Total short run
operating costs
Total short run penalty
for not meeting output S
Slope of tangents = SRMC

Start up costs

0 M R K 0 M S 0 M S RK
MWh MWh MWh
Figure 4.7b Short run total cost curve (based on a cubic function) for a plant with
avoidable fixed costs that is subject to penalties for not meeting contracted output

4.8 SRMC for a plant with contracted commitments to

output
Again, the SRMC of a plant is determined by finding the rate of change of the short run
total cost curve for the trading interval. Because the penalty curve can be downward
sloping, the short run total cost curve can also be downward sloping over the portion of
output where the penalties apply (i.e. up to quantity S in Figure 4.7a, Figure 4.7b, and in
Figure 4.8a and Figure 4.8b). This results in negative SRMC as illustrated in Figure 4.8a
and Figure 4.8b, the important features of which are:
1) SRMC can be negative as a result of the financial penalties incurred through not
meeting contractual commitments at output S.
2) The SRMC curve can be increasing or decreasing over a range of output, but
eventually gets very steep with increasing output as thermodynamic limits are
approached at K beyond the manufacturers rating for the plant at R.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 25
Technical Paper
 Economic Regulation Authority

$ SRMC

M
0 R K MWh
S

Figure 4.8a SRMC curve (based on a U shaped function) for a plant with sunk startup costs
that is subject to penalties for not meeting contracted output

$ SRMC

M
0 R K MWh
S

Figure 4.9b SRMC curve (based on a U shaped function) for a plant with avoidable fixed
costs that is subject to penalties for not meeting contracted output

26 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

5 Cost measurement and estimation methods

This section turns to the issue of measurement and draws upon engineering based
descriptions of electricity supply in economic modelling. Given the short run components
of economic cost identified in Section 2 of this paper, the short run total cost curve for an
individual plant will comprise of:
a) a variable cost component based upon empirical observation of fuel and
operational and maintenance inputs; and
b) if the plant requires startup, an avoidable fixed cost component determined by
empirical measurement of plant startup costs; or
c) a shutdown cost component that applies for a plant already started if that plants
output falls below mingen.

Means by which empirical data can be used to estimate these cost components are
described below. The goal is to construct a short run total cost curve for each plant in a
firms portfolio. Plant-specific SRMC can be derived as the rate of change of the plants
short run total cost curve.

Below is a list of the minimum types of cost data requiring collection and estimation for
each generating facility in the WEM:
1) input-output, average heat rate, or efficiency data;
2) planned and unplanned outage costs;
3) startup costs; and
4) shutdown costs.

In addition to the above, the following data would also improve the estimation of SRMC:
1) Reliability data by plant or plant type with respect to generation load.
2) Half hourly water use by plant with respect to plant output.

Provided an effective metering technology exists, the relationship between an input

variable and electricity production can be estimated via regression analysis. Econometric
techniques allow a mathematical function to be estimated that not only describes the
relationship, but also indicates the degree of confidence that can be attached to the
estimation.

5.1 Determination of variable cost components

5.1.1 Fuel input
In electricity supply, fuel requirements dominate all other components of short run cost.
Therefore, fuel input and fuel efficiency become the primary focus of empirical analyses.
Engineers typically collect input-output data for each plant by measuring the fuel
requirement (in GJ/hr) for given levels of electricity output (in MW) that the plant sends
out. 23 Alternatively, plant efficiency data can be collected from which the input-output

23
The Input-output relationship is the data that is actually measured in the field, Klein, J. (1998), The Use
of Heat Rate Curves in Production Cost Modeling and Market Modeling, Electricity Analysis Office,

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 27
Technical Paper
 Economic Regulation Authority

relationship can be mathematically derived. Both types of data are usually measured at a
number of discrete levels of output for each plant. This is sometimes referred to as point
data, segment data, or valve best point data.

Table 5.1 provides an example of input-output data for the Potrero 3 gas turbine plant in
California, with point 0 corresponding to the origin of the input-output curve, point 1
corresponding to mingen output, points 2 through 4 corresponding to intermediate levels
of output in the operational range of the plant, and point 5 corresponding to the plants
rated capacity. 24 Basic algebra enables easy conversion of the input-output relationship
to average heat rate and/or efficiency measures. Average heat rate is energy input
divided by energy output, whereas efficiency is energy output divided by energy input (i.e.
the inverse of average heat rate) given as a percentage after converting to a common unit
of measurement.

Table 5.1 Input-output and heat rate data for the Potrero 3 gas turbine plant

Output Input-Output Av. Heat Rate Efficiency

% Rated Cap. MW MWh/GJ GJ/MWh %
Point 0 0 0 Undefined Undefined Undefined
Point 1 23 47 543.31 11.56 31.14
Point 2 25 52 590.63 11.36 31.70
Point 3 50 104 1080.81 10.39 34.64
Point 4 80 166 1687.64 10.17 35.41
Point 5 100 207 2135.05 10.31 34.90

Source data: Klein, J. (1998)

An input-output curve can be estimated by fitting a cubic function to the data. This can be
done routinely on statistical or spreadsheet packages. Figure 5.1 depicts the fitted input-
output curve and the relevant equation for the Potrero 3 gas turbine plant using Excel.

2500
y=0 if x = 0
3 2
y = 1E-04x - 0.0318x + 12.689x + 1.714 if x 47
2000
y = Input (GJ/hr)

1500

1000

500

0
0 50 100 150 200 250
x = Output (MW)

Figure 5.1 Input-output curve for the Potrero 3 gas turbine

California Energy Commission. See also Schweppe, F., Caramanis, M., Tabors, R. & Bohn, R. (1988),
Spot Pricing of Electricity, Kluwer Academic Publishers, Boston, p. 283-285.
24
Data sourced from: Klein, J. (1998), The Use of Heat Rate Curves in Production Cost Modeling and Market
Modeling, Electricity Analysis Office, California Energy Commission.

28 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

If a constant load is assumed over a half hour period, a fitted input-output curve can be
converted to a total fuel consumption curve by multiplying both sides of the fitted equation
by hours and dividing by two. This is then multiplied by the opportunity cost of fuel to
produce a total fuel cost curve for a trading interval. For example, assuming an
opportunity cost of fuel of $9.00/GJ results in the half hourly fuel cost curve for the Potrero
3 gas turbine depicted in Figure 5.2.

12000
y=0 if x = 0
3 2
10000 y = 0.0035x - 0.5715x + 114.2x + 7.7131 if x 23.5

8000
y = Cost ($)

6000

4000

2000

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.2 - Half hourly fuel cost curve for the Potrero 3 gas turbine

For some plants, the actual relationship between fuel energy input and electrical energy
output may not be smooth. For example, in the case of a coal fired plant with multiple
steam valves that are opened sequentially, a more exact representation of the input-
output representation may be to refine a quadratic or cubic function by a sine function. 25
Similarly, the input-output curve for a plant with multiple operating modes such as a
combined cycle plant may be more in the form of a piecewise function rather than a fitted
cubic function. Both cases are depicted in Figure 5.3.

Plant with multiple steam valves Plant with multiple operating modes
Input (GJ)
Input (GJ)

sine-modified piecewise function

function
e
d cubic
c function
b
a valve opening points designated a to e
0 0
Output (MWh) Output (MWh)

Sources: Kim et al, (2002) and Wood, A. & Wollenberg, B. (1996)

Figure 5.3 - Possible inputoutput curve of plant with multiple steam valves (e.g. large
thermal generators) and plant with multiple operating modes (e.g. combined
cycle generators).

However, given that the measurement of input-output data for a plant is taken at a limited
number of points, it is unlikely that the requisite level of detail would be available to
improve the input-output representation. Therefore, in the context of portfolio SRMC
25
Kim, J. Shin, D. Park. J, & Singh, C. Atavistic genetic algorithm for economic dispatch with valve point
effect, Electric Power Systems Research, 62, 2002, pp 201-207.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 29
Technical Paper
 Economic Regulation Authority

estimation, an approximation of these complex curves might be made by use of a simpler

function, such as a quadratic function, a cubic function, or quadratic functions over two
ranges of output. Alternatively, more detailed input-output data could be sought from the
generating firm and the problem solved numerically on a spreadsheet.

5.1.2 Operational and maintenance costs: Planned outages

Where output causes a costly deterioration of equipment, wear and tear can be thought of
as a productive input, and thus can be described by an input-output curve similar to that of
fuel. Equipment reliability is often the major focus of engineering based studies.
However, such data is usually not of a form that can be readily placed into an economic
short run price-quantity framework, except by averaging the data. There is, for example,
no obvious way by which to conduct a regression analysis of wear and tear against
equipment load. Therefore, in the case of normal wear and tear of plant and equipment
covered by scheduled maintenance, a practical approach is to divide the cost of a
scheduled maintenance service by the total amount of MWh produced by the plant
between such services. This averaging can then be adjusted by the time value of money:
the further away the scheduled maintenance is in time, the more discounted will be the
per MWh cost of wear and tear in the current trading interval. Note that this method
assumes a linear relationship between load and wear and tear on plant and equipment.

For example, during the period between scheduled maintenances a plant produces
1,500MWh of electricity. If the total cost of a scheduled maintenance, in net present value
terms, is $7,080, the average operational and maintenance cost for the current period
would be:

7,080
= 4.73 ,
1,500

or $4.73 per MWh. Figure 5.4 transposes these operational and maintenance cost
assumptions onto the Poterero 3 gas turbine, with the points marked corresponding to half
hourly operation at the mingen and rated capacity outputs of the plant.

600
y=0 if x = 0
500 y = 4.73x if x 23.5

400
y = Cost ($)

300

200

100

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.4 Assumed half hourly operational and maintenance costs for the Potrero 3 gas
turbine

30 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

5.1.3 Operational and maintenance costs: Unplanned outages

The extent to which a statistical relationship exists between the probability of an
unplanned outage event and a plants load is relevant to the modelling of short run
variable cost. Equipment failure leading to an unplanned outage can result in substantial
financial loss, both to a firm and to society as a whole. From the point of view of the firm,
such costs include not only direct expenditure on unscheduled maintenance, but also any
loss of net revenue that would have been generated by the plant if it were still in service,
along with any regulatory penalties imposed. Therefore, the average value of an
unplanned outage is likely to be very large.

Of particular interest is whether operating near or beyond a plants rated capacity leads to
an increased rate of equipment failure. If evidence can be found of such a relationship,
the average cost of an outage event can be multiplied by the probability of it occurring to
develop an expected cost function. In other words, correlation between the probability of
equipment failure and plant output will see unplanned outage expenses treated as
variable costs. Conversely, if a statistical relationship between output and the probability
of failure cannot be established, unplanned outage expenses should be added to the short
run total cost function as an avoidable fixed cost component.

Figure 5.5 suggests one feasible probability relationship. It seems apparent that, given
plants can and do occasionally operate beyond their rated capacity, there must be some
reliability or other variable cost related reason why firms tend to avoid operating a plant at
this level. Thus, one might postulate a rapidly rising probability of plant failure near and
beyond rated capacity.

1
y=0 if x = 0
y = Probability of unplanned

y = 1.25E-4 if 0.23 < x < 1.0

0.8 3
y = 1000(x - .995) if 1.0 x 1.095
y=1 if x > 1.095
0.6
outage

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2
x = Output (proportion of rated capacity)

Figure 5.5 Hypothetical probability of outage for a gas turbine plant

An expected value of failure could be applied to any plant where such a relationship is
found. For example, using an assumed cost of $35,000 for an unplanned outage, simple
algebra converts the cumulative distribution function of Figure 5.5 to the variable factor
cost function of Figure 5.6. Note that in Figure 5.5 the bottom axis refers to the proportion
of rated capacity whereas the bottom axis of Figure 5.6 refers to actual output in MWh,
after applying the assumptions of Figure 5.5 to the Poterero 3 gas turbine.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 31
Technical Paper
 Economic Regulation Authority

35000

y = Expected cost of unplanned

y=0 if x = 0
30000 y = 4.375 if 23.5 x < 100
3
25000 y = 3.5E7 (x /103.5) - 0.995) if 100 x 112
y = 35000 if x > 112

outage ($)
20000

15000

10000

5000

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.6 Hypothetical expected cost of unplanned outage applied to the Poterero 3 gas
turbine plant

It must be recognised that, in practice, some instances of plant failure, for example those
attributable to human error, will be associated more with startup than with high levels of
output. Therefore, before the correct relationship can be established, time series and
cross sectional data will need to be sought indicating a plants load immediately prior to an
unplanned outage event. Depending on the number of unplanned outages per year, it
may take a considerable period of time to collect enough observations for a statistical
model to be developed. Any relationship is likely to be first detected in aggregated,
market-wide data of plant load as a percentage of rated capacity. This would essentially
develop an average value that could be applied to all the plants in a firms portfolio.

As such, until a correlation with output can be demonstrated, it is proposed that costs
relating to unplanned outages be treated as correlated with startup. In other words, until
an output based cumulative distribution function is developed, expected costs associated
with the risk of plant failure should be treated as a fixed cost component as opposed to a
variable cost component. That is, the expected cost of an unplanned outage should be
included as part of startup costs. This would entail the averaging of plant failure cost
across total startups per plant over a time period. In this case, however, it would not be
appropriate to include regulator imposed financial penalties in the determination of
expected cost, because these would then translate to higher STEM prices, a result which
is contrary to the intention of the penalties: i.e. the enhancement of risk avoidance
incentives.

