2.3.

2 Water pumping systems A study on the benefits of pump scheduling for potable water
suppliers was done by Bunn and Reynolds [97]. A SCADA or similar systems are installed in most
developed utilities and pump stations. These systems continuously log data of important operations.
Substantial historical data of operations has been gathered [97].

The focus is on pumping systems that supply potable water from the treatment plants. These
systems are dependent on systems such as SCADAs. These systems read, store and provide detailed
time-stamped historical data for parameters such as flow, pressure, and reservoir levels. Valuable
information can be extracted from the data to support decision making and the performance
improvement for these processes [97].

Literature Review 46 A number of methods have been developed to optimise scheduling of pump
run times. Most of the scheduling has been done to minimise energy consumption during peak
periods of ToU tariffs.

According to Bunn and Reynolds [97] it is considered impractical to explicitly solve the scheduling of
pumps mathematically, due to the complexity of most water distribution systems and production
constraints. Evolutionary algorithms have been developed to search for a solution of these complex
mathematical problems [97].

Evolutionary algorithms use the theory of evolution to apply random solutions to simulations to find
a better solution. The genetic algorithm, a derivative of the evolutionary algorithm, has been widely
used on pump scheduling problems. The genetic algorithm is well suited for binary problems, where
the pumps are either on or off [97].

Previously pump efficiency improvements focused on the static process of calibrating a pump curve,
assuming an operation point and then modifying the pump characteristics. However, many of these
solutions are not focused on the real-time optimisation of a distribution system. A real-time solution
to optimising pump scheduling is required as a pumping system is a very dynamic system [97].

A program has been developed to avoid using complicated genetic algorithms. It makes use of linear
programming and mixed integer programming due to the faster operating speed. The software is
able to explicitly calculate the efficiency of a pump based on the performance data, such as flow,
pressure and power. This approach reduces electricity costs by exploiting ToU tariffs and additionally
optimising the peak period operation, by operating only the most efficient equipment [97], [98].

The program can schedule pumps based on flow and pressure constraints, storage capacity and level
constraints, operating run times of pumps and cool down time. It can schedule the operation of
pumps with on/off control as well as for pumps with continuous control such as VSD-driven pumps
[97].

The program collects, automatically corrects and stores the data. The data is obtained from
instrumentation, telemetry and recreates missing data by using a hydraulic model of the system.
Literature Review 47 These specific data sets are then compiled into larger data sets which are used
in an optimisation algorithm [97].

The program identifies equipment which is not operating effectively. The program can therefore also
be used for equipment replacement assessments. To reduce electricity costs, a program is required
to automatically schedule pumps, taking into account predicted values to select the most efficient
pump combinations to be operated [97].
The program was implemented in a case study in California in 2006. The implementation yielded an
electricity cost saving of 12% and an improved efficiency of 9%. The improved efficiency was
achieved by optimising the operation schedules of cheaper but less efficient gas engine driven
pumps and more expensive but more efficient electrical driven pumps using a ToU tariff structure
[97], [99].

Tang et al.[100] investigated the issue of load shifting and energy efficiency. The study has been
treated as an efficiency improvement of the operation, since the cost of operation will decrease
when electricity costs decrease due to load shifting.

Previously the number of pump switches and total operation time were not considered in pump
station optimisation. In this study the general optimal control of pump stations were considered to
form an optimal control with consideration to pump switches and operating time. Mathematical
optimisation algorithms were implemented to solve these problems [100].

The formulation of the optimisation algorithms takes the ToU tariffs into account and the load is
consequently shifted out of the peak periods. Theoretically these control approaches can save nearly
30% on energy costs. The saving is achieved by shifting the operation of pumps into offpeak periods.
The control approach, however, was not implemented on a case study to verify the achievable
impact [100].

Similar studies have also been done by Lansey and Awumah [101] and McCormick and Powell [102].
Lansey and Awumah explored the multi-objective optimisation to evaluate an optimisation strategy
taking into account the number of pump stops and starts [101]. McCormick and Powell evaluated
the optimisation strategy by minimising the electrical demand Literature Review 48 charges during
peak periods. Both achieved significant electrical cost savings and peak period load reduction [102].

