International Journal on Electrical Engineering and Informatics - Volume 11, Number 3, September 2019

Determination of Maximum Grid-Connected Photovoltaic Penetration

Level Based on Unit Commitment Solution
Penang Rahimta Mara, Sarjiya, Lesnanto Multa Putranto, and Muhammad Yasirroni

Department of Electrical and Information Engineering,

Universitas Gadjah Mada, Yogyakarta, Indonesia
[email protected]

Abstract: The technological development and the decline of PV investment have resulted in the
rapid growth of PV power plant. However, the power generated from the PV power plant can
not be controlled because it depends on the natural state of the sun. Therefore PV only supplies
power during the day. Consequently, the demand load supplied by the thermal generator becomes
smaller during the day with greater PV penetration. When the penetration is too large, there is a
possibility of a thermal generator scheduling failure. This paper presents a study of the
permissible photovoltaic (PV) power penetration related to generation scheduling by considering
system and generator units constraints. The unit commitment problem is solved using Mixed
Integer Quadratic Programming (MIQP). PV power plants with various sizes are put into the
system until the algorithm fails to find a unit commitment solution. The infeasible solution of
unit commitment problem occurred at PV penetration 69% due to the load ramping needed by
the system at certain hours is higher than the maximum ramping capability of the thermal
generator unit.

Keyword: PV penetration, MIQP, Economic Load Dispatch, Unit Commitment, Ramp rate

1. Introduction
The development of PV technology, the decline of PV investment, and government incentive
policy leads to rapid growth in grid penetration of PV plants [1][2]. Moreover, awareness of the
effects of climate change has led to the desire of the world community to use clean energy. The
total capacity of PV installed worldwide in 2018 was 480 GW[3]. However, there are some
problems with the large PV capacity in the power system because the PV power output cannot
be controlled. The power generated from the PV plants depends on the natural state of the sun.
Furthermore, it is concerned with generation scheduling consisting of thermal and PV. PV
generators are seen as a negative load. Hence, the system load is reduced by the PV power output,
and it is called a net-load [4], then the net-load will be supplied by thermal generators. Because
PV only supplies power during the day, the PV penetration causes the difference between the
peak load and the lowest load to be wider. Moreover, at a certain time, there is an increase or
decrease in electricity demand quickly, in a short time [5].
In order to meet the load requirement, generation scheduling is needed. Unit commitment
(UC) determines the on/off status of generators and power contribution of each online unit at the
same time. In solving the unit commitment problem, generally two basic decisions are involved,
namely the “unit commitment” decision and the “economic dispatch” decision. The “unit
commitment” decision involves the determination of the generating units to be running during
each hour of the planning horizon [6]. The “economic dispatch” decision involves the allocation
of the system demand and spinning reserve capacity among the operating units during each
specific hour of operation. The objective of UC is to minimise the total production cost over the
planning horizon while fulfilling the system constraint and physical constraints of the generator
unit. However, UC is a complex problem due to its complex mix-integer nonlinear formulation
[7][8].
There are various kinds of optimisation techniques to solve UC problems. For instance,
priority lists, dynamic programming, quadratic programming, mix-integer linear programming,
and metaheuristic methods [9][10]. Mixed integer programming capable of solving scheduling

Received: June 20th, 2019. Accepted: September 30th, 2019

DOI: 10.15676/ijeei.2019.11.3.11
610
Penang Rahimta Mara, et al.

problem and economic dispatch simultaneously. Also, it has been widely used to solve problems
related to UC. From a computational perspective, Mixed Integer Linear Programming (MILP)
can efficiently solve unit commitment problems in thermal generators [11]. Generation
scheduling using Mixed Integer Quadratic Programming (MIQP) focused on the impact of ramp-
rate constraints has been done by Yang et al. [12]. MIQP has also been used by Papageorgiou et
al. [13] to solve economic dispatch of electrical power generators with prohibited zones of
operation. MIQP is used by Lopez et al. [14] to solve unit commitment problems in large-scale
electric power systems with various types of power plants.
On the other hand, different studies have been carried out to determine PV penetration effects
on UC. Previous works have found that the more non-dispatchable units in the system, the more
dispatchable units are required to commit [15]. Other study shows the challenge of a sudden
change in the output of PV caused by moving clouds [16]. The generation scheduling changes
to respond the sudden change of PV generation. PV capacity that is allowed in the system must
consider the ramp-up/down capability, technical minimum load capability and response time for
starting up/down of thermal units [17].
In this paper, MIQP is used to solve the UC problem. The objective is to find the penetration
limits of PV based on generation scheduling that does not violate generator physical constraints
and system operational constraints.