5.1.4 Water input

Water input into fossil fuel based electricity generation is, generally, not considered a
significant cost component and therefore can be ignored without much impact on SRMC
calculations. 26 However, where required, a similar framework to fuel input or operational
and maintenance input can be used to determine the relationship between volumetric
water consumption and electricity output (by plant and half hour trading interval). Clearly,
water metering is a well established technology and therefore, as with fuel input data, one
would expect water flow data to be collected by a firm. This, however, may not be
accurate to the half hour, in which case averaging will need to be conducted with
reference to longer timeframes.

26
See, for example, figures given in: National Energy Technology Laboratory, Power Plant Water Usage and
Loss Study, August 2005.

32 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Once the required amount of water input to produce a given level of electricity output is
known, that amount can then be multiplied by the current price of water (i.e. its opportunity
cost) to determine the relationship between the factor cost and electricity output. Studies
on water use by electricity plants often report the results as an average value in gallons or
litres per MWh. This constrains estimation to a linear function and thus may conceal an
economically important non-linear input-output relationship. Therefore, it would be better
to collect actual field data from generating firms, rather than to rely on aggregated data.

5.1.5 Other variable costs

As regulation in the area of pollution control permits, renewable energy certificates and
carbon taxes continues to evolve, the costs associated with such instruments are likely to
increase in significance. Where appropriate, these should be included in economic cost
calculations. However, from the regulators point of view, it should be noted that any
penalties, actual or expected, associated with not meeting regulatory obligations should
not be allowed to be passed on to consumers. That is, regulatory penalties should not be
included in legislated SRMC calculations.

5.2 Determination of non-variable cost components

5.2.1 Startup costs
Startup costs include the opportunity costs of fuel, water, internal power, additional labour
and lost asset value directly attributable to the startup, including the expected costs
associated with the risk of plant failure (if it is not deemed to be correlated with output).
Startup is defined as the period from zero output to mingen. Therefore, to avoid double
counting, the variable cost component between zero output and mingen output for fuel,
water, and wear and tear on plant and equipment is not included in the startup cost
calculation. In other words, only those costs additional to those associated with variable
output should be included in the startup cost calculation. This is because the short run
total factor cost of fuel, water and operation and maintenance already include such costs,
as is evident in the relevant diagrams in Section 4 of this paper.

Startup costs are placed in three categories depending upon the period elapsed since the
plants previous shutdown: hot starts, warm starts, and cold starts. The decision as to
which of these should be used in short run total cost calculations will reflect the firms unit
commitment plan, thus relating to the reasonable expectations allowance in clause 6.6.3
of the Market Rules.

Startup costs are described by a constant function, as in the hypothetical example of

Figure 5.7 (assumed for the Poterero 3 gas turbine), where the marked points correspond
to the plants mingen and rated capacity output levels. 123456

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 33
Technical Paper
 Economic Regulation Authority

5000

4000 y=0 if x = 0
y = 2600 if x 23.5

y =Cost ($) 3000

2000

1000

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.7 Assumed startup costs for the Poterero 3 gas turbine

5.2.2 Shutdown costs

For most plants, the direct costs of shutdown are comprised of fuel and internal power
components and are not significant. However, the decision to shutdown and inevitably
start the plant up again must be weighed against the alternative cost of operating a plant
at mingen over the intervening period (minus any avoided cost of variable base-load
generation). Shutdown costs are likely to be significant in the case of coal fired power
stations. They are avoided only if the plants output remains at or above mingen. As with
startup costs, only those costs additional to those associated with variable output should
be included in the shutdown cost calculation. To prevent double counting, the variable
costs associated with the output a plant produces as its generation is gradually allowed to
slow down to zero are not added to shutdown costs. 27 Figure 5.8 provides a hypothetical
example of shutdown costs assumed for the Poterero 3 gas turbine, where the marked
point corresponds to zero output.

5000

4000
y = Cost ($)

3000
y = 1400 if x = 0
2000 y=0 if x 23.5

1000

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.8 Assumed shutdown costs for the Poterero 3 gas turbine

27
Where a facility such as a base-load plant (which typically operates as a facility with sunk startup costs in
every trading interval) needs to be shutdown periodically for maintenance, the cost of its shutdown and
startup should not be considered a non-variable cost. These are costs that, while paid in a lumpy fashion,
accrue due to wear and tear on plant and equipment and are therefore variable costs. In this case it is
correct to average the cost over the intervening period and then adjust this average for the time value of
money (see Section 5.3.2 above).

34 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

5.3 Derivation of a plants short run total cost curve

5.3.1 Plants with sunk startup costs
For a plant already started prior to the half hour trading interval, the short run total cost
curve is calculated by vertical summation of all the short run total factor costs (i.e. variable
fuel costs, variable water costs, and variable operational and maintenance costs) in
excess of mingen, plus the avoidable non-variable cost of shutdown. This is
accomplished via the following two step procedure:
1. The half hourly fuel cost curve, the half hourly operational and maintenance
cost curve, and any other relevant half hourly factor cost curves (e.g. water)
are added together to create an interim variable cost curve for the plant.
2. Shutdown costs are then added to the curve.

5.3.1.1 Example based on Poterero 3 gas turbine heat rates

Using the assumptions given in Sections 5.1 and 5.2 for the Poterero 3 gas turbine, the
steps for generating the plants short run total cost curve (with sunk startup costs) are
preformed as follows:

Step 1
An interim curve is determined by the vertical summation of Figure 5.2 and Figure 5.4.
The result is depicted in Figure 5.9.

12000
y=0 if x = 0
3 2
10000 y = 0.0035x - 0.5715x + 118.93x + 7.7131 if x 23.5

8000
y = Cost ($)

6000

4000

2000

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.9 Interim variable cost curve (based on Poterero 3 gas turbine heat rates)

Step 2
The shutdown costs appropriate for that half hour interval (given reasonable expectations
of unit commitment) are then added to the curve. Here it is assumed the shutdown cost is
$1,400, as given in Figure 5.8. The result, depicted in Figure 5.10, is the plants short run
total cost curve for the half hour.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 35
Technical Paper
 Economic Regulation Authority

12000
y = 14000 if x = 0
10000 3 2
y = 0.0035x - 0.5715x + 118.93x + 7.7131 if x 23.5

y = Cost ($) 8000

6000

4000

2000

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.10 Short run total cost curve (based on Poterero 3 gas turbine heat rates) for a
plant with sunk startup costs

5.3.2 Plants requiring startup to produce in a half hour trading

interval
For a plant not started prior to the half hour trading interval, the short run total cost curve
is calculated by vertical summation of all the short run total factor costs (i.e. variable fuel
costs, variable water costs, and variable operational and maintenance costs) in excess of
mingen, plus the avoidable fixed cost of startup. This is accomplished via a two step
procedure similar to that given in Section 5.3.1 above:

1. The half hourly fuel cost curve, the half hourly operational and maintenance
cost curve, and any other relevant half hourly factor cost curves (e.g. water)
are added together to create an interim variable cost curve for the plant.
2. Startup costs are then added to the curve.

Example based on Poterero 3 gas turbine heat rates

Using the assumptions given in Sections 5.1 and 5.2 for the Poterero 3 gas turbine, the
steps for generating the plants short run total cost curve (with sunk startup costs) are
preformed as follows:

Step 1
This step is identical to that outlined in Section 5.3.1 above.

Step 2

The startup costs appropriate for that half hour interval (given reasonable expectations of
unit commitment in relation to a cold, warm or hot start) are then added to the curve. Here
it is assumed the startup cost is $2,600, as given in Figure 5.7. The result, depicted in
Figure 5.11, is the short run total cost curve.

36 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

14000 y = 5154.8087 if x = 0
3 2
y = 0.0035x - 0.5715x + 118.93x + 2607.7131 if x 23.5
12000

10000
y = Cost ($)
8000

6000

4000

2000

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.11 Short run total cost curve (based on Poterero 3 gas turbine heat rates) for a
plant requiring startup

5.4 Derivation of a plants SRMC curve

The optimisation process given in Section 7 and Appendix 1 below, does not require the
derivation of plant-specific SRMC. Rather, it optimises plant-specific short run total cost to
generate an optimal portfolio short run total cost curve, and then derives the optimal
portfolio SRMC from this curve. However, if desired, a plant-specific SRMC can be
produced by taking the first derivative of the plant level short run total cost curve. In the
case of the hypothetical plant given above in Sections 5.2 and 5.3 (i.e. based on Poterero
3 gas turbine heat rates at a $9.00/GJ gas price), the derivation of plant-specific SRMC
results in the curve depicted in Figure 5.12 below.

2
120 y = 0.0105x - 1.143x + 118.93 if x 23.5

100

80
y = Cost ($)

60

40

20

0
0 20 40 60 80 100 120
x = Output (MWh)

Figure 5.12 - SRMC curve (based on Poterero 3 gas turbine heat rates)

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 37
Technical Paper
 Economic Regulation Authority

6 Line losses
The spatial location of a plant relative to load and the physical nature and topography of
transmission infrastructure will affect the plants effectiveness in supplying electricity to the
market. This is because the transmission of electricity over distance is subject to loss. In
practice, a relatively efficient plant that is distant from consumers may be more expensive
at meeting demand when compared with a relatively inefficient plant that is located close
to consumers. 28 Therefore, the quantity of electricity sent down a transmission line must
be adjusted by a loss calculation before the optimisation process described in Section 7
and Appendix 1 below is applied.

6.1 Line losses and the SRMC optimisation problem

In a general sense, there are three main determinants of transmission loss in an electricity
supply system. These are outlined below.

1) Losses are proportional to line resistance. The line resistance has a constant
value in the short run as determined by ambient conditions and the physical nature
of a transmission line (a major aspect of this being the distance between
generation and load). Therefore, a generating firm must take line resistance as
given for a particular half hour trading interval.

2) Losses are inversely proportional to the square of line voltage. Line voltage is a
parameter determined by system management requirements. While it may be
influenced in the short run by actions of generating firms, for all practical purposes
each firm must take line voltage as given for a half hour trading interval.

3) Losses are proportional to the square of the power flow along a line. Power flow is
a variable parameter in the short run and is determined by the magnitude of load
during a trading interval as well as the relative location of that load vis--vis
generation. Hence, if the network distribution of load for each trading interval were
apparent and predictable, the choice of power flow would be within the complete
control of the entity responsible for plant dispatch in a day ahead market.

It can be seen from the above that the only determinant relevant to the short run economic
decision is power flow. This is because, from a generating firms point of view in a short
run market, net power flow is variable at the portfolio level to the extent that the share of
portfolio output can be allocated to different plants within the half hour. Therefore, power
flow is a lever upon which a generating firm can optimise its resource allocation in a half
hourly price-quantity framework.

Given the above, plant optimisation is complicated by the fact that, because increased
power flow results in a greater-than-proportional increase in line losses, the total loss from
a combination of plants will be greater than the sum of its parts. Figure 6.1 provides a
demonstration. Assume that generators A, B, C and D are identical plants with similarly
identical short run total cost curves, but that while generators A and B are distant from the
380MW load at the reference node (i.e. bus 2) and therefore subject to loss, generators C

28
Good overviews of line loss economics are provided by: Stoft, S. (2002), Power System Economics, IEEE
Press, New Jersey, pp. 415-423, and by; Wood, A. & Wollenberg, B. (1996), Power Generation, Operation
and Control, 2nd. ed., Wiley-Interscience, New York.

38 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

and D are not. Losses along the line are equal to 0.0002 P{2A, B} , where P{ A, B} is the
combined quantity of power sent out by generators A and B.

Generator A Generator C
Capacity = 207MW Capacity = 207MW

Losses = 0.0002 P{2A, B}

Bus 1 Bus 2

Load = 380MW

Generator B Generator D
Capacity = 207MW Capacity = 207MW

Figure 6.1 A simple generation and transmission system

Table 6.1 Two plant optimisation with line losses

Plant combination to meet 190MWh demand at bus 2 in Figure 6.2

Plants C and D Plants A and C Plants A and B

Optimal portfolio SRMC $105.53 $111.57 $136.13

Optimal plant share of Plant C: 95MWh Plant A: 88.5MWh Plant A: 95MWh
energy arriving at Plant D: 95MWh Plant C: 101.5MWh Plant B: 95MWh
bus 2
Optimal plant share of Plant C: 95MWh Plant A: 92MWh Plant A: 103.5MWh
energy sent out Plant D: 95MWh Plant C: 101.5MWh Plant B: 103.5MWh
Total energy generated 190MWh 193.5MWh 207MWh
Total energy generated 0MWh 93MWh 207MWh
at bus 1
Total losses 0MWh 3.5MWh 17MWh

Using the Poterero 3 gas turbine given in Section 5 above as the basis for the four
identical plants in the system described by Figure 6.1, gives the results summarised in
Table 6.1. 29 Each of the optimal two plant combinations are described after taking line
losses into account. By assuming demand is constant over the trading interval, the
optimisation problem is able to be converted to the usual price-energy relationship, i.e.
where 380MW of power is treated as 190MWh of energy demand for the half hour.