Moreira and Ramos [103] identified that the majority of the life cycle cost of a pump is spent on
electricity, approximately 90% in a 30-year lifetime [97]. Maintenance and purchase cost attributing
to the rest of the costs. A genetic algorithm was used to determine the best daily pumping schedule
[103].

By analysing and modelling different pump characteristics the most efficient operating point could
be achieved. The main goal is to analyse a system and simulate several pumping schedules in order
to reduce the electricity costs. Different tariff structures and different demands throughout the day
need to be considered. The simulation took into account the number of starts allowed per pump
[103]. A case study in Portugal was used as a base for the study. The system was fully analysed with
parameters such as [103]:

 Types of pump installed;

 Flows supplied by pumps;

 Overall efficiency of pumping system;

 Discharge head needed;

 Pipeline length;

 Reservoir capacities and elevations;

 Profile of water demand;

 Average water demand; and

The simulation model consisted of two scenarios. Two pumps with three starts and one pump with
six starts. A different pump and motor required for each scenario was selected based on the system
curves, operating curves and best efficiency point of the combination. The effect of fixed speed
drives and VSDs on the pumps was investigated. Small variations in the speed of the pumps can
translate to significant power consumption variations. This can increase the cost efficiency of the
system.

Literature Review 49 After implementing the energy efficiency component and the pump scheduling
it was possible to reduce the electricity costs by almost 44%. Significant savings could be achieved
with a manual override. During the manual override the primary criteria of operation was to avoid
operating the pumps in the peak periods and to prevent large variations in the reservoir level.

It is recommended that an optimisation model be done for more complex systems. Systems with
interaction between pumping stations would benefit from a detailed optimisation model using a
genetic algorithm. From the simulation it was also determined that a system with one pump allowing
for six starts was more efficient and versatile compared with a system with two pumps allowing for
three stops. It is recommended to implement a system which will allow real-time optimisation and
scheduling of a pump system.

Nortjé [21] implemented a load shift study on a water transfer scheme, which provides water to
power stations and Sasol entities in the Mpumalanga Highveld, South Africa. These Pump Stations
(PS) supply water to power stations such as Tutuka, Matla and Duvha. These facilities have large
storage capacities ranging from 885 Mℓ – 900 Mℓ. The pumping system supplies water to various
other users connected along the line [21].

Infrastructure and communication systems were installed at the pumping stations. These
infrastructure upgrades are paid for by Eskom DSM. The infrastructure upgrades included replacing
relay logic controls with PLCs, installation of HMIs and the introduction of a SCADA system.

Additional to the infrastructure installations, an EMS, similar to the system used in the mines, was
installed. The installations were necessary to allow for control via the EMS. Previously, plant
operators received information about reservoir levels and demand. The operator then had to make a
decision based on the information received, taking various other factors into account, whether a
pump needs to start or stop [21].

The EMS eliminates the operator as the SCADA communicates the necessary information directly to
the EMS. Based on a programmed algorithm the EMS can decide whether to start or stop a pump.
The peak periods are programmed into the EMS and subsequently the minimum necessary pumps
can be scheduled to operate during the peak periods [21].

Literature Review 50 Table 6 presents the load shift impact achieved by Nortjé. PS 1 and PS 2
achieved a combined impact due to the fact that these stations are interconnected. A combined
evening peak period impact of 12.6 MW was achieved on average at all the stations. This relates to a
financial saving of approximately R 4.765-million annually [21].
Slade [104] identified the success rate of DSM initiatives on the clear water pump stations in the
mining sector. A water distribution system has been identified as an ideal candidate for DSM. This is
due to the similarities shared between mine pumping systems and potable water distribution
systems. The specific water distribution system referred to is situated in the Northern Cape, South
Africa [104].

Slade identified that mine pumping systems and water distribution systems have similarities
including abstracting water from a source, high flow rates, high pressure heads, storage capacities
and large installed capacity of pumps. The pumps are used to transfer water to a higher elevation,
such that the water can gravitate to lower elevations. Large capacity pumps are required to
overcome the pressure head and supply high volume flow [104], [105].

The distribution system extracts water from a river, purifies the water and pumps the potable water
to a reservoir from where it is gravity-fed throughout the system. The distribution system consists of
one high lift pump station and two booster pump stations. Three high lift pumps and six booster
pumps are installed, each with an installed capacity of 780 kW.