2. Problem Formulation
UC problem is to find generators unit to be on and off within a specified period that can meet
load requirements, usually in day ahead horizon. This UC problem would be optimal when the
allocation of generator power can be optimally allocated during each time frame, which results
in the most economical generation costs, and it is satisfied all constraints.

A. Objective Function of Unit Commitment

The purpose of generation scheduling is to find minimal generation costs with available
generators to meet load requirements within a specified period. The objective function of unit
commitment can be formulated as follows
min ∑𝐼𝑖=1 ∑𝑇𝑡=1((𝑓(𝑃𝑔𝑖,𝑡 )𝑈𝑖,𝑡 ) + 𝑆𝑈𝑖,𝑡 + 𝑆𝐷𝑖,𝑡 ) (1)
Here, t index shows the time, stated in hour, from 1 to the specified time interval and i index
shows the number of thermal generator units. 𝑈𝑖,𝑡 represents the status of unit i at time t. U will
be 1 if the generator is on and 0 if the generator is off. 𝑃𝑔𝑖,𝑡 represents the generation of unit i at
time t in MW. 𝑓(𝑃𝑔𝑖,𝑡 ) is the fuel cost of unit i which is a function of the generator power output
indicated by the quadratic function shown in equation (2).
𝑓(𝑃𝑔𝑖,𝑡 ) = 𝑎𝑖 (𝑃𝑔𝑖,𝑡 )2 + 𝑏𝑖 (𝑃𝑔𝑖,𝑡 ) + 𝑐𝑖 (2)
Here, ai, bi and ci represent the fuel cost coefficient of unit i.
𝑆𝑈𝑖,𝑡 and 𝑆𝐷𝑖,𝑡 is the cost of starting up and shutting down the generator i at time t,
respectively. If the status 𝑈𝑖,𝑡 is “1” and 𝑈𝑖,𝑡−1 is “0” then the 𝑆𝑈𝑖,𝑡 value is calculated by (3). If
the status 𝑈𝑖,𝑡 is “0” and 𝑈𝑖,𝑡−1 is “1” then the 𝑆𝐷𝑖,𝑡 value is calculated by (4).
𝑆𝑈𝑖,𝑡 = 𝐶𝑠𝑡𝑎𝑟𝑡 (𝑈𝑖,𝑡 − 𝑈𝑖,𝑡−1 ) (3)
𝑆𝐷𝑖,𝑡 = 𝐶𝑑𝑜𝑤𝑛 (𝑈𝑖,𝑡−1 − 𝑈𝑖,𝑡 ) (4)
Here, Cstart is start-up cost, and Cdown is shut down cost.

B. Power Balance Constraint

The operation of a power plant is always to be able to meet load requirements. With PV and
thermal generator, the amount of power from the thermal power plant and the power from the
PV must be the same as the load.
∑𝐼𝑖=1 𝑃𝑔𝑖,𝑡 + 𝑃𝑠𝑡 − 𝑃𝑑𝑡 = 0 (5)

611
Determination of Maximum Grid-Connected Photovoltaic Penetration

Here, 𝑃𝑔 denote the power of the thermal generator, 𝑃𝑠𝑡 denote the power of the PV generator
at time t, and 𝑃𝑑𝑡 denote the load demand at time t.

C. System Reserve Requirement

Spinning reserve (SR) is the generating capacity available from all generating units connected
in the system which is reduced by the load served. SR must be available in anticipation of a
sudden load change or loss of a unit. Spinning reserve constraint is represented as follow
∑𝐼𝑖=1 𝑃𝑔𝑚𝑎𝑥𝑖,𝑡 𝑈𝑖,𝑡 + 𝑃𝑠𝑡 ≥ 𝑃𝑑𝑡 + 𝑃𝑠𝑟𝑡 (6)
Here, 𝑃𝑔𝑚𝑎𝑥𝑖,𝑡 is maximum power limit of generator i, 𝑃𝑠𝑟𝑡 is the reserved power provided at time
t.