29
In each case, the plant is modelled as already started prior to the half hour interval i.e. shutdown costs
are included in short run total cost, but startup costs are not.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 39
Technical Paper
 Economic Regulation Authority

The starting point is to consider the situation where generators C and D are the only
plants available for the trading interval. In this case the 190MWh of demand is met
without any line loss. Each plant produces 95MWh of energy to meet the load and the
optimisation method can be conducted without modification to the plants short run total
cost curves. If, however, only generators A and C, (or alternatively, any two plant
combination comprising a plant located at bus 1 and a plant located at bus 2) are
available for the half hour trading interval, line losses will need to be taken into account for
the energy sent out at bus 1. Figure 6.2 describes how, in these cases, the correct
optimisation method first requires the short run total cost curve for generator A or B to be
shifted back by an amount equal to any line losses. This enables the energy arriving at
bus 2 to be evaluated in an apples with apples comparison of the plants total cost
curves. Not taking line losses into account will produce erroneous results.

Generator A or B Generator C or D
$ $
Total cost of energy Total cost of energy
sent out from Bus 1 produced at Bus 2

Total cost of energy

arriving at Bus 2

Shutdown cost Shutdown cost

0 0
MWh MWh

Figure 6.2 The backward shift in plant level short run total cost to account for line loss.
To elaborate, as Table 1 indicates, taking line losses into account when generators A and
C are combined results in an optimal plant share of energy sent out compromising
92MWh from plant A and 101.5MWh from plant C, at an optimal portfolio SRMC of
$111.57. However, not taking line losses into account will lead to the same portfolio
SRMC estimation as that of combining generators C and D, i.e. $105.53, an error of $6.04
or about 5.7 per cent. Moreover, line losses of 3.5MWh of energy will be unaccounted for
in the calculation. Such errors are greatly magnified as more power is sent down the line
and losses increase at a quadratic rate. If, for example, only generators A and B (as
described in Figure 6.1) are available to meet system demand, Table 6.1 shows that
17MWh of energy will be unaccounted for and the SRMC estimation error will be $30.60
or about 29 per cent.

6.2 Line loss factors in the WEM

It should be clear from the above discussion that an accurate consideration of line losses
is critical in determining the optimal short run economic dispatch of plant. 30 However,
under the previous market structure of a vertically integrated monopoly there existed no
need to allocate the financial responsibility for line losses among competing economic
agents. Therefore, methods for determining the financial allocation of line losses are,
worldwide, a relatively recent development, having arisen under industry deregulation. All

30
Similarly, the optimal location of transmission and generation investment in the long run requires an
accurate assessment of line losses.

40 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

such accounting methods have entailed a difficult trade-off between accuracy and
tractability, given the dynamic spatial and temporal irregularity between load and
generation, the inability to physically separate one generators transmission loss from
anothers when sharing a line, and the nonlinearity of the loss function. A universal
approach to tackling the allocation problem is yet to emerge. Given the large amounts of
money involved, the particular method chosen for a given market has often invoked
debate.

The approach taken in the WEM is to collect yearly load and generation data at each node
weighted by half hour and to compress this data into a single annual metric, known as a
loss factor, for each node. The loss factor is an annualised static number that attempts to
compare the tendency for loss at any given node in the network relative to the systems
reference node (defined in the market rules as the Muja 330 bus-bar). A loss factor of
1.00 suggests that, on average, the node in question is equally effective at meeting
system demand compared to the reference node. Conversely, a loss factor that differs
from 1.00 indicates that the average network distribution of supply and demand causes
the node to be, on average, either more or less effective at meeting system demand
compared to the reference node. In practice, the meter reading for each generator is
multiplied by the loss factor to give an adjusted output, and it is this adjusted output
against which payment is made for electricity generation. For example, a loss factor of
0.95 would suggest the average effective output at the node is 5 per cent less than that at
the reference node, whereas a loss factor of 1.05 would suggest the average effective
output at the node is 5 per cent more than that at the reference node. Loss factors for
each node are posted on the IMOs web site, as are the rules pertaining to their
determination. 31

It is clear that the loss factor methodology was not designed for the purposes of half
hourly optimisation between generation facilities. Hence, given that the multiplication of a
loss factor against nodal output is a linear operation, it must be recognised that its use in a
short run optimisation framework will produce a nonlinear error that increases the further
away nodal output diverges from its average value. By way of example, assume the line
loss factor determined for Bus 1 in Figure 6.1 is deemed to be 0.96 based on an average
half hourly output over the year of 100MWh generated at that node. For no error to occur
under these circumstances, generation at bus 1 for the half hour would have to be exactly
100MWh. If, however, 190MWh were produced at bus 1 applying the loss factor of 0.96
would suggest losses of 7.6MWh. Yet, as Table 6.1 shows, this figure is incorrect. More
power sent down the line results in more than proportional losses: a nonlinear
relationship. Actual losses would be 17MWh, 223.7 per cent greater than the amount
estimated by use of the linearised loss factor.

Therefore, errors should be expected if the loss factor method is used for SRMC
optimisation. However, given the immense complexity of the supply and demand
relationship in instantaneous network distribution terms, it is highly unlikely that, in the
context of day ahead predictability, line loss estimation error will ever be fully reduced.
The dynamics of power flow throughout a power system can change moment by moment
with locational shifts in demand that are very difficult to anticipate. As such, given that the
requirements of clause 6.6.3 of the market rules pertain to the reasonable expectations
of SRMC in a day ahead market, the use of line loss factors might be tolerable at present.
However, in time, a truly marginal, nonlinear, line loss metric could be developed by cost
minimising firms and/or government agencies. It should be noted that errors caused by
the use of line loss factors for short run optimisation will tend to be quadratic with output.

31
Wholesale Electricity Market Procedure (2006), Market Procedure for Determining Loss Factors, available
IMO website: www.imowa.com.au/Attachments/Loss%20Factor%20Procedure.pdf

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 41
Technical Paper
 Economic Regulation Authority

Development of a metric that causes such errors to be small and linear with output would
be a clear improvement. 32

32
The author may explore this possibility at a future date.

42 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

7 Portfolio short run cost over a half hour time

frame
In the short run, a typical electricity generation firm will have a portfolio of plants to utilise
in the production of electricity. Each plant in such a portfolio is likely to differ significantly
in its cost characteristics, utilising different fuels at different prices and at varying levels of
thermal efficiency which itself depends on factors such as plant design, vintage and level
of maintenance. Therefore, a firm faces a cost minimising decision directly related to how
much of its portfolio output should be shared between each facility. An optimal mix of
output from the various plants will minimise the portfolio short run total cost in a half hour
trading interval. In other words, to maximise profit over a half hour interval, electricity
should be sourced in such a way that each plant is producing a level of output that
minimises the portfolio short run total cost of meeting demand. This results in an optimal
portfolio total cost curve. The rate of change of this curve determines the optimal portfolio
SRMC curve.

7.1 Short run total cost for a portfolio of plants with

sunk startup costs
In considering the construction of the optimal portfolio short run total cost curve, the
simplest case is where all available plants have already been started prior to a half-hour
trading interval. Under sunk startup cost assumptions, the following sections will describe
how the optimal proportion of each plants contribution to portfolio output will vary as
portfolio output varies.

7.1.1 Example 1: Linearly increasing total cost technology,

portfolio short run total cost determined by a SRMC
order of merit
Figure 7.1 depicts a portfolio of plants that have already been started prior to the relevant
half hourly trading interval, each of which will be required to generate electricity in
subsequent trading intervals. The portfolio consists of three generating facilities with
linearly increasing short run total cost functions: one low cost, one mid cost and one high
cost plant.

Plant A: low cost Plant Plant B: mid cost Plant C: high cost
$ $ $

a b c

Shutdown costs
0 0 0
91 MWh 26 MWh 16 MWh

Figure 7.1 A portfolio of three linear short run total cost plants with sunk startup costs and
equalised shutdown and mingen costs between plants

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 43
Technical Paper
 Economic Regulation Authority

So as to enable a relatively less complicated introduction to the calculation of optimal

portfolio short run total cost, shutdown costs in Figure 7.1 are treated as being: (a) equal
for each plant, and; (b) equal to the total cost at mingen for each plant. 33 The treatment of
differing shutdown cost assumptions and differing total costs at mingen between plants
will be considered in Section 7.1.6.

The linear cost assumption at the plant level in Figure 7.1 combined with an assumption of
equalised shutdown costs reduces the determination of the optimal portfolio short run total
cost curve to a relatively trivial problem. This is because optimisation can be achieved by
building an aggregated short-run total cost curve from Figure 7.1 according to a strict,
plant-specific SRMC order of merit. In doing so, ensuing quantities of output (moving to
the right for each plant) are sourced so as to cause the total cost of the combined
electricity generation (i.e. the portfolio short run total cost curve) to rise at the slowest
possible rate. In other words, electricity is sourced from the plant with the lowest sloped
short run total cost curve right up to its maximum level of output, before moving on to the
next lowest SRMC plant. In the example of Figure 7.1 it is assumed that, for any given
level of output, the SRMC of plant A is less than the SRMC of plant B, which in turn is less
than the SRMC of plant C. Therefore, a cost minimising firm would set a rule that,
whenever possible, variable output should be sourced from plant A first, plant B second,
and plant C last.

Table 7.1 describes the constraints affecting the optimal sourcing of portfolio output from
the three available plants depicted in Figure 7.1 for various ranges of portfolio output,
given the aforementioned assumptions. For any level of portfolio output left of the pooled
mingen of all three plants (i.e. outputs 91 + 26 + 16 = 133 from Figure 7.1), the decision
must incorporate the shutdown cost of at least one plant. Conversely, if portfolio output is
greater than the pooled mingen of all three plants, no plant is required to shutdown,
although the option remains available.

The optimal portfolio short run total cost curve derived from this optimisation process is
depicted in Figure 7.2. The diagram shows a saw tooth pattern for the curve for a portfolio
level of output left of the pooled mingen of all three plants, reflecting differing optimal
combinations of variable output and shutdown from each plant as mingen constraints
come into play. Right of the pooled level of mingen, energy is sourced incrementally in a
SRMC order of merit: first from plant A up to its maximum capacity, then from plant B up
to maximum capacity, and finally from plant C up to its maximum capacity.

33
The total costs for the three plants, A, B and C, were modelled as follows:
2,000 if Q A = 0
CA = ,
20Q A if 91 Q A 340
2,000 if QB < 26
CB = , and
75QB if 26 QB 90
2,000 if QC = 0
CC = ,
150QC if 16 QC 50
where Ci denotes the total cost of plant i, and Qi denotes the output of plant i, {i A, B, C} .

44 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Table 7.1 The optimal choice of plant operation and shutdown for the portfolio of plants
given in Figure 7.1
Discrete
portfolio
output Determining factor for Optimal decision
range bottom of output range Run constraints Shutdown vs. Run
MWh Plant Decision
0 - 15 Zero output. All three plants must shutdown. A Shutdown
B Shutdown
C Shutdown

16 - 25 Mingen plant C. Plants A and B must shutdown. A Shutdown

B Shutdown
C Run

26 - 41 Mingen plant B. Plant A must shutdown and if plant B A Shutdown

runs, plant C must shutdown.
B Run
C Shutdown

42 - 90 Mingens plants A + B. Plant A must shutdown. A Shutdown

B Run
C Run

91 - 106 Mingen plant A. If plant A runs, plants B and C must A Run

shutdown.
B Shutdown
C Shutdown

107 - 116 Mingens plants A + C. If plant A runs, plant B must shutdown. A Run
B Shutdown
C Run

117 - 132 Mingens plants A + B. If plant A runs, plant C must shutdown. A Run
B Run
C Shutdown

133 + Mingens plants A + B + C No run constraints. A Run

(pooled mingen of the three
plants). B Run
C Run

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 45
Technical Paper
 Economic Regulation Authority

$ Optimal Portfolio Short Run Total Cost Curve

shutdown costs incurred for

one or more plants below
pooled mingen
Plant C

Plant B

Plant A
Lower envelope of all
feasible cost paths

Pooled mingen of
the three plants

0 MWh

Figure 7.2 Optimal portfolio short run total cost curve derived by combining linearly
increasing total cost technology plants with equalised shutdown and mingen costs

7.1.2 Optimal portfolio SRMC

Once the short run total cost curve for a portfolio of plants is determined for a half hour
trading interval, it is then a matter of finding the rate of change of that curve to determine
the SRMC of the portfolio for the trading interval. This is defined in the same way for a
portfolio of plants as it is for a single plant, i.e. by the slope of the tangents to the short run
total cost curve.

$ Optimal Portfolio SRMC

Note: Quantity band average Plant C

marginal cost measure used
below pooled mingen
Plant B
Plant A

0
M MWh

Figure 7.3 Optimal portfolio SRMC comprised of plants with a linearly increasing total cost
technology and equalised shutdown and mingen costs between plants
Figure 7.3 depicts the SRMC curve for the portfolio of plants described in Figure 7.1.
Portfolio output M corresponds to the sum of all three plants mingen levels. The results
are intuitive. With each plant exhibiting a flat SRMC curve, the portfolio SRMC curve
becomes a step function, with each plant stacked in its SRMC order of merit. Note that,
below the pooled mingen level, where shutdown costs cause sudden jumps in the short

46 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

run total cost curve, SRMC is treated as the change in price over the change in quantity
from output zero to output M. 34

7.1.3 Example 2: Cubic based short run total cost technology,

the (sub-optimal) SRMC order of merit approach
As with Section 7.1.1 above, Figure 7.4 depicts a portfolio of plants that have already
been started prior to the relevant half hourly trading interval. However, this time the
portfolio consists of generating facilities with short run total cost curves based on a cubic
functional form. In this case, the method applied in Section 7.1.1 right of the pooled
mingen level of output of all three plants (i.e. building an aggregated short-run total cost
curve from Figure 7.4 according to a strict SRMC order of merit) produces a suboptimal
curve.

