1. No category

African Utility Week, 12-14 May 2015 (Cape Town), Technical Workshop
Bridging the Power Gap in Africa:
Testing options to manage constrained short to medium term system
operations
Jarrad G. Wright
Craig L. Hart
Senior Consultant
Consultant
Energy Exemplar (Africa)
Energy Exemplar (Africa)
Johannesburg, South Africa
Johannesburg, South Africa
[email protected]
[email protected]
Abstract—Although significant historical investment in power
system capacity placed African countries in a strong position
to meet increasing demand, this capacity has been absorbed by
significantly increased levels of demand as a result of economic
growth and intensive electrification programmes. Even though
long term capacity expansion is planned, system operators
face increasingly constrained system operations in the short to
medium term with available capacity. Adding to this, ageing
capacity is requiring increased levels of maintenance with increasing outage durations. Performing the required maintenance is not
always possible as a result of the abovementioned constraints
resulting in further increases in maintenance required as well as
increased levels of forced outages. This paper presents options
for system operators to use in an attempt to prevent and stop
what seems like a downward spiral of system adequacy in
the short to medium term. Using a sufficiently detailed but
R
generic model developed in PLEXOS,
a number of supply
and demand side options are presented and analysed to present
their effect on a constrained power system. These options include
non-selective load-shedding, Demand Side Participation (DSP),
introduction of country/regional time zones, labour shifts in
selected sectors of the economy, leveraging existing but unused
distributed generation, fast-tracked emergency power generation,
strategic energy storage in adjacent countries/regions and the
effects of security constrained operations.
I. N OMENCLATURE
MIP Mixed Integer Programming
QP Quadratic Programming
LP Linear Programming
DSP Demand Side Participation
COUE Cost of Unserved Energy
IEA International Energy Agency
OCGT Open Cycle Gas Turbine
PV Photovoltaic
CSP Concentrated Solar Power
II. I NTRODUCTION
It is well known that capacity expansion plans have been developed at a country and regional level throughout Africa [1–
5]. Despite these plans being developed, it seems like there
is a disconnect between these plans and the implementation
thereof as system operators face increasingly constrained
system operations. In recent years, examples of this in all
African regions include Ghana [6], Nigeria [7], Botswana [8],
Namibia [9], South Africa [10], Tanzania [11], Ethiopia [12]
and Egypt [13] to name a few. Contributing to the constrained
system operations experienced by system operators is the
reduced levels of planned maintenance performed on ageing
generation capacity for various reasons including available
human resources, financial constraints and pressures to supply
end-users at all times (even if to the detriment of the generation
capacity requiring maintenance). This means that performing
required maintenance is not always possible which results
in increased levels of forced outages as generation capacity
reliability degrades.
This paper presents an analytical assessment of options for
system operators to use to stop/reduce what seems like a
downward spiral of system adequacy in the short to medium
term. A model is developed which assesses a number of supply
and demand side options for system operators to use to help
weather the storm in the short to medium term. These options
include:
a Demand side:
i Non-selective load-shedding
ii Demand Side Participation (DSP)
iii Introduction of country/regional time zones
iv Introduction of labour shifts
b Supply side:
i Improvement in generation availability
ii Leveraging existing but unused distributed generation
iii Fast-tracked emergency power generation
iv Strategic energy storage in adjacent countries/regions
v Security constrained operations
The paper is structured to include an overview of the model
developed, the presentation of the effect each supply and
demand side option can have followed by a concluding section.
III. T HE M ODEL
The model used for the analysis has been developed in
R PLEXOSis
R
PLEXOS.
a best in class software tool used
Page 1 of 8
African Utility Week, 12-14 May 2015 (Cape Town), Technical Workshop
around the globe primarily for energy market optimisation and
analysis in short, medium and long term.
The model developed uses mathematical programming (
Quadratic Programming (QP) and Mixed Integer Programming
(MIP)) focusing on short to medium term constrained system
operations. In one fully integrated environment, the model
includes the optimal placement of maintenance events (excluding forced outages), decomposition of any medium-term
constraints to short-term constraints and a complete production
cost model including unit commitment and economic dispatch.
The analysis is performed over a representative one year time
horizon to assess the effects of the various supply and demand
side options in the short to medium term that the system
operator could use to mitigate constrained system operations.
Hardware/software configuration to run the required simulations include a 64-bit architecture in a Windows environment
R Xeon
R E5-2670 processors operating at
with two Intel
2.6 GHz (8 cores each) and 100 GB RAM.
The supply side of the model includes a mix of generation
capacity technologies including steam turbines, Open Cycle
Gas Turbines (OCGTs), hydraulic turbines, wind turbines,
solar Photovoltaic (PV) and Concentrated Solar Power (CSP).