D. Thermal Unit Constraint

Each generator has specific characteristics, and this will be a limitation in the operation of
the generator. In thermal generators, there are limits on the form of maximum power, minimum
power, minimum uptime, minimum downtime, ramp up and ramp down.
1. Generation limit
Equation (7) shows maximal and minimum power constraint which is the power supply from
the generator must be smaller or equal to the maximum capacity and must be greater or equal to
the minimum capacity of the generator.
𝑃𝑔𝑚𝑖𝑛𝑖 ≤ 𝑃𝑔𝑖,𝑡 ≤ 𝑃𝑔𝑚𝑎𝑥𝑖 (7)
Here, 𝑃𝑔𝑚𝑖𝑛𝑖 is minimum power limit of unit i, 𝑃𝑔𝑚𝑎𝑥𝑖 is the maximum output of unit i, and
𝑃𝑔 𝑖,𝑡 is a power output of unit i at time t.

2. Minimum up and down-time

Generators also have a limit called minimum up-time (MUT) and minimum down-time
(MDT). MUT represents how long a generator is “on” before it can be turned off. On the
contrary, MDT represents how long a generator is “off” before it can be turned on. Equation (8)
and (9) show MUT and MDT constraints.
𝑇𝑜𝑛 𝑖 ≥ 𝑀𝑈𝑇𝑖 (8)
𝑇𝑜𝑓𝑓 𝑖 ≥ 𝑀𝐷𝑇𝑖 (9)
Where 𝑇𝑜𝑛 is the duration of unit i is “on”, and 𝑇𝑜𝑓𝑓 is the duration of unit i is “off”, 𝑀𝑈𝑇𝑖 is
minimum up-time of unit i, and 𝑀𝐷𝑇𝑖 is minimum down-time of unit i.

3. Ramp Constraint
Ramp up / ramp down constraint shows the ability of a generator to increase and decrease
power during a particular time.
𝑃𝑔𝑖,𝑡 − 𝑃𝑔𝑖,𝑡−1 ≤ 𝑈𝑅𝑖 (10)
𝑃𝑔𝑖,𝑡−1 − 𝑃𝑔𝑖,𝑡 ≤ 𝐷𝑅𝑖 (11)
Here, 𝑃𝑔𝑡 is the power of generator i at time t, 𝑃𝑔𝑡−1 is the power of generator i at time t-1,
𝑈𝑅𝑖 is the ramp-up limit of generator i, DRi is the ramp-down unit of generator i.

E. PV Power Constraint
The power generated from the PV unit depends on the natural conditions of the sun. The
maximum power limit of PV is limited by the installed capacity of the PV generator
𝑃𝑠 ≤ 𝑃𝑠𝑚𝑎𝑥 (12)
Here, 𝑃𝑠 is the PV generator power output, and 𝑃𝑠𝑚𝑎𝑥 is the maximum capacity of a PV
generator.

612
Penang Rahimta Mara, et al.

3. Proposed Method
In this paper, MIQP method uses CPLEX to get an optimal solution of generation scheduling
and dispatch. The quadratic programming is for thermal generator fuel cost functions that are in
the quadratic formulation. Constraints in MIQP are divided into two, namely equality constraints
and inequality constraints. Equality constraints in UC in the form of a balance of power between
load and generation. Inequality constraints in the form of spinning reserve, generation limit,
minimum uptime, minimum downtime, and ramp rate.
The simulation is carried out in two steps to obtain maximum PV penetration, as shown in
Figure 1. Furthermore, The object used is a system of 10 units of thermal generators [18]. System
parameters are shown in Figure 2 and Table I. Figure 2 shows the demand load profile and Table
I shows the parameters of the thermal generator. The amount of PV penetration is base on the
percentage of peak load with PV profile shown in Figure 2.
Step one, the purpose of step one is to find the limit of PV penetration with a 10% penetration
interval. First, input the demand load profile parameters and parameters of the thermal generator.
For initial conditions, simulations are carried out without PV penetration. Then, solve UC
problem by MIQP by adding PV penetration of 10%. If a solution is obtained, then PV
penetration is added by 10%. Nevertheless, if there is no feasible solution, the simulation is
stopped. Record the highest PV penetration that still has a UC solution and proceed to step two.

Start

Input load profile and PV

penetration profile. Initial
condition is no PV
penetration

Add PV penetration Solve UC Problem by

by 10 % MIQP

yes
Find feasible
solution

no

Input load profile and PV

penetration profile with last
feasible solutin.