Plant A: low cost Plant B: mid cost Plant C: high cost

$ $ $

h
C1
A2 e B2 f
A1 b B1 d

a c g

Shutdown costs
0 0 0
a0 a1 a2 b0 b1 c0 c2
b2 c1
MWh a3 MWh b3 MWh

Figure 7.4 A portfolio of three plants exhibiting a cubic short run total cost technology
with sunk startup costs and equalised shutdown costs between plants

To demonstrate: ensuing quantities of output (moving to the right for each plant) are again
sourced so as to cause the total cost of the combined electricity generation (i.e. the
portfolio short run total cost curve) to rise at the slowest possible rate. 35 For example,
consider portfolio output right of the pooled level of mingen for the portfolio of plants in
Figure 7.4 (i.e. a1+b1+c1). Because the slope of the tangent at point a is less than the
slope of the tangent at point c or g, plant A is first in SRMC order of merit. This is so up
until output a2, where the slope of the tangent at point b is equal to the slope of the
tangent at point c. Here the incremental cost of electricity from plant A becomes higher
than that from plant B. This sends a signal that fuel should now be burned in plant B in
lieu of plant A to meet the increasing load. This remains the case up until output b2 where
it is better to increase plant B and plant A fuel input in staggered unison (corresponding to
SRMC merit) as the portfolios output increases. Once outputs b3 and a3 are reached

C
34
Using a difference equation of C& = with even increments of Q = 1 results in several large and
Q
negative sharp spikes in portfolio SRMC left of M in Figure 7.3 and Figure 7.8, and left of m in Figure 7.11.
M
C
However, as a general rule,
0
C& M =
Q
if Q = M . While this corresponds to the discrete

estimation of SRMC (see Box 1.1), for large Q , it should actually be regarded as a quantity band
average marginal cost measure.
35
The same simplifying assumptions of Section 7.1.2 remain: startup costs are considered sunk, each plants
mingen levels are the same, and the shutdown costs for each plant are equal to the total cost at mingen.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 47
Technical Paper
 Economic Regulation Authority

however, the slopes of total cost for both plant A and plant B become greater than the
slope of total cost for plant C at point g. It is then cheaper to supply additional quantities
of electricity from plant C only. Eventually, when point h is reached, the slowest possible
rate of total cost increase with respect to quantity is obtained by ramping up output from
all three facilities in staggered SRMC merit. In this final stage each plant is nearing its
theoretical output capacity and so total costs for the portfolio as a whole are rising very
rapidly.

The portfolio short run total cost curve derived from this SRMC order of merit process is
depicted in Figure 7.5. The points labelled above the curve indicate where a plants
output begins to be ramped up and added to the portfolio. Conversely, points labelled
below the curve indicate where a plants output is held constant over a subsequent portion
of the portfolios output. The labels in Figure 7.5 correspond to the points with the same
labels in Figure 7.4. For output lower than the pooled level of mingen the same method
was applied as that given in Table 7.1 resulting in a similar saw tooth pattern as that given
in Figure 7.2.

$ SRMC Order of Merit Portfolio Short Run Total Cost Curve

f
e Pooled ramp-up
of output from all
Shutdown costs three plants
incurred for one or Plant C
g f
more plants below e
pooled mingen b

c Plant B
b
a Plant A
Pooled ramp-up of plant A
and plant B output

Pooled mingen of
the three plants
0 MWh

Figure 7.5 Sub-optimal portfolio short run total cost curve derived by combining cubic
function based plants in a SRMC order of merit

However, the stringent use of a SRMC order of merit (i.e. one based on the idea that
output is ramped-up as more output is required) neglects the fact that, for non-linear short
run total cost functions such as those in Figure 7.4, plant level SRMC is non-constant with
output. 36 Consider the cost path based on a strict SRMC order of merit overlayed in
Figure 7.6 with an optimal cost path specific to portfolio output Q. For portfolio output M
through to X, both curves coincide. However, for a portfolio output greater than X they

36
One common error is to think of economic representations such as those described by Figure 7.4 and
Figure 7.5 as dynamic, as opposed to static, models. In reality, while Figure 7.4 and Figure 7.5 describe a
cost-quantity relationship within a half hour slice of time they describe no aspect of temporal change.
Hence linear thinking in terms of output along a SRMC path is not actually appropriate: i.e. it is not correct
to think of optimisation in terms of one source of output being brought on earlier or later than another
source of output. Rather, over the half hour period, the three plants operate simultaneously, each at its
own average output throughout the interval. The optimal mix of outputs from the available plants will
produce the lowest cost curve for a particular level of portfolio output.

48 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

begin to diverge with the dashed curve representing the SRMC merit order case and the
solid line describing an optimal cost path to portfolio output Q. The reason for this
divergence is that, given portfolio output Q, optimality requires a higher share of plant B
generation than that provided by the SRMC merit order short run total cost curve.

More specifically, in the sub-optimal SRMC order of merit short run total cost curve case,
the mix of generation that provides portfolio output Q consists of:
a) the mingen levels of output M, plus
b) an additional amount of plant A output equal to S minus M, plus
c) an additional amount of plant B output equal to Q minus S;

whereas the optimal mix of generation that provides portfolio output Q consists of:
a) the mingen levels of output M, plus
b) an additional amount of base-load output equal to X minus M, plus
c) an additional amount of plant B output equal to Q minus X.

The optimal mix of portfolio output at Q results in a lower cost compared to the sub-
optimal mix by an amount equal to p2 minus p1.

$ Sub-optimal cost path to output Q

Optimal cost path to output Q

p2
p1 Y
Z
a
b
Optimal mix
of generation
to output Q

0 M X S Q MWh
Pooled Plant A Plant B
mingen variable output variable output

Figure 7.6 An optimal portfolio short run total cost path to portfolio output Q compared to
a SRMC order of merit portfolio short run cost path

7.1.4 Optimal short run total cost curve: The lower envelope of
all feasible cost paths
The particular optimal cost path to output Q shown in Figure 7.6 was obtained by
connecting the plant B and plant A short run total cost curves at the point marked Z (as
opposed to the point marked Y in the sub-optimal case). 37 Here the breakaway from the

37
Note that more than one optimal cost path might be able to describe an optimal generation mix for a given
level of output. Hence the term an optimal price path to a specific level of portfolio output is preferable to
the term the optimal price path to a specific level of portfolio output. In Figure 7.6, for example, an
alternative optimal cost path to that shown might have been drawn as a pooled ramp up between plant A
and plant B diverging from the sub-optimal curve at a portfolio output less than X.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 49
Technical Paper
 Economic Regulation Authority

strict SRMC order of merit rule can be seen, because at point Z tangent a (along the sub-
optimal curve using plant A output) is not as steep as tangent b (along the optimal curve
using plant B output). Note also that for the portfolio output between X and S the SRMC
merit order short run total cost curve is lower than the optimal path shown. This indicates
that each level of portfolio output has its own optimal generation mix associated with it. It
follows that the lower envelope of all feasible price paths describes the optimal portfolio
short run total cost curve. An example is depicted in Figure 7.7 which corresponds to the
set of plants depicted in Figure 7.4.

$ Optimal Portfolio Short Run Total Cost Curve

Lower envelope of all

feasible cost paths

0 M MWh

Figure 7.7 The optimal portfolio short run total cost curve as the lower envelope of all
feasible cost paths

7.1.5 Optimal portfolio SRMC

Figure 7.8 depicts the SRMC curve for the portfolio of plants described in Figure 7.9.
Portfolio output M corresponds to the sum of all three plants mingen levels. The saw
tooth shape that is evident in the portfolio SRMC curve comes about mainly due to the
heat rate dominated cost assumptions that result in a U shaped SRMC curve for fossil
fuel based plants. These occur at the particular tipping points in portfolio output when it
first becomes optimal to include a higher merit plant in the generating mix. The exact
location of these points along the portfolio quantity axis depends upon the unique
confluence of the SRMC curves associated with the available plants. As with Figure 7.3,
portfolio SRMC below pooled mingen is treated as the change in price over the change in
quantity from output zero to output M.34, 38

38
See Appendix 1 (Section A1.02) below for a mathematical interpretation of a similar set of plants to those
given in this example.

50 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

$
Optimal Portfolio SRMC

Note: Quantity band average

marginal cost measure used
below pooled mingen

0
M MWh

Figure 7.8 Optimal portfolio SRMC comprising plants with technology based on a cubic
short run total cost function

7.1.6 Example 3: The effect of differing shutdown and mingen

costs between plants
Plants of varying design, vintage, and level of maintenance are likely to face differing
shutdown costs and mingen costs. Base-load facilities, for example, are likely to have
large shutdown costs relative to mid-merit or peaking plants. Figure 7.9 depicts a portfolio
of linear cost plants that have already been started prior to the relevant half hourly trading
interval. However, this time the portfolio consists of generating facilities with differing
shutdown cost assumptions. 39

39
The total costs for the three plants, A, B and C, were modelled as follows:
18,000 if Q A = 0
CA = ,
20Q A if 91 Q A 340
1,000 if QB = 0
CB = , and
75QB if 26 QB 90
2,100 if QC = 0
CC = ,
150QC if 16 Q 50
where Ci denotes the total cost of plant i, and Qi 0 denotes the output of plant i, {i A, B, C}

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 51
Technical Paper
 Economic Regulation Authority

Plant A: low variable cost Plant B: mid variable cost Plant C: high variable cost
$ high shutdown cost low shutdown cost mid shutdown cost
$ $
18,000
Large shutdown
cost

Small shutdown
costs
a c
b 2,100
1,000
0 0 0
91 340 26 90 16 50
MWh MWh MWh

Figure 7.9 A portfolio of three linear short run total cost plants with sunk startup costs and
differing shutdown costs between plants

As with the example given in Figure 7.1, for any given level of output, the SRMC of plant A
in Figure 7.9 is less than the SRMC of plant B, which in turn is less than the SRMC of
plant C. However, the differing shutdown costs between plants in Figure 7.9 means that,
unlike the example in Figure 7.9, where shutdown costs were equalised with mingen costs
for each plant, a comparison must now be made between the cost of running a plant at
mingen (or above) and shutting it down. Indeed, despite the fact that the variable costs of
the plants in Figure 7.9 are similar to those in Figure 7.1, the change in shutdown costs
changes the optimal shutdown decision significantly.

Table 7.2 describes the constraints affecting the optimal sourcing of portfolio output from
the three available plants, depicted in Figure 7.9, for various ranges of portfolio output.
Compare Table 7.1 with Table 7.2. The two tables are identical up until a portfolio level of
demand of 116MWh. In Table 7.2, however, the differing shutdown costs between each
plant means it is best to shut plant B down between 117MWh and 359MWh of portfolio
output, and to run all three plants for portfolio output of 360MWh or above. In contrast,
Table 7.1 indicates that it is best to shut plant C down between 117MWh and 132MWh of
portfolio output, and beyond 133MWh of portfolio output it is best to keep all three plants
running.

The optimal portfolio short run total cost curve derived from the optimal shutdown decision
for the plants in Figure 7.9 is depicted in Figure 7.10 (contrast with Figure 7.2). The
diagram shows that shutdown costs impact upon optimal short run total cost even for
levels of portfolio output beyond the pooled mingen of all three plants (i.e. 133MWh in
Figure 7.10). In fact, in Figure 7.10 the minimum total cost in the range of portfolio output
between 133MWh and 359MWh is achieved by shutting down plant B and running plants
A and C. In other words, the particular combination of variable and shutdown costs given
in this example means it is only optimal to keep all three plants running if demand
exceeds 359MWh.

52 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Table 7.2 The optimal choice of plant operation and shutdown for the portfolio of plants
given in Figure 7.9
Discrete
portfolio
output Determining factor for Optimal Shutdown
range bottom of output range Run constraints vs. Run decision
MWh Plant Decision
0 - 15 Zero output. All three plants must shutdown. A Shutdown
B Shutdown
C Shutdown

16 - 25 Mingen plant C. Plants A and B must shutdown. A Shutdown

B Shutdown
C Run

26 - 41 Mingen plant B. Plant A must shutdown and if plant B A Shutdown

runs, plant C must shutdown.
B Run

C Shutdown

42 - 90 Mingens plants A + B. Plant A must shutdown. A Shutdown

B Run
C Run

91 - 106 Mingen plant A. If plant A runs, plants B and C must A Run

shutdown.
B Shutdown
C Shutdown

107 - 116 Mingens plants A + C. If plant A runs, plant B must shutdown. A Run
B Shutdown
C Run

117 - 132 Mingens plants A + B. If plant A runs, plant C must shutdown. A Run
B Shutdown
C Run

133 - 359 Mingens plants A + B + C No run constraints. A Run

(pooled mingen of the three
plants). B Shutdown
C Run

360 + Total cost with plants A and C No run constraints. A Run

running and B shutdown
exceeds total cost of all plants B Run
running. C Run

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 53
Technical Paper
 Economic Regulation Authority

$ Optimal Portfolio Short Run Total Cost Curve

Pooled mingen of
the three plants

0 133 360 MWh

Figure 7.10 Optimal portfolio short run total cost curve derived by combining linearly
increasing total cost technology plants in a SRMC order of merit

7.1.7 Optimal portfolio SRMC

Figure 7.11 depicts the SRMC curve for the portfolio of plants described in Figure 7.4.
Portfolio output M corresponds to the sum of all three plants mingen levels, while portfolio
output m corresponds to the mingen level of plant A (i.e. the low variable cost, high
shutdown cost, plant). The sharp spikes that are evident in the portfolio SRMC curve
come about at the particular tipping points in portfolio output where a change occurs in the
optimal plant shutdown decision (see Table 7.2 above). Portfolio SRMC below plant A
mingen is treated as the change in price over the change in quantity from portfolio output
zero to portfolio output m.34

$ Optimal Portfolio SRMC

0
m M MWh

Note: Quantity band average

marginal cost measure used
below plant A mingen

Figure 7.11 Optimal portfolio SRMC comprised of plants with a linearly increasing total
cost technology and with sunk startup costs

54 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

7.2 The equimarginal principle

The method applied in Section 7.1.4 is consistent with a fundamental principle of
economics: the equimarginal principle. In the current context this principle refers to the
fact that, for a given level of portfolio output, portfolio short run total cost will be minimised
when each of the plants that have been started operate at the same SRMC. In other
words, with multiple sources of production, optimisation requires a distribution of output
between each source so that SRMC is equalised across them. This principle remains in
place regardless of the combination of available plants and the particular shape of their
total cost curves.