These technologies are fired by a number of fuels including
coal, nuclear, gas/diesel, water, solar and wind. Each generation technology is modelled with specific assumed technical
and financial parameters as summarised in Appendix A. Hydro
generation capacity is energy constrained on a monthly basis
based on an assumed annual rainfall seasonality profile. Maintenance parameters chosen have been intentionally increased
from typical maintenance rates and forced outage rates in order
to present a constrained system. A supply side with installed
capacity of 46 711 MW (rated capacity of 43 119 MW) has
been modelled with a total of 598 units and unit sizes ranging
from 1.3 MW to 900 MW.
The demand side of the model is built up using sector
specific participation in the overall demand profile as well as
daily, weekly and seasonal variations. Pertinent characteristics of the demand model are summarised in Appendix A.
The annual seasonality of the demand model is shown in
Figure III while the hourly demand profile over a typical
week is shown in Figure III. As can be seen, demand is
dominated by the industrial sector dominates with residential,
mining an commercial sectors having quite similar shares
while the agricultural sector has a small demand requirement.
The demand side has a peak demand of 35 850 MW and
minimum demand of 19 775 MW with an annual energy
requirement of 247 141 GWh.
IV. R ESULTS AND A NALYSIS
A. Non-selective loadshedding
Non-selective loadshedding is performed by a system operator when there is insufficient generating capacity available to
meet demand. Under these circumstances, the system operator
is forced to shed a certain amount of demand to ensure that the
supply-demand balance is maintained. The model that has been
developed includes a balancing parameter (unserved energy)
that simulates loadshedding as the last option available to the
Figure 1.
Demand model of daily energy per sector of the economy
Figure 2. Demand model of typical weekly profile per sector of the economy
system operator to maintain the supply-demand balance. it is
automatically invoked by the optimisation when the supplydemand balance cannot be maintained. Unserved energy is
penalised heavily in the objective function by a parameter
known as Cost of Unserved Energy (COUE). The COUE can
be interpreted to be the value placed on a unit of energy not
supplied and can be sector specific. There are many detailed
methods available to determine COUE but an estimate of
USD 7 000/MWh is used in this model for all sectors.
The annual supply-demand balance of the model is shown
in Figure 3 where the effect of constrained system operation
is clearly seen via the unserved energy required to maintain
the supply-demand balance. The supply stack for a typical
week is given in Figure 4 where the available generation
capacity is shown along with all supply resources against
the demand. As can be seen, nuclear and coal generation
capacity are dispatched first as base-load. Hydro capacity is
energy constrained and thus provides mid-merit and sometimes
peaking capacity while renewables are dispatched at their
rated capacity as they are self-dispatched (based on assumed
wind and solar profiles). The thermal (gas/diesel) generation
capacity is typically dispatched as peaking capacity but due
to the constrained nature of the system it is dispatched as
both mid-merit and peaking generation capacity. The effect
of not having sufficient available generation capacity is clear
where loadshedding is required at a number of times during
the week shown. It should be noted that even when there
appears to be sufficient available generation capacity to meet
the demand, certain technical constraints of the available
generation capacity necessitate loadshedding.
Page 2 of 8
African Utility Week, 12-14 May 2015 (Cape Town), Technical Workshop
Figure 5.
Figure 3.
Figure 4.
Annual supply-demand balance (non-selective loadshedding)
Supply stack for a typical week (non-selective loadshedding)
B. Demand Side Participation (DSP)
Demand Side Participation (DSP) is an alternative to peaking generating capacity as it is the reduction of demand
by certain end-users (compensated accordingly) when the
power system supply-demand balance is constrained. Many
governments, utilities and energy service companies around
the world have implemented successful DSP programs with
varying levels of end-user participation including International
Energy Agency (IEA) countries like Spain, Denmark, USA,
Canada, Sweden, Germany and Netherlands being very active [14] while local African examples include Eskom (South
Africa) [15] and Nampower (Namibia) [16].
DSP is included in the model by defining a certain level
of capacity, number of hours of operation per day, number of
activations per day and number of activations per year. DSP
is included with a total capacity of 1 000 MW. The number
of activations per day is one, number of hours per operation is
two hours while the number of activations per year is assumed
to be 150. The cost of activating DSP is assumed to be slightly
higher than the typical large end-user tariff level (so as to
incentivise the participation in DSP).
Similar to Figure 4, Figure 5 shows a typical week of operations but includes DSP. As can be seen, DSP assists the system
operator at times of peak demand and reduces the amount of
loadshedding required to maintain the supply-demand balance.