Add PV Solve UC Problem by

penetration by 1 % MIQP

yes
Find feasible
solution

no

Find maximum PV
penetration

End

Figure 1. Flowchart of Proposed Method

613
Determination of Maximum Grid-Connected Photovoltaic Penetration

The second step, the purpose of step two is to find the limit of PV penetration with a 1%
penetration interval. Similar to step 1, first, input the demand load profile parameters and
parameters of the thermal generator. The simulation starts with the largest PV penetration in step
one that is still has a feasible solution. Do MIQP to find a UC solution with 1% PV penetration
interval. If a solution is still obtained, then proceed with increasing PV penetration by 1 %.
However, if MIQP failed to find UC solution then the simulation is stopped. The highest PV
penetration that simulation still has a feasible solution becomes the maximum limit of PV
penetration.

4. Simulation Result
In this study, a 10-generator power system is used for the simulation. Generator parameters
are shown in Table I. Generator G1 and G2 have large capacity with the most economical fuel
costs, and large ramp rates, but have the longest minimum uptime and minimum downtime.

Table 1. Generator Data

Unit1 Unit2 Unit3 Unit4 Unit5 Unit6 Unit7 Unit8 Unit9 Unit10
Pmin (MW) 150 150 20 20 25 20 25 10 10 10
Pmax (MW) 455 455 130 130 162 80 85 55 55 55
2
a ($/MW ) 0.0005 0.0003 0.002 0.0021 0.004 0.0071 0.0008 0.0041 0.0022 0.0017
b ($/MW) 16.19 17.26 16.6 16.5 19.7 22.26 27.74 25.92 27.27 27.79
c ($) 1000 970 700 680 450 370 480 660 665 670
RU (MW) 113.75 113.75 43.333 43.333 54 40 42.5 55 55 55
RD (MW) 113.75 113.75 43.333 43.333 54 40 42.5 55 55 55
MUT (hour) 8 8 5 5 6 3 3 1 1 1
MUD (hour) 8 8 5 5 6 3 3 1 1 1
Cstart ($) 9,000 10,000 1,100 1,120 1,800 340 520 60 60 60
C down ($) 0 0 0 0 0 0 0 0 0 0
T0 8 8 -5 -5 -6 -3 -3 -1 -1 -1

1500 1
1400
1300
Demand Load (MW)

1200 0.8
1100
PV Output (pu)

1000
900 0.6
800
700
600 0.4
500
400
300 0.2
200
100
0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time

Demand PV Output

Figure 2. Load Profile and PV Profile

Generator G3, G4 and G5 have medium-capacity with moderate fuel costs. Generators G3
and G4 have lower fuel costs than G5 but have a longer minimum uptime and downtime and a
lower ramp rate. Generator G6, G7, G8, G9, and G10 have small-capacity but have the highest
fuel cost, where generators G 6 and G 7 have lower fuel costs and ramp rates than generators
G8, G9, and G10, but minimum uptime and the downtime is longer.

614
Penang Rahimta Mara, et al.

The 24-hour load profile is shown in Figure 2, which the peak load is 1500 MW at hour 12
and lowest load is 700 MW at hour 1. Figure 2 also shows the PV power profile in pu. From the
Figure 2, it can be seen that the peak load demand and the peak output of PV occur at the same
time.

1500
1400
1300
1200
Load (MW)

1100
1000
900
800
700
600
500
400
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time

Without PV PV 68% PV 69%

Figure 3. Load profile with and without PV penetrstion

Result of step one, simulation shows that UC with PV penetration of 70% does have a feasible
solution. The current net load profile condition is the lowest net load is 450 MW at hour 12, the
peak load is 1400 MW at hour 20, and there is a sharp increase of 631 MW at hour 13 to 14.
Thus, the maximum PV penetration for step one is 60 %. Then, PV penetration 60% used for the
initial condition in step two.
Then, the simulation continues to step two to get more precise results. Simulations performed
for PV penetration of 60% to 70%. The results from step two show that the maximum PV
penetration is 68% because at PV penetration 69% failed occur when getting a UC solution. To
find out the cause of failure to get a feasible solution can be done by considering the net load on
68% and 69% PV penetration shown in Figure 3 and the characteristics of the generator shown
at Table 1.
In unit commitment, the problem-solution must respect both generator physical constraints
and system operational constraints namely power balance constraint, system reserves
requirement, generation limit, and ramp constraint. Failure of generation scheduling can occur
due to violation of constraints.