Consider the portfolio of two plants given in Figure 7.12. Assume demand for electricity
for the half hour is at 100MWh. One (sub-optimal) way of producing this amount of
portfolio output would be to supply 50MWh from each plant. However, given that plant A
has a SRMC of $10 at 50MWh compared to $18 for plant B at 50MWh, reducing
production from plant B by 1MWh would save $18 at a cost of $10 if it were replaced by a
MWh of production from plant A. The overall marginal saving in this case would be $8.
For a cost minimising firm, this process of marginal reallocation of production from the
high marginal cost plant to the low marginal cost plant would continue until the plants
marginal costs were equalised. In Figure 7.12, this occurs when output from plant A is at
60MWh and output from plant B is at 40MWh.

Alternatively, Figure 7.13 shows how the equimarginal principle applies in the case of two
constant SRMC technology plants. Again, if the demand for electricity is assumed to be
100MWh for the half hour trading interval, 50MWh of supply from each plant would be
sub-optimal. This is because, increasing output from plant A to its capacity of 60MWh
costs $100, while decreasing output from plant B by the same amount saves $180: a net
saving of $80. Because plant As SRMC curve becomes infinite at its capacity of 60MWh,
the marginal cost of both plants is equalised at a SRMC of $18.

$ Plant A $ Plant B
SRMCA SRMCB

18
10
0 0
50 60 MWh 40 50 MWh

Figure 7.12 The equimarginal principle for U shaped SRMC plants

$ Plant A $ Plant B
SRMCA SRMCB

18

10
0 0
50 60 MWh 40 50 MWh

Figure 7.13 The equimarginal principle for constant SRMC plants

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 55
Technical Paper
 Economic Regulation Authority

7.3 Short run total cost for a portfolio comprising one

or more plants with avoidable fixed costs
When some plants in a portfolio are not already operating prior to a half hour trading
interval, a firm faces the decision as to whether or not the costs associated with starting
those plants should be incurred. While startup costs are a fixed cost at the level of the
plant, they can be thought of as variable costs at the level of the portfolio. The following
sections will provide examples of optimisation in the case where one or more plants in a
portfolio are available, but not started, prior to the half hour trading interval.

7.3.1 Example 4: Linearly increasing total cost technology with

differing shutdown costs and startup costs between
plants
In Figure 7.14 a firm with four linearly increasing total cost technology plants in its portfolio
has two already started prior to the half hour interval and two that could be started up, if
required, to meet expected demand during the interval. The costs of startup for the first
two plants are considered sunk, while for the two unstarted plants all costs are avoidable
(by not engaging in startup) with startup costs comprising an avoidable fixed cost
component. The firm faces the decision as to whether or not the costs associated with
starting additional plants to meet that demand should be incurred. Note that plant C
awaiting startup has a lower SRMC than the already operating plant B, as indicated by the
lesser slope of its short run total cost curve. 40

40
The total costs for the four plants A, B, C and D were modelled as follows:
18,000 if Q A = 0
CA =
20Q A if 91 Q A 330
1,400 if QB < 26
CB =
75QB if 26 QB 95
2,100 if QC = 0
CC = , and
15QC if 26 QC 95
18,000 if Q A = 0
CD = ,
150Q A if 12 Q A 340
where Ci denotes the total cost of plant i, and Qi denotes the output of plant i, {i A, B, C , D}.

56 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Plant A: Plant B: Plant C: Plant D:

sunk startup sunk startup avoidable startup avoidable startup
$ $ $ $
18,000

3,632 3,675
1,400
0 0 0
91 MWh 26 MWh 0 26 MWh 12 MWh

Figure 7.14 - A portfolio comprising plants with technology based on a cubic short run
total cost function: two with sunk startup costs and two with avoidable fixed startup costs

The lower envelope of all feasible cost paths for the portfolio of plants in Figure 7.14 is
described by Figure 7.15. Despite the apparent difference between the sunk cost and
avoidable fixed cost cases, the decision as to whether or not to incur startup costs is the
same optimisation problem as presented in Section 7.1.4. That is, a profit maximising firm
will seek to minimise the total cost of generating a given portfolio quantity of electricity
over the half hour interval. Hence, the derivation of the optimal portfolio short run total
cost curve as the lower envelope of all feasible cost paths applies equally well to the case
where avoidable fixed costs are included in the short run total costs for one or more
generating facilities in the portfolio.

The various optimal cost paths in Figure 7.15 were obtained:

a) by considering the optimal shutdown decision for plants A and B.

b) by connecting the short run total cost curve for plant C (with avoidable fixed
startup and mingen costs) to the curve that describes the optimal mix of production
between plant A and plant B; and
c) by connecting the short run total cost curve for plant D to the curve that describes
the optimal mix of production between plant A, plant B and plant C.

For each given level of demand an optimal mix of output from the available plants
produces the required electricity at the lowest feasible cost. Note the sudden jumps in
optimal short run total cost as a result of the startup costs, in particular for plant D. These
result from the assumption of linearity up to capacity in the plant level short run total cost
functions.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 57
Technical Paper
 Economic Regulation Authority

$ Optimal Portfolio Short Run Total Cost Curve

0
91 MWh

Figure 7.15 - The lower envelope of all feasible short run total cost paths in a portfolio of
four plants with linear short run total costs and with two plants having avoidable fixed costs

Table 7.3 describes the constraints affecting the optimal sourcing of portfolio output from
the four plants depicted in Figure 7.14 for various ranges of portfolio output. For any level
of portfolio output left of the pooled mingen of the two plants already started (i.e. mingen
outputs A + B = 91 + 26 = 117 from Figure 7.14), the decision must incorporate the
shutdown cost of at least one of those two plants. Similarly, constraints apply to the
decision to start a plant up. It is only beyond the pooled mingen of all four plants (i.e.
outputs 91 + 26 + 26 + 12 = 155 from Figure 7.14) where there are no constraints as to what
plants can be utilised to meet demand. Note that such constraints are only one aspect of
the cost minimisation decision for a given level of portfolio output. That is, the optimal
decision as to what plants should operate or shutdown may differ at varying levels of
output, even within a range of output under which identical run constraints apply.

58 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
 Economic Regulation Authority

Table 7.3 The optimal choice of plant operation and shutdown for the portfolio of plants
given in Figure 7.14
Discrete Optimal decision
portfolio
output Determining factor for Shutdown vs. Startup & Run vs.
range bottom of output range Run constraints Run Dont Startup
MWh Plant Decision Plant Decision
0 - 11 Zero output. No plant can run. A Shutdown C Dont Startup
B Shutdown D Dont Startup

12 - 25 Mingen plant D. Plants A, B and C cannot run. A Shutdown C Dont Startup

B Shutdown D Startup & Run

26 - 37 Mingen plant B. Plant A must shutdown; A Shutdown C Dont Startup

if plants B, C or D run, no B Run D Dont Startup
other plant can run.
38 - 51 Mingens plants (B = C) + D. Plant A must shutdown; A Shutdown C Dont Startup
if plants B or C run, the other B Run D Dont Startup
plant cannot run.
52 - 63 Mingens plants B + C. Plant A must shutdown; A Shutdown C Dont Startup
if plants B and C run, plant D B Run D Dont Startup
cannot run.
64 - 77 Mingens plants B + C +D. Plant A must shutdown. A Shutdown C Dont Startup
B Run D Dont Startup

78 - 90 Plants A + B shutdown costs Plant A must shutdown. A Shutdown C Startup & Run
exceed plant C total cost.
B Shutdown D Dont Startup

91 - 101 Mingen plant A . If plant A runs, no other plant A Run C Dont Startup
can run.
B Shutdown D Dont Startup

102 - 116 Mingens plants If plants A and D run, no A Run C Dont Startup
A + D. other plant can run.
B Shutdown D Dont Startup

117 - 128 Mingens plants If plants A and B or C run, no A Run C Dont Startup
A + (B = C). other plant can run.
B Shutdown D Dont Startup

129 - 142 Mingens plants If plants A and D run, and B A Run C Dont Startup
A + (B = C) + D. or C run, the other plant
cannot run. B Shutdown D Dont Startup

143 - 154 Mingens plants If plants A, B and C run, plant A Run C Dont Startup
A + B + C. D cannot run.
B Shutdown D Dont Startup

155 - 330 Mingens plants No run constraints. A Run C Dont Startup

A + B + C + D. B Shutdown D Dont Startup

331 - 404 Maximum output of plant A No run constraints. A Run C Dont Startup
exceeded.
B Run D Dont Startup

405 - 519 Plants A + B total costs No run constraints. A Run C Startup & Run
exceeds plant B shutdown cost
+ plants A and C total costs. B Shutdown D Dont Startup

520 + Maximum output No run constraints. A Run C Startup & Run

plants A + B + C B Run D Startup & Run

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 59
Technical Paper
 Economic Regulation Authority

7.3.2 Optimal portfolio SRMC

Figure 7.16 depicts the discrete portfolio SRMC curve for the portfolio of plants described
in Figure 7.14.34 Note that the step function is not monotonically increasing. For example,
there is a decrease in cost as a result of the startup of plant C which has a lower SRMC
than both plants already operating. Moreover, an off-the-scale spike in portfolio SRMC is
evident at the point where it is optimal to start plant D up to meet demand.

$ Optimal Portfolio SRMC

Thin spike in portfolio SRMC results

from the cost of plant D startup

0
MWh
91
Decrease in SRMC step function results
from the cost of plant C startup

Note: Quantity band average marginal cost

measure used below plant A mingen

Figure 7.16 - Optimal portfolio SRMC for a cubic-based total cost function technology with
one or more plants having avoidable fixed costs

7.3.3 Example 5: Portfolio of plants with cubic based total cost

functions and with differing shutdown costs and startup
costs between plants
Figure 7.17 describes a firm with four plants in its portfolio: two with sunk startup costs,
and two that can be started when optimal. Each plant has a short run total cost function
that is based on a cubic functional form.

60 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Plant A: Plant B: Plant C: Plant D:

sunk startup sunk startup avoidable startup avoidable startup
$ $ $ $

0 0 0 0
m MWh MWh MWh MWh

Figure 7.17 A portfolio comprising plants with technology based on a cubic short run total
cost function: two with sunk startup costs and two with avoidable fixed startup costs
Figure 7.18 reveals the optimal short run total cost curve corresponding to the portfolio of
plants described by Figure 7.17. Note that, as a result of the optimisation process, and
despite the large avoidable fixed costs associated with the two plants requiring startup,
there is no sudden jump in the optimal portfolio short run total cost curve. The lower
envelope of all feasible cost paths produces a relatively smooth curve.

$ Optimal Portfolio Short Run Total Cost Curve

0 m MWh

Figure 7.18 The lower envelope of all feasible short run total cost paths for a portfolio of
four plants with cubic based short run total costs and with two plants awaiting startup

7.3.4 Optimal portfolio SRMC

The saw tooth pattern that is evident in the SRMC curve for the portfolio of plants given in
Figure 7.17 is remarkably similar to that described in Section 7.1.5, despite the
introduction of startup costs for half of the plants in the portfolio. 41 Again, the saw teeth

41
Note that the portfolio depicted in this section has one more plant to that in Section 7.1.5.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 61
Technical Paper
 Economic Regulation Authority

occur at the particular tipping points in portfolio output when it first becomes optimal to
include output from a higher cost plant in the generating mix. 42 In the case of a plant
already started, this applies to non-mingen variable output only (mingen output is already
included in the mix). For plants not yet started however, the tipping points can be quite
dramatic because of the sudden addition of a mingen quantity of output to the mix at zero
marginal cost. This, combined with the U shape of plants SRMC curves, causes the rate
of increase in optimal portfolio short run total cost to decline abruptly before increasing
again. 43

Optimal Portfolio SRMC

$

MWh
m

Note: Quantity band average

marginal cost measure used
below plant A mingen

Figure 7.19 Optimal portfolio SRMC for a cubic-based total cost function technology with
one or more plants having avoidable fixed costs

42
A (perhaps) surprising finding that results from the optimisation process is that, where plants have U
shaped SRMC functions, startup costs do not trigger a sudden jump in portfolio short run total cost and
therefore do not represent an off the scale spike in SRMC. The saw teeth spikes that do occur are similar
in scale for a portfolio comprising plants with sunk startup costs as those for a portfolio comprising one or
more plants that require startup within the relevant half hour trading interval.
43
See Appendix 1 (Section A1.03) below for a mathematical interpretation of a similar set of plants to those
given in this example.