However, the constraints of only being allowed to operate DSP
once a day, for 2 hours and activating it 150 times throughout
the year are clear as significant loadshedding is still required.
Supply stack for a typical week (with DSP included)
C. Time Zones
Constrained operations in a specific country with one time
zone could be assisted by the introduction of time zones.
This is relatively easy to do (even if only temporary until
other capacity is available or existing generation performance
improves).
In the model, certain components of demand in the model
is shifted by 2 hours to assess the effect on the constrained
system. The share of demand in each of the two time zones
is as shown in Table I.
The resultant supply for a typical week is shown in Figure 6.
As can be seen, there is a flattening of the demand profile as
a result of the shifting in demand by 2 hours. It is interesting
to note that peak demand is reduced from 35 392 MW to
34 653 MW (739 MW) while minimum demand increased
slightly (as expected) from 20 253 MW to 20 605 MW
(352 MW) even though the same energy is supplied over the
course of the year.
Table I
D EMAND SHARE PER TIME ZONE
Industrial
Mining
Commercial
Agricultural
Residential
Share (%)
(0 hrs)
Share (%)
(+2 hrs)
23
7
23
34
30
77
93
77
66
70
D. Labour Shifts
An option that could be available to a country experiencing
significantly constrained system operations is the introduction
of labour shifts in certain sectors of the economy that do
not typically have labour shifts i.e. the commercial sector.
Although extreme, this may assist significantly in managing
the demand profile and allowing for the system operator to
better serve end-users.
The model tests the above by introducing three eight hour
labour shifts in the commercial sector of the economy with
the residential load profile assumed to follow similarly. The
mining and industrial sector are assumed to remain as-is
Page 3 of 8
African Utility Week, 12-14 May 2015 (Cape Town), Technical Workshop
more available capacity and thus no loadshedding in the week
presented in Figure 9. Although what seems like a simple solution to a constrained system, the improvement of generation
capacity performance by 10% requires significant planning and
investment in labour and capital by the generating capacity
owners and thus the effort required in pursuing this solution
to relieve a constrained system should not be underestimated.
Figure 6.
Supply stack for a typical week (2 time zones)
(they already use labour shifts) while the agricultural sector is
assumed to be inflexible to labour shifts.
The resultant supply for a typical week is shown in Figure 7.
The significant change in the demand profile is clear with
the demand profile flattening significantly and a reduction of
the system peak demand from 35 392 MW to 32 964 MW
(2 428 MW). As can be seen in Figure 7, this results in less
loadshedding but it should be noted that the flatter demand
profile does increase the minimum demand significantly and
thus reduces the opportunity to perform maintenance in what
were lower demand periods when labour shifts were not
introduced.
Figure 8.
Annual supply-demand balance (improved plant availability)
Figure 9. Supply stack for a typical week (with improved plant availability)
Figure 7.
Supply stack for a typical week (with labour shifts)
E. Improvement in generation availability
In the developed model, the amount of generation capacity
available to meet demand is relatively low (compared to the
installed capacity) as a result of degraded generating capacity
performance i.e. higher maintenance rates and forced outage
rates. With improved generation capacity performance, there
would be more available capacity to meet demand and thus a
less constrained power system. In order to test this, maintenance rates and forced outage rates were reduced by 10%.
The annual supply-demand balance is shown in Figure 8
and can be contrasted to Figure 3 where degraded plant
performance resulted in significant loadshedding throughout
the year. The supply for a typical week is shown in Figure 9.
The improved generation capacity performance allows for
F. Unused distributed generation
Unused distributed generation is typically sourced from
standby small generators typically used by end-users as backup for when an outage does occur on their premises/plant/site.
Under the circumstances of a constrained power system, the
system operator may be able to use these distributed standby
generators to supply power at certain times throughout the
year.
To simulate the use of unused distributed generation, it is
assumed that these generators can be operated once a day for
4 hours at a time with a maximum of 150 operations per year
(1000 MW is assumed to be available for the system operator
to use). The cost of using this capacity is assumed to be in
line with typical diesel engine technology and fuel.
The supply for a typical week is shown in Figure 10. As
can be seen, the distributed is used at times when the system
is constrained but can only be used for 4 hours at a time
an can only be used 150 times throughout the year (thus the
reason for not being used for all days of the week resulting
in loadshedding).
Page 4 of 8
African Utility Week, 12-14 May 2015 (Cape Town), Technical Workshop
Figure 10.
generation)
Supply stack for a typical week (with unused distributed
Figure 12. Weekly energy over the year and effect of emergency generation
on supply-demand balance
G. Emergency generation
From the perspective of a system operator, emergency
generation would play a similar role to the unused distributed
generation with the major difference being that there is an
associated lead time in procuring emergency generation. Thus,
the constrained system operations will continue until the
emergency generation can be installed.