A. Power Balance Constraint.

Table 2. Power Increase/Decrease at PV Penetration 68% and 69% (A negative sign shows a
decrease of the load and a positive sign indicates an increase of the load
Hour PV 68 % PV 69% Hour PV 68 % PV 69%
1 0 0 13 44.20 46.32
2 50 50 14 609.71 620.14
3 100 100 15 -118.37 -118.64
4 100 100 16 -167.74 -168.00
5 50 50 17 123.57 126.13
6 100 100 18 129 129
7 -6.87 -7.71 19 100 100
8 -83.04 -84.99 20 200 200
9 8.20 6.85 21 -100 -100
10 -241.53 -246.56 22 -200 -200
11 -162.82 -165.95 23 -200 -200
12 -133.94 -136.65 24 -100 -100

615
Determination of Maximum Grid-Connected Photovoltaic Penetration

Power balance shows the ability of the generator to meet load requirements. PV penetration
of 68% and 69% have the same peak load. The peak load is 1400 MW at hour 20. With the
total maximum capacity of all generators is 1662, the peak load is still below the total
maximum capacity of the generator.

B. System reserves requirement

In this simulation system reserves is 10% of the load. With a peak load of 1400 MW, so the
reserve system needed is 140 MW. Sum of peak load and system reserves is 1540 MW. It is
lower than the maximum generator capacity of 1662 MW.

C. Generation limit
The generator units have upper and lower limits. To find the upper and lower limits that are
satisfied, the peak load and the lowest load simulation result condition is investigated. The peak
load is 1400 MW that occurs at hour 20. If the entire generator is on, the peak load is still below
the maximum capacity of all generators. The lowest load is 465 MW that happened at hour 12.
Minimum load is still higher than the minimum output capacity of the existing generator, so the
load is still possible to be supplied by a few generators.

D. Ramp constraint
Table 2 shows the load increase or decrease in every hour. A negative sign shows a decrease
of the load and a positive sign indicates a load increase. Form Table I, it can be seen that the total
ramp-up of the system is 620 MW and the total ramp down of the system is 620 MW. At PV
penetration 68 %, the highest ramp-up requirement of 609 MW. However, at 69 % PV
penetration, the load at hour 13 is 511.32 MW rises to 1131.46 MW at hour 14, which means the
load increases by 620.14 MW. The load increase is higher than the maximum ramp-up of the
system. So that, load changes cannot be met by the maximum ramp up. The most significant
decrease in load occurred at hour 10, with a decrease of 246 MW. It is smaller than the ramp
down capability.
Table 3 shows UC with 68% PV penetration. At hour 1 to hour 12, with medium to low net-
load and small to medium ramp-up and ramp-down, 2 to 5 generators are on in every hour, that
is generator G1, G2, G5, G6, and G7. At hour 13, even though the load was still small at 524.2
MW, all the generators turned on Table IV shows the output with 68% PV penetration. It can be
seen that at hour 13 almost all generator output is minimal except G1 and G2. All generators are
turned on because they are prepared for high ramp-up needs of 609.71 MW from hour 13 to hour
14. Because the generator power must be minimum in the first-hour generator is on. Furthermore,
it can be seen that at hour 14, almost all generators increase the power output to the maximum
ramp-up capability. In the following hours, the need for ramp-up / ramp-down was small. At
hour 15 with a load of 1015.53 MW, even though all generators are still on, but the power of
generator G4 to generator G10 is minimum. Beside to satisfy need of the load, these generators
are prepared to turn off at the following hour. Because the generator must be at a minimum
output before it can be turned off. Whereas when peak loads at hour 20 with a load of 1400 MW,
only eight generators are on which the output power of four generators is maximum and the
output ower of three generators is minimum.

616
 Penang Rahimta Mara, et al.

The generator with maximum output is G1, G2, G3, and G4 which are generators with the largest capacity and the lowest cost. The generator with
minimum output is G8, G9, and G10 which is the generator with the smallest MUD (1 hour) because G9 and G10 will only supply peak load and will be
off in the next hour.

Table 3. Unit Commitment at PV Penetration 68 %

Status
Generator
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
G1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
G2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
G3 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0
G4 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 0
G5 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0
G6 0 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 0 0 1 1 1 0 0 0
G7 0 0 0 0 1 1 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0
G8 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 1 0 0 0
G9 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 0 0 0 0
G10 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 0 0 0 0