62 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

8 Discussion
The purpose of this section is to discuss the various findings of the preceding sections
and the modelling conducted in the Appendices of this paper. Much consideration is
given to market efficiency and the effectiveness of the market rules given the likely shape
of the portfolio SRMC curve. Measurement difficulties are also outlined, with specific
regard given to the issue of data quality. Finally, a projection is given in relation to the
competitive evolution of the market and the likely duration of the need for the SRMC rule
requirement.

8.1 Estimation of plant level functional form and its

effect on portfolio SRMC
The use of mathematical modelling, together with spreadsheet software, enables
graphical depiction of the general features of portfolio SRMC in electricity generation. In
Appendix 1 of this paper, the specific functional form chosen to describe fossil fuel plant
short run total cost was a piecewise function that comprised of: (a) some component of
mingen at either zero economic cost (in the case of a plant with sunk startup costs) or at a
positive avoidable fixed cost (in the case of a plant requiring startup); and (b) a rational
function with a cubic function as the numerator and a linear function as the denominator
giving a shape to the curve. Provided three key assumptions are maintained,
choosing the optimal mix of generation out of a portfolio of plants leads to the saw tooth
shaped portfolio SRMC pattern described in this paper. First, a plants SRMC curve is
assumed to initially decrease with improving thermal efficiency beyond some mingen
amount. Second, an individual generating facility faces a capitalised thermodynamic limit
in terms of output capacity that can be approached, but not exceeded, by the SRMC
function. Third, a plants SRMC will eventually increase as this limit is approached.

These assumptions are consistent both with the economic law of diminishing marginal
returns and engineering based observations of thermal efficiency in fossil fuel plants.
However, as is evident from the analysis in Section 5, a different choice of functional form
will change the general features of the portfolio SRMC curve generated from the
optimisation process described herein. 44 For example, the author has conducted a similar
optimisation exercise to that conducted in Appendix 1 on a portfolio of plants assumed to
have sunk startup costs, low shutdown costs, and monotonically increasing SRMC
functions. The result was a portfolio SRMC function that was also monotonically
increasing. In contrast, as described in Section 7 above, an assumption of linearity in
plant level short run total cost results in a step function that may or may not increase
monotonically depending upon the requirement to start a plant up to meet demand.

The important concept that connects the result of portfolio optimisation to the short run
total cost curves of individual generating facilities is that the lower envelope of all feasible
costs paths for a portfolio of plants provides the optimal portfolio short run total cost curve,
with portfolio SRMC being the slope of this curve. At every point along the optimal
portfolio short run total cost curve, all plants operating have equalised SRMC, as is
consistent with the equimarginal principle. Any combination of plant that diverges from
this principle will be inefficient. Therefore, in SRMC terms, for an efficient firm, there is no
such thing as a most expensive plant in operation during a trading interval. The obvious
exception occurs under the assumption of plant level short run total cost linearity across

44
This fact points towards the importance of data collection and empirical analysis.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 63
Technical Paper
 Economic Regulation Authority

the whole portfolio. In this case, the portfolio SRMC step function that results may be
considered to have an identifiably marginal or most expensive plant.

8.2 Basic efficiency criteria with respect to a saw

tooth SRMC function
The fact that portfolio SRMC may not be monotonically increasing does impose some
additional criteria to the setting of efficient output. This, however, does not prove
inconsistent with the concept of market efficiency occurring where price equals SRMC. In
Figure 8.1 one can imagine a competitive industry where market forces determine price P
for a trading interval. In the competitive environment all firms are price takers so no
output can be supplied above this price and profit maximising behaviour will correspond to
the efficient outcome. However, because price P passes through the saw tooth shaped
increasing portfolio SRMC curve at five different levels of output, profit maximisation has a
more complex decision criteria than that usually described in microeconomic textbooks.
In Figure 8.1, assuming average revenue is greater than average cost at all levels of
output, the firm will maximise its producer surplus by choosing output Q3 if areas (c + e) >
(b + d) and [(c + e) (b + d)] > (c b). Otherwise the firm will choose output Q2 if area c
> b, or Q1 if it is not. This is the best a firm can do in an efficient market because in the
competitive environment market price cannot be forced up by any given firm. There is, for
example, no scope for additional profit by adding a margin to the SRMC curve for
commercial risk or for fuel cost recovery on long run contracts. In a competitive market
such thinking will result in reduced profits. However, in a market that is influenced by
market power, a similar approach will push the price up, resulting in monopoly rents for
the firm at the expense of consumers, thus reducing the net benefits of the market as a
whole. Such behaviour is unacceptable under the market rules.

$ Optimal
Portfolio SRMC

b e
d
P
a c

M Q1 Q2 Q3 MWh

Figure 8.1 The efficient output choice for a firm with a portfolio SRMC curve that does not
increase monotonically

A saw tooth portfolio SRMC curve does have implications with regards to the
effectiveness of the market rules. To the extent the word reflect in clause 6.6.3 of the
market rules (see Section 1 above) is deemed to be equivalent to the word reproduce,
there exists a conflict between various requirements of the market rules - specifically: the

64 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

requirement to offer electricity at SRMC and the requirement to do this in monotonically

increasing steps in price intervals bounded by fuel specific caps. Obviously,
monotonically increasing steps make the algorithms that generate market prices easier to
construct. However, any disparity between a firms offer curve and its actual SRMC will
fail to achieve the economic efficiency aspects of the market objectives.

To use a common analogy, a strict interpretation of the word reflect would require
generators to fit a square peg into a round hole. If a generator constructs its portfolio offer
curve in the manner described in the partial equilibrium framework of Figure 8.2, for
example, the result is a market price of Pm as opposed to the efficient price of Pe and
thus a deadweight loss to society of area b + c. Furthermore, because the optimal
portfolio SRMC curve lies below the portfolio offer curve at the level of marginal benefit
(as represented by the bid curve), the result is a large transfer of wealth from consumers
to producers. This amounts to area a + b, thus failing market objectives aimed at
encouraging economic efficiency.

If the choice as to where a step should lie is left open to the interpretation of a generator, it
would be economically rational for that firm to choose a configuration that maximises its
economic rent to the cost of consumers. The potential for gaming such an outcome will
be directly correlated with the firms ability to avoid regulated civil penalties as much as
the scope of those penalties. In this context, the fact that a series of monotonically
increasing steps cannot be fitted to the true nature of a generators SRMC takes on
obvious importance. The problem is further exacerbated by the requirement to fit offer
curves within stipulated price floors and price caps. If, operating under these
contradictions, a portfolio offer curve such as that depicted in Figure 8.2 were deemed to
be noncompliant with the market rules, generators may, without clear guidance, claim they
are being lured into a straw man argument. Conversely, if the portfolio offer curve
depicted in Figure 8.2 were deemed to be consistent with the market rules, the outcome
may approach that of full monopoly pricing, leaving the regulated SRMC rule wanting as a
market efficiency instrument.

$ Optimal
Portfolio SRMC
Bid Curve

Portfolio Offer
Curve

Loss of consumer surplus = Area a + b

Pm

a b
Deadweight loss = Area b + c
Pe c

Qm Qe d

Figure 8.2 Example of total and consumer surplus losses that can occur where a
monotonically increasing portfolio offer curve cannot be fitted to an optimal portfolio SRMC
curve which is not monotonically increasing over the full range of output

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 65
Technical Paper
 Economic Regulation Authority

The extent to which this problem can be mitigated depends on the number of plants
available to a monitored firm in a trading interval as well as the interpretation of the market
rules. Firstly, a firm with many plants (and thus one likely to have a degree of market
power) may face a saw tooth pattern that can be divided into a series of relatively
horizontal bands. As such, a large number of available plants may naturally limit the
extent to which monopoly profits can be extracted via an offer curve strategy such as that
depicted in Figure 8.2. Secondly, if the average marginal cost over a quantity band were
deemed to reflect SRMC, a second best outcome would be achieved in economic
efficiency terms, on average, over repeated iterations. In fact, a quantity band average
marginal cost interpretation could be applied to any series of offer steps that a firm places
through a saw tooth SRMC curve, regardless of the steepness of its general upwards
trend. Such an interpretation would also smooth out any thin spikes in SRMC that might
result from startup (shutdown) costs associated with bringing a plant online (offline) during
a trading interval.

Figure 8.3 provides an example. A firm divides its output into five steps. In calculating the
price of each step, the firm takes the average of the portfolio SRMC curve over the range
of output that corresponds to each step. For example, the step between outputs a and b
has an average portfolio SRMC of price p. In other words, price p represents the average
value of the portfolio SRMC curve bounded by outputs a and b. The same procedure is
followed in the construction of other steps. The exception is where the average portfolio
SRMC is higher than a fuel specific price cap over a range of output. In this case the
price of the step equals the price cap. Box 8.1 defines this approach in mathematical
terms.

Price cap Optimal

Portfolio SRMC

Portfolio Offer
Curve

a b MWh

Figure 8.3 A quantity band average marginal cost approach to reflecting portfolio SRMC
with an offer curve comprising non-monotonically decreasing steps.

66 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Box 8.1 Mathematical definition of quantity band average portfolio SRMC steps

Consider the following generic function for a portfolio of plants:

dC
C (Q) = , Q 0,
dQ

where portfolio SRMC (denoted C ) is a function of portfolio output Q and is the first
derivative of the short run total cost function. In this case, the average portfolio SRMC on
the interval [a,b] is given by:

1 b

b a a
C (Q) dQ ,

provided the specified definite integral exists.

In most cases, the portfolio SRMC will not be continuous over a range of output. It is
therefore appropriate to approximate the above by means of the following discrete
method:

1 n
C (Qi )Q ,
b a i =1

where portfolio SRMC (denoted C ) over the interval [a,b] is divided into n subintervals
between i and n, each of width Q . The above is the average portfolio SRMC function
for the n subintervals and the given values of Qi . The greater the value of n, the closer
the result will be to that of the definite integral method.

8.3 Empirical issues

Turning to the question of empirics, measurement difficulties arise due to the unavailability
and quality of suitable data from market participants. For example, the real time metering
of power stations for both fuel input and generation output is inherently inaccurate and
unsuitable for SRMC estimation techniques (Waters, 2007; per com). 45 The main issue
in this regard is the lag between fuel use and electricity output, which prohibits an
accurate half hourly estimation of the relevant input-output relationship. However, a good
substitute exists in the unit heat rate curves which are available to market participants.
Using the cubic estimation technique described in Section 5 of this paper, these curves
can be readily converted into a total fuel cost curve by half hour, which can in turn be
converted into a marginal factor cost curve for the trading interval. This method allows
practical evaluation of the most significant of the short run variable costs associated with
electricity supply.

The quality of startup cost, shutdown cost, operational and maintenance cost, and line
loss data, however, remains an issue. Thus far, the Authority is in possession of such
information in its summary form, rather than in a form that reveals the method by which it

45
Waters, J. Verve Energy, Per com., email, 9th August 2007.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 67
Technical Paper
 Economic Regulation Authority

was derived. This does not assist auditing on the part of the Authority to determine
whether the data is robust by engineering and economic standards.

A further problem arises due to the gap between engineering and economic based
representations of electricity supply. The software packages typically used by generators
to optimise plant and equipment operate as a black box, i.e. without the users detailed
understanding of how its output is generated. Another constraint is that SRMC is an
economic concept that is not well understood within the electricity industry. This second
factor was evident when Short & Swan (2002) attempted to estimate the SRMC of
generators in the Australian national electricity market. As they put it, they were forced
to use the information revealed by the generators themselves through their publicly
available supply offers as a means of estimating their generator marginal costs because
of a lack of accurate and detailed information on marginal costs for Australian
generation stations. 46 In reference to Short & Swans paper, and to marginal cost
based measures of market power more generally, Brennan (2002) pointed towards the
common, economically erroneous, use of the average variable or operating cost of the
last generator that would be dispatched to meet energy demand 47 as a proxy for
marginal cost in models of electricity generation. 48 ROAM Consulting, which was
commissioned to model SRMC in the WEM, sees the textbook definition of SRMC as
those costs of a firm that vary directly with the short run level of production. 49 This
approach is consistent with Brennans concerns. A similar difficulty arises from the use of
the PowrSym3 tool by Verve Energy in determining its portfolio marginal cost for each
half hour as the operating cost of the most expensive generating unit dispatched in that
half hour. 50

8.4 Real competition vs. replicated competition

The SRMC rule creates the paradox of taking a centralised, legislated approach to
regulation while at the same time seeking an efficient, decentralised, market-based
allocation of resources. Conversely, under real competition, market forces tend towards
an efficient outcome without the need for participants to be aware of their SRMC. Ideally,
the condition of real competition is preferable to the condition of replicated competition.
Hence, the legislated SRMC approach should be thought of as a temporary, second best,
measure that is expected to be removed at such time when the first best condition (i.e. the
dissolution of market power) can be achieved.