Apart from the electricity consumption of a system, many other factors also need to be investigated.
These factors include [105]:

 The capacities of the reservoirs; Literature Review 51

 The minimum and maximum percentage levels of the reservoirs;

 The number of columns available;

 The minimum and maximum number of simultaneously operated pumps;

 The flow rates of the pumps; and

 The installed capacities of the pumps.

In many DSM initiatives it is preferred and often required to automate a system. It has been
observed in the DSM initiatives implemented in the mining sector that control room operators do
not adhere to intervention strategies. The human element prevents DSM initiatives to be
sustainable. An EMS can be implemented to ensure that the contractual DSM initiative is achieved
[104], [105].

If the information is available on the SCADA, the control room operator is able to see the reservoir
levels, pump status, and water flow. Based on this information the operator makes a decision to
start or stop a pump to control the level of the upstream or downstream reservoir. With the EMS it
is possible to replace the control room operator [105].

The EMS is able to directly communicate with the SCADA system via a common network. All the
factors that were investigated, as mentioned above, can be programmed as input parameters. The
EMS communicates with the SCADA and gathers information such as reservoir levels, pump statuses
and flow rates [104], [105].

Using mathematical models programmed into the EMS, the program can schedule the operation of
pumps. The schedule can be optimised in real-time to continuously insure that the best combination
is operated. The EMS communicates the schedule to the SCADA. The SCADA communicates the
command to the PLCs which will then start or stop a pump, depending on the command given by the
EMS [104], [105].

The EMS was implemented on a potable water distribution system. The distribution system is
obligated to supply a specified volume of water on a daily basis. This volume of water that needed to
be transferred, together with the reservoir levels, played a critical role in the simulation model to
determine the optimal operating schedule [104], [105].

Literature Review 52 The EMS proposed a schedule for the operation of the pump station. The
proposed schedule realised a morning peak period load shift impact of 3.6 MW and an evening peak
period impact of 3 MW. It was also discovered that the pump station was billed on an Eskom
Nightsave tariff structure [104], [105].

The feasibility of switching to a ToU tariff structure was investigated to fully benefit from the DSM
initiative. The ToU tariff offers a greater flexibility and saving potential for energy intensive users. By
switching to a ToU tariff structure a possible electricity cost saving of R 1.5- million could be achieved
based on the 2006/2007 tariffs [104], [105]. All the above mentioned case studies are proof that
load shift initiatives are plausible within the water distribution sector.

2.3.4 Energy efficiency

An estimated 30% of water is lost worldwide. A similar proportion of energy is also lost as a result.
Energy in water distribution systems are lost due to factors such as inefficient pumps, poor design of
pump stations, maintenance not done on equipment, old pipes with high head loss, excessive supply
pressures and inefficient operations [107].

Energy efficiency optimisation of water distribution systems can be as simple as reducing water
leakage, and as complex as demand prediction, pump system optimisation and reservoir storage
optimisation [93].

According to Feldman [107] energy efficiency improvements can be done by improvement of pump
station design, improvement of system design, installation of VSDs, efficient operation strategies and
leakage reduction.

Reducing leakage of a system will reduce the waste of energy. Reducing leakage will also decrease
the loss in pressure head of a system and increase the volume of water supplied to users. The most
effective method to reduce water leakage is by pressure modulation. Operating more pumps in the
off-peak period leads to an efficient operation. The pressure is increased by doing this during the off-
peak periods [93], [107].
This is in conflict with another optimisation strategy to dynamically modulate the pressure in a
system. The aim of pressure modulation is to reduce leakages by minimising the pressure in offpeak
periods. The solution for these conflicting strategies is to isolate distribution areas [93], [107].

Inefficient control and oversized systems cause inefficient pump stations. An opportunity for energy
efficiency exists as a result of the oversized systems. Many pump stations are oversized by more
than 20%. To control the flow of the pump station, throttling valves, pump speed adjustments and
bypass lines can be used. A large dissipation of energy, causing an increased pressure head, occurs
when flow is controlled by a throttling valve [93].

Pump speed control is achieved by VSDs. VSDs for pumps allow for operation with fixed flow and
variable pressure operation or with fixed pressure and variable flow operation. This reduces
Literature Review 55 the number of pump switches and reduces the likelihood of water hammer.
According to Gellings [108] VSDs can potentially save 10 – 20% of pump energy. Kiselychnyk [109]
indicates that a 10% speed reduction can realise a reduction of 27% in energy usage [93].