Table 4. Generator Dispatch at PV Penetration 68 %

Generation (MW)
Generator
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
G1 455 455 455 455 455 455 455 455 455 420 374 320 247 361 455 455 455 455 455 455 455 455 455 455
G2 245 295 370 455 455 455 455 455 455 341 228 150 150 264 251 150 246 360 406 455 455 452 380 345
G3 0 0 0 0 0 0 0 0 0 0 0 0 20 63 106 101 116 130 130 130 106 63 20 0
G4 0 0 0 0 0 0 0 0 0 0 0 0 20 63 106 120 130 130 130 130 106 63 20 0
G5 0 0 25 40 65 145 139 83 92 25 25 25 25 106 25 25 25 25 59 140 148 67 25 0
G6 0 0 0 0 0 20 20 20 20 0 0 0 20 60 20 0 0 0 20 60 20 0 0 0
G7 0 0 0 0 25 25 25 0 0 0 0 0 25 68 25 0 0 0 0 0 0 0 0 0
G8 0 0 0 0 0 0 0 0 0 0 0 0 10 55 10 0 0 0 0 10 10 0 0 0
G9 0 0 0 0 0 0 0 0 0 0 0 0 10 55 10 0 0 0 0 10 0 0 0 0
G10 0 0 0 0 0 0 0 0 0 0 0 0 10 42 10 0 0 0 0 10 0 0 0 0

617
Determination of Maximum Grid-Connected Photovoltaic Penetration

From the discussion above, it can be concluded that the ramp-up constraint that caused the
failure to find a feasible solution at 69% of the PV penetration. At 68% PV penetration, at hour
13 the entire generator is on, not only to meet current load requirements but also to prepare for
ramp-up requirements that are very high in the following hours.
The UC simulation shows the total production cost. From Figure 4, it can be seen that higher
level of PV penetration, lower the total operating cost. The presence of PV reduces total
operating cost because the operating cost of PV is negligible. Although the total operating cost
decrease, it does not happen in thermal operating cost per MWh. Figure 4 shows that the
operating cost of thermal generator per MWh decrease at 10 % PV penetration until 40 % PV
penetration. Then increase at PV penetration is higher than 40 %. The lowest cost per MWh is
20.61 $/MWh at 40 % PV penetration.
Furthermore, the ten generators used in the simulation are divided into three categories based
on cost characteristics. This categorisation is to find out the cause of the difference in cost per
MWh for different PV penetration. The first category is the most economical generators that are
G1 and G2. The second are generators with moderate cost that are G3, G4, and G6. The other
five generators, G6 until G10, is in the last category. The last category is generator with
expensive cost. However, G6 and G7 are cheaper than G8, G9, and G10. Figure 5 shows the
comparison of energy share of the generator at PV penetration 10 %, 40, and 60 %. Figure 5
shows comparison of energy share by generator at PV penetration 10 %, 40 % and 60 % and
Table V shows comparison of the number of generator start-up that occurred in PV penetration
of 10%, 40% and 60%.
As shown in Figure 5, at PV penetration, 10% compares with PV penetration 40 % show that
energy that supplied by ten generators are in the same pattern but different percentages. Both at
PV penetration 10 % and 40 %, the highest supply form G1 and G2. Supply energy from G1 and
G2 at PV penetration 10 % supply is smaller than at PV penetration 40 %, that is 78,2 compared
to 83,9%. For the expensive generator, share energy supply at both conditions is almost same,
that is about 1,26 %. However, supply energy form the moderate generators, at PV penetration
10 % is much bigger than at PV penetration 60 %, that is 20,4% compared to 14,7 %. From
Table V, start-up of generator at PV penetration 10 % is less than start-up of generator at PV
penetration 60 %, that 9 times compared to 10 times. The number of generator start-ups is almost
same in all generator except G8. So, the operating cost of thermal generator per MWh at PV
penetration 10 % is higher than operating cost of thermal generator per MWh at PV penetration
40 % caused by energy that is supplied by G3, G4 and G5, that has moderate fuel cost, more
significant at PV penetration 10 %.

Figure 4. Operating cost with various PV penetration

21.2 580,000
570,000
21.1
560,000
Total Operating Cost ($)
Cost per MWH ($)

550,000
21
540,000
20.9 530,000
520,000
20.8
510,000
500,000
20.7
490,000
20.6 480,000
0% 10% 20% 30% 40% 50% 60%
PV Penetration

Cost Per MWh Total Operating cost

618
Penang Rahimta Mara, et al.

Table V. Comparison of Frequency of Generator Start-up

PV Frequency of Start-up
Penetration
G1 G2 G3 G4 G5 G6 G7 G8 G9 G10
10% 0 0 1 1 1 2 1 1 1 1
40% 0 0 1 1 1 2 1 2 1 1
60% 0 0 1 1 1 3 2 2 1 1