While sound economic science is vital in good market design the main feature of a well
designed market is that it should not rely on the participants sound application of
economic science. However, clause 6.6.3 of the market rules obliges participants to
implement advanced economic techniques to determine their portfolio SRMC curve.
46
The assumption of perfect competition precludes this method in any evaluation of the abuse of market
power by participants (see, Short, C. & Swan, A. (2002), Competition in the Australian National Electricity
Market, ABARE, Current Issues 02.01, January, p. 4).
47
Brennan, T. (2002), Preventing Monopoly or Discouraging Competition? The Perils of Price-Cost Tests for
Market Power in Electricity, Resources for the Future Discussion Paper 02-05, p. 2. See also Brennan, T.
(2003), Mismeasuring Electricity Market Power, Regulation, Vol. 26, No. 1, pp. 60-65.
48
Brennans criticisms are relevant if SRMC is either U shaped (the standard theoretical description) or
linearly increasing. If, however, SRMC can be approximated by a horizontal straight line, SRMC will equal
short run average variable cost. Therefore, in the latter case, short run average cost would be a
reasonable proxy for SRMC.
49
ROAM Consulting (2007), SRMC STEM Offer Modelling, Report IMO 00010/MA to Independent Market
Operator, 22 May, p. 1.
50
Verve Energy (2007), STEM Pricing Using PowrSym, Report No VE 02/07, January, p. 3. Document not
publicly available: permission given for this quotation. Note: the statement quoted may reflect the view of
the papers author rather than Verve Energy as an organisation.

68 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Problems associated with this requirement were anticipated by Ruff (2002) when a similar
proposal to clause 6.6.3 of the market rules was tabled in California before its Senate
Judiciary Committee:
The fundamental economic and policy flaw in [the legislated SRMC approach] is its
focus on market outcomes rather than market structure, just the opposite of the approach
generally taken in competition policy and law. As a general matter, competition policy
does not try to determine what the outcome of a truly competitive market would be and
then impose that outcome on a basically uncompetitive industry. Instead, competition
policy tries to create an industry structure that is inherently competitive because it has
many, competing players, or at least to induce the players in the industry to act as though
they were competitive, and then lets the competitive process determine prices and other
features of the industry. A policy focused on market structure and (secondarily) behaviour
is far more feasible to implement than a policy focused on market outcomes, and is
economically preferable because it allows the operations and evolution of the industry to
be determined by competition rather than by calculations or judgments of competition
51
authorities and judges.

It should be noted that the legislated SRMC proposal was rejected in California. Indeed,
to the authors best knowledge, no wholesale electricity market beyond WA polices SRMC
in this manner. Economists tend to view the encouragement of competition via industry
disaggregation as a longer term objective. 52 That is, the actual or threatened entry of
competitors into a market, over time, not only forces prices down towards the efficient
level, but also favours those firms that are able to lower their costs. Given that a firms
economic management skills are as much a part of this natural selection process as any
other aspect of business operations, it should not be the role of a regulator to perform the
function on the firms behalf. This would be contrary to the concept of allowing the profit
motive to drive technical change and gains in efficiency.

The introduction of competition into a market should be thought of as an evolutionary

process both at the level of the market and at the level of regulation, with the fostering of
real competition being the best long term regulatory strategy. That is, regulatory
resources are best directed towards the promotion of a market that can evolve to actual
competitiveness, thus enabling a move away from the need to replicate efficiency under
conditions of acute market power. Hence the SRMC rule is viewed as a temporary
measure that seeks to emulate an efficiency concept which would otherwise occur
naturally in a market under real competition.

In WA, the competitive process is currently constrained by the States geographical

isolation and the dominance of the incumbent public generating firm. The upside to this is
that there is much scope for structural improvement, given that the dominant player
currently has in excess of 40 generating facilities at its disposal. In WA, the evolution of
energy market policy has followed a deregulation process starting with the break up of a
combined gas/electricity utility into separate gas and electricity utilities, through to the
current situation where the electricity industry is vertically disaggregated, with the publicly
owned generator constrained by legislation preventing it from adding to its portfolio
beyond an aggregate generating capacity of 3,000MW (not including renewable energy
sources). 53 In August 2007, the total installed capacity in the SWIS stood at 4,106MW.

51
Ruff, L. (2002), Statement of Larry E. Ruff, PhD on California State Senate Bill No. 2000 Unlawful Electric
Power and Natural Gas Practices, Before the Senate Judiciary Committee, April 23, p. 7.
52
See, for example, Winston, C. (1998), U.S. Industry Adjustment to Economic Deregulation, Journal of
Economic Perspectives, Vol. 12, No. 3, pp. 89-110; and Borenstein, S. & Bushnell, J. (2000), Electricity
Restructuring: Deregulation or Reregulation?, Regulation, Vol. 23, No. 2, pp. 46-52.
53
Ministerial Direction under the Electricity Corporations Act 2005, Effective 1 April 2006, Signed 21st March
2006.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 69
Technical Paper
 Economic Regulation Authority

An installed capacity of 4,786MW was anticipated by mid 2010 (see Figure 8.4). 54 In
August 2007, Verve Energys installed capacity in the SWIS stood at 3,240MW. 55

Therefore, in a static policy environment (with respect to further market disaggregation),

the evolution towards a truly competitive market would be expected to occur gradually
over the next one to two decades. A linear projection beyond the current schedule for
plant commissioning and decommissioning, while maintaining Verve Energys capacity
constant at 3,000MW, indicates Verve Energys market share may reach 50 per cent by
early 2016. 56 It is anticipated that the SRMC rule may still be required well beyond this
date until such time that Verve Energys share of capacity is matched by one or more
generating firms.

100
Verve projected percentage share
of installed market capacity

75

50

25

0
31/08/2007 6/01/2008 14/05/2008 20/09/2008 27/01/2009 5/06/2009 12/10/2009 18/02/2010 27/06/2010

Date

Figure 8.4 Expected time evolution of Verve Energys market share based on System
Management projections

54
System Management (2007), Medium Term Projected Assessment of System Adequacy Report, August,
Available IMO website: http://www.imowa.com.au/10_5_1_n_mt_pasa.htm
55
Verve Energy Home Page, Accessed 24 August 2007, Available: http://www.verveenergy.com.au/
subContent/companyProfile/Company_Profile.html
56
The accuracy of this estimate will depend on factors such as economic growth, technological change and
changes to government policy.

70 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

9 Conclusions
In the WEM, and in wholesale electricity markets more generally, an efficient rationing of
scarce resources will only occur if: (a) generators are paid their marginal costs of
production, while; (b) consuming firms pay (equivalent to) a competitive price for their
business inputs. In other words, efficiency will only occur if the wholesale price of
electricity equals its SRMC of supply. A wholesale price above SRMC will lower the
wealth generation potential of the market as a whole. Critically, given that no market
operates in isolation, an inefficient wholesale electricity price will distort the broader
economy, leading to a widely distributed loss of wealth.

The presence of market power in the WA electricity industry has resulted in regulation
intended to replicate the outcomes of a competitive market. Clause 6.6.3 of the market
rules requires generators to offer their electricity at a price that reflects their reasonable
day ahead expectations of SRMC. These offers must be submitted as a portfolio supply
curve in an increasing sequence of steps and under the constraint of price floors and fuel
specific price caps. To the extent that the shape and range of the portfolio supply curve
under the above constraints does not match the portfolio SRMC curve, an inconsistency
arises in the requirements of the market rules. In effect, there is a trade-off between the
desirability of a simple user-friendly information technology system through which bid and
offer information is submitted to the IMO, the actual nature of cost, and the requirements
of economic efficiency.

However, use of the word reflect in clause 6.6.3 of the market rules provides scope for
flexibility. If the portfolio supply curve were determined by means of averaging marginal
cost over a submitted quantity band, a second best outcome would be averaged out over
repeated iterations. The conditions for this improve as the number of generating facilities
in the portfolio increases, thus tending to be better suited to firms with market power.
Therefore, provided price caps are based on worst case SRMC scenarios, a portfolio
supply curve could be fitted to SRMC in a manner that is consistent with the full
requirements of the market rules.

SRMC modelling has great potential in delivering commercial gains through improvement
in business operations that have not yet been exposed to the rigour of a competitive
market. Developing an understanding of a generators cost structure should, therefore, be
seen not only as a social benefit but also as a private benefit to market participants.
Projections suggest that the current condition of acute market power is likely to persist in
the WEM in the medium term at least, with acute market power defined as the absolute
majority share of installed capacity in the SWIS by a single generating firm. Hence the
SRMC monitoring approach appears likely to remain a key part of the regulatory regime
for some time.

The information provided within this paper is intended to improve the understanding of the
SRMC concept within the Western Australian electricity industry. Clearly, information flow
of this nature is requisite for the effectiveness of clause 6.6.3 of the market rules.
Likewise, a flow of information in the opposite direction (i.e. from market participants to the
Authority) will be of critical importance. Legislative powers granted to the Authority under
the Electricity Industry (Wholesale Electricity Market) Regulations 2004 provide the
Authority with an ability to request such information. The quality of the data remains an
area of priority. However, provided the data input is sound, the modelling techniques
developed by the IMO and the Authority in conjunction with this paper place the Authority
in a good position to determine a firms reasonable expectations of portfolio SRMC.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 71
Technical Paper
 Economic Regulation Authority

References
Borenstein, S. & Bushnell, J. (2000), Electricity Restructuring: Deregulation or
Reregulation?, Regulation, Vol. 23, No. 2, pp. 46-52.

Brennan, T. (2002), Preventing Monopoly or Discouraging Competition? The Perils of

Price-Cost Tests for Market Power in Electricity, Resources for the Future Discussion
Paper 02-05.

Brennan, T. (2003), Mismeasuring Electricity Market Power, Regulation, Vol. 26, No. 1,
pp. 60-65.

Economic Regulation Authority (2008), Short Run Marginal Cost, State of Western
Australia.

Electricity Industry (Wholesale Electricity Market) Regulations 2004, Western Australia,

Available, Available World Wide Web, URL: http://www.eriu.energy.wa.gov.au/

Johnston, J. (1952), Statistical Cost Functions in Electricity Supply, Oxford Economic

Papers, New Series, Vol. 4, No.1, pp. 68-105.

Kim, J. Shin, D. Park. J, & Singh, C. Atavistic genetic algorithm for economic dispatch
with valve point effect, Electric Power Systems Research, 62, 2002, pp 201-207.

Klein, J. (1998), The Use of Heat Rate Curves in Production Cost Modeling and Market
Modeling, Electricity Analysis Office, California Energy Commission.

Market Procedure for Determining Loss Factors, Wholesale Electricity Market Procedure
(2006), Available IMO website: www.imowa.com.au/Attachments/Loss%20Factor%20
Procedure.pdf.

Ministerial Direction under the Electricity Corporations Act 2005, Government of Western
Australia, Effective 1 April 2006, Signed 21st March 2006.

Muth, J. (1961), Rational Expectations and the Theory of Price Movements,

Econometrica, Vol. 29, No. 3, pp. 315-335.

National Energy Technology Laboratory, Power Plant Water Usage and Loss Study,
August 2005

ROAM Consulting (2007), SRMC STEM Offer Modelling, Report IMO 00010/MA to
Independent Market Operator, 22 May.

Ruff, L. (2002), Statement of Larry E. Ruff, PhD on California State Senate Bill No. 2000
Unlawful Electric Power and Natural Gas Practices, Before the Senate Judiciary
Committee, April 23.

Schweppe, F., Caramanis, M., Tabors, R. & Bohn, R. (1988), Spot Pricing of Electricity,
Kluwer Academic Publishers, Boston.

Short, C. & Swan, A. (2002), Competition in the Australian National Electricity Market,
Australian Bureau of Agricultural and Resource Economics, Current Issues 02.01,
January.

72 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Stoft, S. (2002), Power System Economics, IEEE Press, New Jersey.

System Management (2007) , Medium Term Projected Assessment of System Adequacy

Report, August, Available IMO website:
http://www.imowa.com.au/10_5_1_n_mt_pasa.htm.

Verve Energy (2007), STEM Pricing Using PowrSym, Report No VE 02/07, January.

Verve Energy Home Page, Accessed 24 August 2007, Available: http://www.verveenergy.

com.au/subContent/companyProfile/Company_Profile.html.

Wang, H. & Yang, B (2001), Fixed and Sunk Costs Revisited, Journal of Economic
Education, Vol. 31, No. 2, Spring, p. 178-185.

Waters, J. Verve Energy, Per com., email, 9th August 2007.

Wholesale Electricity Market Amending Rules (December 2006), Western Australia,

Available IMO website: http://www.imowa.com.au/market_rules.htm.

Wholesale Electricity Market Procedure (2006), Market Procedure for Determining Loss
Factors, Available IMO website:
www.imowa.com.au/Attachments/Loss%20Factor%20Procedure.pdf.

Winston, C. (1998), U.S. Industry Adjustment to Economic Deregulation, Journal of

Economic Perspectives, Vol. 12, No. 3, pp. 89-110.

Wood, A. & Wollenberg, B. (1996), Power Generation, Operation and Control, 2nd. ed.,
Wiley-Interscience, New York.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 73
Technical Paper
 Economic Regulation Authority

APPENDICES

74 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Appendix 1. Optimisation examples

A1.01 Individual plant models
Consider the following generic function for an individual plant:

C i = C i ( Fi , S i , Qi , K i , M i ; j ,i ) , Fi 0, M i 0, 0 Qi < K i , i = 1,,n, j = 1,,m.

where short run total cost (denoted C) for plant i of n plants is a function of its output Qi ,
any mingen output M i , avoidable fixed costs Fi , shutdown costs S i , the maximum
thermodynamic output capacity of the facility K i (under current ambient conditions), and
where j ,i constitutes a set of m parameters for plant i.