Where flow rate variation is not necessary, pump resizing is a more efficient intervention. Pump
resizing includes reduction of impeller diametres or even replacement of a pump. Other methods to
enhance the overall efficiency of a water distribution system include [93]:

 Replacing inefficient equipment;

 Managing leaks by regular monitoring;

 Selecting a suitable electricity tariff; and

 Incorporating renewable energy sources.

Some of these methods require high capital expenditure, such as replacement of equipment. Other
methods, however, require little to no additional costs, such as selecting suitable electricity tariffs. It
is possible to reduce the electricity costs by 50% by utilising a suitable tariff structure [93].

Vilanova and Balestieri [13] summarise state-of-the-art methods to energy and hydraulic efficiency
and conservational approaches in water supply systems. An overview of energy efficiency and
conservation alternatives is provided. The results of these selected methods are presented in the
picture.

FYI

According to JP de Wet
3 Use of VSDs on pumping systems Varying flow rates are often a requirement for pumping systems.
In many cases the maximum required flow for the system is used to select the pumps and pumping
configuration. On average, the pumping system only operates at a fraction of its maximum installed
capacity and some kind of flow control is necessary [59]. Several methods of flow control are
available on pumping systems of which the most commonly used ones are throttling, bypassing, on-
off control and VSD control. Figure 19 provide illustrations of the mentioned pump control methods.
The methods in the following example are applied on identical pumps, thus the only constant
variable are the flow. Figure 19:

Methods for pump flow control [59]. Throttling Bypassing On-off control VSD control Load
management through the utilisation of VSDs on water transfer schemes. 25 The resultant power
consumption for each of these methods differ. In Figure 20, the red arrows indicate the effect that
each flow control method has on the pump characteristic curve. The green square represents the
power consumption of the pump after the corresponding flow control method is applied. Figure 20:

Power consumption for flow control methods. (Adapted from [59]) The power consumption (P) of
each method can be calculated by multiplying the flow (Q) with the head (H). This corresponds with
the area between the x- and y-axes and the operating point of the pump. The relative power
consumption of each method then calculates to:[59],[60]  No Control: P = 15 x 15 = 225  Throttling:
P = 12 x 17.3 = 207.6  Bypassing: P = 9.5 x 17.1 = 162.5  On-off control: P = (60% x 225.0) + (40% x
0.0) = 135.0  VSD Control: P = 9 x 8.3 = 74.7 Load management through the utilisation of VSDs on
water transfer schemes. 26 From the above results, it is clear that VSD control is the most energy
efficient method for flow control on centrifugal pumps. The pump affinity laws explain this energy
saving through speed reduction. These laws state that the power of a pump varies in proportion to
the cube of the speed. Therefore, reducing the speed of a pump will result in a power reduction
equal to the cube of the speed change. This means that a pump running at 70% of its maximum
speed will consume (0.7)3 = 34% power [61]. In addition to power savings, VSDs also provide the
following benefits to a pumping system: Reduced maintenance through smooth starting [61] - VSDs
function as soft starters for pumps, eliminating the high peak current associated with conventional
starting methods. This reduces stress on the electrical system and also reduces the mechanical stress
on the motor and pump due to lower start-up torque. Improved system reliability and
performance[49],[62] - Reduced pump wear, especially on pumps and seals, is a direct result of any
speed reduction by a VSD. Small process variations can quickly be matched to the process
requirements in a system by a VSD, which improves the overall performance of the system. Reduced
risks of water hammering [59] - The main cause of water hammer is rapid changes in flow. This result
is a pressure wave travelling inside the pipelines which can damage the pipes, valves and pipe
supports. VSD allows gradual flow reduction to limit water hammering. Reduced risk of cavitation on
pumps [59] - Cavitation occurs in pumps when the static pressure drops below the liquid vapour
pressure, causing vapour bubbles to form. These vapour bubbles then collapse, releasing high
amounts of energy that knock small pieces of metal from surfaces inside the pump. Figure 21 show
resultant damage that cavitation has on a pump impeller. With a VSD the suction pressure of a pump
can be managed to prevent this drop in pressure.