Figure 5 also shows the comparison energy share at PV penetration 40 % and 60 %. For the
most economical generator, there is a different pattern. G1 have bigger share at PV penetration
60 %, but G2 has bigger share at PV penetration 40 %. The sum of energy share by G1 and G2
is higher at PV penetration 60 % that is 85,1 % at PV penetration 60 % and 83,9 % at PV
penetration 60 %. Likewise, the energy share of the generator G3, G4, and G5 are higher at 40%
PV penetration, that is 14,7% compared to 11,8 %. However, the percentage of energy supplied
from the most expensive generator is higher at PV penetration 60%. The sum of energy supplied
by generator G6, G7, G8, G9, and G10 is 2,9 % at PV penetration 60 % and 1,2 % at PV
penetration 40 %. At PV penetration 40%, the start-up occurred 9 times, while at the PV
penetration 60%, the start-up occurred 12 times. G6 and G7 generators more often start-up at PV
penetration 60% that has an impact on higher start-up costs. So, the fuel cost per MWh at PV
penetration 60% is higher than at PV penetration 40% caused by the use of more generators with
more expensive costs that are G3, G4, and G5 generators, and higher start-up costs due to more
frequent occurrence start-up.
At PV penetration with higher operating costs per MWh, the more expensive generators being
used more widely. Also, in some cases, more start-up generators occur with the consequence of
higher start-up costs.

5. Conclusion
In this paper, the UC was solved using MIQP. UC solutions must be able to meet load
requirements and at the same time, respect to generator physical constraints and system
operational constraints. High PV penetration causes changes in net-load. The higher PV
penetration causes changes in load between the times to be more significant, especially during
the daytime. In the first step, it was found that failure to get a UC solution occurred at PV
penetration 70%. Which means, the maximum penetration that still gets the UC solution is when
the PV penetration is 60%. Then the second step starts with 60% PV penetration. In the second
step, at PV penetration 69 %, a failure in the generation scheduling occurs because of high ramp-
up that is higher than ramp-up capability of generators. So, the maximum PV penetration is 68%.
Furthermore, from the total operating cost results, it was found that generators operating cost per
MWh decrease from without PV until PV penetration 40%. Hereafter, at PV penetration is above
40%, generator operating cost per MWh increase. By observing to generator energy mix, it is
known that at the PV penetration with higher operating costs per MWh, generators with more
expensive costs are used more widely. Also, in some cases, more start-up generators occur with
the consequence of higher start-up costs.

6. Acknowledgment
This work was funded by the Rekognisi Tugas Akhir (RTA) Program of the Research
Directorate of Universitas Gadjah Mada. Any opinions, findings, and conclusions or
recommendations expressed herein are those of the authors and do not necessarily reflect those
of the Research Directorate of Universitas Gadjah Mada.

7. References
[1] M. Obi and R. Bass, “Trends and challenges of grid-connected photovoltaic systems – A
review,” vol. 58, pp. 1082–1094, 2016.

619
Determination of Maximum Grid-Connected Photovoltaic Penetration

[2] IRENA, “Solar Power,” 2019. [Online]. Available: https://www.irena.org/costs/Power-