The mathematical formulations that follow are intended to provide theoretical descriptions
of short run total cost and SRMC for individual plants and are for illustrative purposes only
i.e. the functional forms are not based on actual data. Indeed, the functions were
deliberately made strongly convex to better demonstrate the method with a small number
of plants (actual plant level SRMC functions may appear less convex). Nonetheless, the
specific assumptions that are made which are transparent within the functional forms
themselves reflect, in a general sense, thermodynamic principles and the law of
diminishing returns. That is, the SRMC functions initially decrease with improving thermal
efficiency before increasing asymptotically to a maximum output capacity.

The following functional form for plant-specific short run total cost is used forthwith:

S i if Qi = 0

C i = 1,i (Qi M i ) + 2,i (Qi M i ) + 3,i (Qi M i )
3 2

+ 4,i + Fi if M i Qi < K i
K i Qi

Taking the first derivative results in the SRMC function:

[ ]
( K i Qi ) 31,i (Qi M i ) 2 + 2 2,i (Qi M i ) + 3,i +
dC (Q M i ) + 2,i (Qi M i ) + 3,i (Qi M i )
3 2

Ci = i = 1,i i if M i Qi < K
dQi ( K i Qi ) 2

for,

0 M i K i , M i Qi < K i .

Consider five individual plants where {i 1,2,3,4,5 }. First, a low cost plant with sunk
startup costs:

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 75
Technical Paper
 Economic Regulation Authority

18,000 if Q1 = 0
1
C1 = (Q1 92) 3 18(Q1 92) 2 + 1,750(Q1 92)
16 + 5,000 if 92 Q1
291 Q1

(291 Q1 ) 3 (Q1 92) 2 36(Q1 92) + 1,750 +

16
1
(Q1 92) 3 18(Q1 92) 2 + 1,750(Q1 92)

C1 =
16
if Q1 92 .
(291 Q1 ) 2

Second, a mid cost plant with sunk startup costs:

1,400 if Q2 = 0
1
C 2 = (Q2 12) 3 56(Q2 12) 2 + 3,200(Q2 12)
4 + 2,500 if 12 Q2
111 Q2

(111 Q2 ) 3 (Q2 12) 2 112(Q2 12) + 3,200 +

4
1
(Q2 12) 3 56(Q2 12) 2 + 3,200(Q2 12)

C2 =
4
if Q2 12 .
(111 Q2 ) 2

Third, a mid cost plant that has not yet been started, technologically identical to plant 2:

0 if Q3 = 0
1
C3 = (Q3 12)3 56(Q3 12) 2 + 3,200(Q3 12)
4 + 2,500 + 1,100 if 12 Q3 < 111
111 Q3

(111 Q3 ) 3 (Q3 12) 2 112(Q3 12) + 3,200 +

4
1
(Q3 12) 3 56(Q3 12) 2 + 3,200(Q3 12)

C3 =
4
if Q3 12 .
(111 Q3 ) 2

Fourth, a high cost plant with sunk startup costs:

76 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

1,400 if Q4 = 0
1
C 4 = (Q4 7) 3 100(Q4 7) 2 + 3,400(Q4 7)
2 + 2,600 if 7 Q4 < 50
50 Q4

(111 Q4 ) 3 (Q4 7) 2 200(Q4 7) + 3,400 +

2
1
(Q4 7) 3 100(Q4 7) 2 + 3,400(Q4 7)

C4 =
2
if Q4 7 .
(50 Q4 ) 2

Fifth, a high cost plant that has not yet been started, technologically identical to plant 4:

1
(Q5 7) 3 100(Q5 7) 2 + 3,400(Q5 7)
C5 = 2 + 2,600 + 1,400 if 7 Q5 < 50
50 Q5

(111 Q5 ) 3 (Q5 7) 2 200(Q5 7) + 3,400 +

2
1
(Q5 7) 3 100(Q5 7) 2 + 3,400(Q5 7)

C5 =
2
if Q5 7 .
(50 Q5 ) 2

Geometric representations of the functions are provided on the following page (Figure
A1.1 through to Figure A1.10).

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 77
Technical Paper
 Economic Regulation Authority

180
20000 C ' 2 (Q 2 )
C 1 (Q 1) 150
15000

Cost units
Cost units

120
10000 90
60
5000
30
0 0
0 100 200 300 0 100 200 300
Output units (Q 1) Output units (Q 2 )

Figure A1.1 Figure A1.2

Figure A2.2

20000 180
C ' 2 (Q 2 )
C 2 (Q 2 ) 150
15000

Cost units
Cost units

120
10000 90
60
5000
30
0 0
0 100 200 300 0 100 200 300
Output units (Q 2 ) Output units (Q 2 )

Figure A1.3 Figure A1.4

Figure A2.4

20000 180
C 3 (Q 3 ) C ' 3 (Q 3 )
150
15000
Cost units
Cost units

120
10000 90
60
5000
30
0 0
0 100 200 300 0 100 200 300
Output units (Q 3 ) Output units (Q 3 )

Figure A1.5 Figure A1.6

Figure A2. 6

20000 180
C ' 4 (Q 4 )
C 4 (Q 4 ) 150
15000
Cost units
Cost units

120
10000 90
60
5000
30
0 0
0 100 200 300 0 100 200 300
Output units (Q 5) Output units (Q 4 )

Figure A1.7 Figure A1.8

Figure A2. 8

20000 180 C ' 5 (Q 5)

C 5 (Q 5) 150
15000
Cost units
Cost units

120
10000 90
60
5000
30
0 0
0 100 200 300 0 100 200 300
Output units (Q 4 ) Output units (Q 5)

Figure A1.9 Figure A1.10

78 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Figure A2.10

A1.02 Model of a portfolio comprising plants with

sunk costs
Consider a simple portfolio of three plants corresponding to those with sunk startup costs
given in Section A1.01: i.e. plant 1, plant 2 and plant 4. Portfolio short run total cost
C{1, 2, 4} = C ( ) , is a composite function that combines the short run total cost functions of
each plant C1 (Q1 ) , C 2 (Q2 ) and C 4 (Q4 ) at a portfolio quantity of total output
= Q1 + Q2 + Q4 .

Developing the optimal curve C{1, 2, 4} ( ) is a two step process. First, the function
C{1, 2} ( ) provides an interim curve that combines the low cost and the mid cost functions
C1 ( ) and C 2 ( ) as a composite function with Q4 = 0 , i.e. = Q1 + Q2 . This is
performed as follows:

C{1, 2} ( , p) = C1 ( p ) + C 2 [(1 p) ] ,

where p is the proportion of low cost plant production in the portfolio for the half hour
period, i.e. Q1 = p and Q2 = (1 p ) . The optimal interim curve is then developed by
minimising incremental C{1, 2} ( ) as follows:

C{1, 2} = C1 ( p ) C 2 [(1 p) ]
d
dp

= 0.

Therefore:

C1 ( p ) = C 2 [(1 p ) ] ,

which is the equimarginal principle of economics.

Let p 0 be the optimal proportion of output from plant 1 for any given and C{01, 2} ( ) be
the optimised total cost function of plants 1 and 2, then:

( )
C{01, 2} ( ) = C1 ( p 0 ) + C 2 (1 p 0 ) .

A graphical depiction of a spreadsheet model of the optimal interim curve is provided by

Figure A1.11.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 79
Technical Paper
 Economic Regulation Authority

24000
C {01, 2} ( )
20000

Cost units 16000

12000

8000

4000

0
0 50 100 150 200 250 300 350 400 450
Output units ( )

Figure A1.11

The second step is similar to the first, where the optimal interim curve C{01, 2} ( ) is
combined with C 4 ( ) to form a composite function that describes portfolio short run total
cost C{1, 2, 4} ( ) . This is performed as follows:

C{1, 2, 4} ( , p ) = C 4 ( p ) + C{01, 2} ((1 p ) )

where p is now the proportion of high cost plant production in the portfolio for the half
hour period. The optimal portfolio short run total cost curve is then developed by
minimising incremental C{1, 2, 4} ( ) as follows:

C{1, 2, 4} = C 4 ( p ) C{01, 2} [(1 p) ]
d
dp

= 0

C 4 ( p ) = C{01, 2} [(1 p) ]

Again, this corresponds to the equimarginal principle of economics. It follows:

( ) (
C{01, 2, 4} ( ) = C 4 p 0 + C{01, 2} (1 p 0 ) , )
where p 0 is the optimal proportion of output from plant 4 for any given and C{01, 2, 4} ( )
is the optimised total cost function of plants 1, 2 and 4.

A spreadsheet model of the optimal portfolio short run total cost curve is provided by
Figure A1.12.

80 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

24000
C {01, 2 , 4} ( )
20000

16000
Cost units
12000

8000

4000

0
0 50 100 150 200 250 300 350 400 450
Output units( )

Figure A1.12

d 0 C{01, 2, 4}
In Figure A1.13, a spreadsheet approximation of C{1, 2, 4} is given by using
d
= 1 when > 92 , and = 92 with C{01, 2, 4} = C{01, 2, 4} (0) C{01, 2, 4} (92) when
0 92 .
300
C{01, 2, 4}
250

200

150
Cost units

100

50

0
0 100 200 300 400
Output units( )
-50

-100

-150

Figure A1.13

A1.03 Portfolio model with a mix of sunk cost and

avoidable fixed cost plants
Consider a portfolio of four plants plant 1, plant 2 and plant 3 and plant 5 described in
Section A1.01. Portfolio short run total cost C{1, 2,3,5} = C ( ) , is a composite function that
combines the short run total cost functions of each plant C1 (Q1 ) , C 2 (Q2 ) , C 3 (Q3 ) and
C 5 (Q5 ) at a portfolio quantity of total output = Q1 + Q2 + Q3 + Q5 .

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 81
Technical Paper
 Economic Regulation Authority

The optimal interim curve from Section A1.02, C{01, 2} ( ) , which describes the optimal
combination of output from plant 1 and plant 2, is used as a baseline. Developing the
portfolio short run total cost function C{1, 2,3,5} ( ) is a two step process. First, the function
C{1, 2,3} ( ) combines the functions C{01, 2} ( ) and C 3 ( ) , creating a composite function
with Q5 = 0 , i.e. = Q1 + Q2 + Q3 . This is performed as follows:

C{1, 2,3} ( , p ) = C 3 ( p ) + C{01, 2} [(1 p ) ] ,

where p is the proportion of production from plant 3 in the portfolio for the half hour
period, i.e. Q3 = p and Q1 + Q2 = (1 p ) . The optimal curve is then developed by
minimising incremental C{1, 2,3} ( ) as follows:

C{1, 2,3} = C 3 ( p ) C{01, 2} [(1 p ) ]
d
dp

= 0

which corresponds to the equimarginal principle of economics:

C 3 ( p ) = C{01, 2} [(1 p ) ] .

The optimal proportion of output from plant 3, denoted p 0 , determines the optimised short
run total cost function, C{01, 2,3} ( ) , of plants 1, 2 and 3, as follows:

C{01, 2,3} ( ) = C 3 ( p 0 ) + C{1, 2} ((1 p 0 ) ) .

A spreadsheet model of the optimal short run total cost curve C{01, 2,3} ( ) is provided by
Figure A1.14.

82 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

32000
C {01, 2 , 3} ( )
28000

24000

20000
Cost units
16000

12000

8000

4000

0
0 100 200 300 400 500 600
Output units ( )

Figure A1.14

The second step is similar to the first step, where C{01, 2,3} ( ) is combined with C 5 ( ) to
form the composite portfolio short run total cost function C{1, 2,3,5} ( ) :

C{1, 2,3,5} ( , p ) = C 5 ( p ) + C{01, 2,3} ((1 p ) )

where p is now the proportion of high cost plant (i.e. plant 5) production in the portfolio
for the half hour period, i.e. Q5 = p and Q1 + Q2 + Q3 = (1 p ) . The optimal portfolio
short run total cost curve is then developed by minimising incremental C{1, 2,3,5} ( ) as
follows:

= C 5 ( p ) C{01, 2,3} [(1 p) ]
d
C{1, 2,3,5}
dp

= 0

C 5 ( p ) = C{01, 2,3} [(1 p) ].

It follows:

C{01, 2,3,5} ( ) = C 5 ( p 0 ) + C{01, 2,3} ((1 p 0 ) ) ,

where p 0 is the optimal proportion of output from plant 5 for any given and C{01, 2,3,5} ( )
is the optimised total cost function of plants 1, 2, 3 and 5.

A spreadsheet model of the optimal portfolio short run total cost curve is provided by
Figure A1.15.

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 83
Technical Paper
 Economic Regulation Authority

32000
C {01, 2 ,3, 5} ( )
28000

24000
Cost units 20000

16000

12000

8000

4000

0
0 100 200 300 400 500 600
Output units ( )

Figure A1.15

d 0 C{01, 2,3,5}
In Figure A1.16, a spreadsheet approximation of C{1, 2,3,5} is given by using
d
= 1 when > 92 , and = 92 with C{01, 2, 4,5} = C{01, 2, 4,5} (0) C{01, 2, 4,5} (92) when
0 92 .

525 C{01, 2,3,5}

450
375

300
Cost units

225

150

75

0
0 100 200 300 400 500 600
-75

-150
Output units ( )

Figure A1.16

84 Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals
Technical Paper
Economic Regulation Authority

Portfolio Short Run Marginal Cost of Electricity Supply in Half Hour Trading Intervals 85
Technical Paper