Generation-Costs/Solar-Power.
[3] International Renewable Energy Agency, Renewable Capacity Statistic 2019. 2019.
[4] M. Cui, J. Zhang, C. Feng, A. R. Florita, Y. Sun, and B. Hodge, “Characterizing and
analyzing ramping events in wind power , solar power , load , and netload,” Renew. Energy,
vol. 111, pp. 227–244, 2017.
[5] P. Denholm, M. O’Connell, G. Brinkman, and J. Jorgenson, Overgeneration from Solar
Energy in California : A Field Guide to the Duck Chart, no. November. 2015.
[6] T. Ting, C. K. Loo, and S. S. Ngu, “Solving Unit Commitment Problem Using Hybrid
Particle Swarm Optimization,” J. Heuristics, vol. 9, pp. 507–520, 2004.
[7] A. Y. B. Eua-Arporn Sarjiya, “Short-term Operating Strategy with Consideration of Load
Forecast and Generating Unit Uncertainty Short-term Operating Strategy with
Consideration of Load Forecast and Generating Unit Uncertainty,” IEEJ Trans. Power
Energy, vol. 127, no. 11, pp. 1159–1167, 2007.
[8] R. Baldick, “The Generalized Unit Commitment Problem,” IEEE Trans. Power Syst., vol.
10, no. 1, pp. 465–475, 1995.
[9] N. P. Padhy, “Unit Commitment — A Bibliographical Survey,” IEEE Trans. Power Syst.,
vol. 19, no. 2, pp. 1196–1205, 2004.
[10] S. Y. Abujarad, M. W. Mustafa, and J. J. Jamian, “Recent approaches of unit commitment
in the presence of intermittent renewable energy resources : A review,” Renew. Sustain.
Energy Rev., vol. 70, pp. 215–223, 2017.
[11] M. Carrión and J. M. Arroyo, “A computationally efficient mixed-integer and linear
formulation and for the thermal and unit,” IEEE Trans. Power Syst., vol. 21, no. 3, pp.
1371–1378, 2006.
[12] L. Y. R. Quan, J. Jian, H. Zheng, “A Two-Stage Method with Mixed Integer Quadratic
Programming for unit commitment with Ramp Constraints,” in 2008 IEEE International
Conference on Industrial Engineering and Engineering Management, 2008, pp. 374–378.
[13] L. G. Papageorgiou and E. S. Fraga, “A mixed integer quadratic programming formulation
for the economic dispatch of generators with prohibited operating zones,” Electr. Power
Syst. Res., vol. 77, no. 10, pp. 1292–1296, 2007.
[14] J. Á. López, J. L. Ceciliano-Meza, I. Guillén Moya, and R. Nieva Gómez, “A MIQCP
formulation to solve the unit commitment problem for large-scale power systems,” Int. J.
Electr. Power Energy Syst., vol. 36, no. 1, pp. 68–75, 2012.
[15] A. M. Z. Alabedin, S. Member, S. Member, M. M. A. Salama, and A. Indices, “Generation
Scheduling in Microgrids under Uncertainties in Power Generation,” 2012 IEEE Electr.
Power Energy Conf. Gener., pp. 133–138, 2012.
[16] W. T. Jewell and T. D. Unruh, “Limits on cloud-induced fluctuation i photovoltaic
generation,” IEEE Trans. Energy Convers., vol. 5, no. 1, pp. 8–14, 1990.
[17] E. Y. Pramono and S. Isnandar, “Criteria for integration of intermittent renewable energy
to the Java Bali Grid,” Int. Conf. High Volt. Eng. Power Syst. ICHVEPS 2017 - Proceeding,
vol. 2017-Janua, pp. 91–94, 2017.
[18] V. P. S. a. Kazarlis, a. G. Bakirtzis, “A genetic algorithm solution to the unit commitment
problem,” EEE Trans. Power Syst., vol. 11, no. February, pp. 83–92, 1996.

620
Penang Rahimta Mara, et al.

Penang Rahimta Mara received his B.Eng in electrical engineering from

Universitas Gadjah Mada in 2005. He has worked at Ministry of Energy and
Mineral Resources Republic of Indonesia since 2008 and served as an electrical
inspector since 2014. He is currently pursuing a master degree in Electrical
Engineering Universitas Gadjah Mada.

Sarjiya received the B.Eng and M.Eng degree in Electrical Engineering from
Universitas Gadjah Mada in 1998 and 2001 respectively. He received his Ph.D
degree in 2008 from Department of Electrical Engineering, Chulalongkorn
University Thailand. He has been with Department of Electrical and
Information Engineering, Universitas Gadjah Mada since 1999. Currently, his
position is head of department and an associate professor in power system
engineering. His research interest includes power system operation and
planning, variable renewable energy integration and hybrid renewable energy
system.

Lesnanto Multa Putranto received a B.Eng. from Universitas Gadjah Mada,

Indonesia at 2007, an M.Eng. from Chulalongkorn University, Thailand at 2012
and a Ph.D. degree at Hokkaido University, Sapporo, Japan at 2017. He has
worked for Department of Electrical and Information Engineering at
Universitas Gadjah Mada since 2009. Currently, He is the head of Transmission
and Distribution Laboratory. His research interest is focused on analysis,
operation, and control of electric power systems. He is interested in renewable
energy resource development, real-time monitoring of power systems, and new
technology in power system engineering.

Muhammad Yasirroni received his B.Eng in electrical engineering from

Universitas Gadjah Mada in 2018. Currently, he is a research assistant in Power
Systems Research Group at the Department of Electrical and Information
Engineering, Universitas Gadjah Mada. He is interested in power systems
modelling and simulation.

621
 © 2019. This work is published under
https://creativecommons.org/licenses/by/4.0(the “License”).
Notwithstanding the ProQuest Terms and Conditions, you may
use this content in accordance with the terms of the License.