For the analysis, two types of emergency generation are assumed (small and large). Small emergency generation totalling
300 MW is installed in increments of 100 MW with an initial
lead time of 6 months followed by 2 month intervals. Large
emergency generation of 1000 MW is assumed to have a lead
time of 8 months. This is feasible as outlined in [17] and
many service providers are available to provide this type of
fast tracked emergency generation capacity.
An annual view of the phasing in of emergency generation
is shown in Figure 12 where the lead time is clearly seen
(emergency generation only generates from 6 months into the
year). The supply for a typical week is shown in Figure 11. The
inclusion of the emergency generation helps significantly as it
can generate when required by the system operators (compared
to standby generation and DSP with their inherent constraints).
capacity to send power to the adjacent country, allowing the
country to reduce generation from their hydro capacity and
keep energy in storage reservoir(s). When the system operator
then requires generation capacity at times of system peak, the
energy ”stored” in the adjacent country can be called upon to
supply demand.
Storage capacity of 1000 MW is assumed and the model is
left to determine the appropriate energy storage requirement
and the effect on constrained system operations. As can be
seen in Figure 13, the envelope of energy storage operation
throughout the day shows how the storage charges overnight
and before the evening peak while discharging for the morning
and evening peaks (as expected). A similar envelope is shown
in Figure 14 for the energy stored on a daily basis. As can be
seen, the energy requirement is in the range of 3 600 GWh
(just over 3.5 hours at full capacity) and peaks in the mornings
while discharging for the rest of the day.
Figure 13.
Figure 11.
Daily range of strategic storage operation
Supply stack for a typical week (with emergency generation)
I. Security constrained operations
H. Strategic storage
A system operator could use a storage medium in an
adjacent country effectively as a pumped storage generating
plant. When system demand is low (overnight, early morning
and weekends), the system operator can use excess generating
Typically, system operators operate a power system taking into consideration certain key contingencies (generators/transmission lines) and operating the power system in a
manner that ensures the system unit commitment and energy
dispatch meets pre and post-contingency constraints. In the
Page 5 of 8
African Utility Week, 12-14 May 2015 (Cape Town), Technical Workshop
Figure 16. Example of monthly generation from pool of security constrained
generators (no constraint)
Figure 14.
Daily range of strategic storage energy requirement
model developed, security constrained operation is simulated
via the grouping of a number of hydro and gas/diesel generators and constraining their generation to their available
capacity less a large contingency in the same zone of the
system as these generators in order to ensure demand can still
be met in that zone. The contingency assumed is the loss of
a nuclear unit of 900 MW.
The result of security constrained operation is clear in
Figure 15 and Figure 16 (for an example one month period in
the year). The group of generators chosen is limited to their
available capacity less the 900 MW contingency previously
mentioned resulting in a further constrained system (increased
levels of loadshedding required). From this representative
example, allowing for concessions on security constrained
system operations by the could help to maintain the supplydemand balance even if only applied in the short to medium
term until new generation capacity or demand side alternatives
are brought online.
Figure 15. Example of monthly generation from pool of security constrained
generators (constrained)
V. C ONCLUSIONS
In the context of the challenges faced by system operators
to balance supply and demand in increasingly constrained
power systems a model has been developed to assess various
supply and demand side options to improve short to medium
term operational adequacy. Using a model built The model
developed was based on mathematical programming (QP and
MIP) and included the optimal placement of maintenance
events, decomposition of medium-term constraints to shortterm constraints and a complete production cost model including unit commitment and economic dispatch in one fully
integrated environment.
On the demand side, the effect of not having sufficient
available generation capacity is clear when loadshedding is
required to balance supply and demand by the system operator
(an undesirable result). DSP can assist at times of system peak
but is limited in that it can only be activated for a pre-defined
number of hours per day and per year. A novel solution which
could be relatively easily to implement is the introduction of
time zones into a country in order to shift demand and attempt
to reduce the peaks present in the demand profile. A reasonable
reduction in system peak was possible with two time zones two
hours apart. A more extreme option being the introduction of
labour shifts in sectors of the economy where shifts were not
already in place (commercial) resulted in a significant change
in the demand profile and significant reduction in system peak.
On the supply side, the use of standby distributed generation
capacity by the system operator (typically by end-users as
backup capacity) allowed for an improvement in the system
but with limited hours of operation per day and per year
(similar to DSP) this does not assist in providing mid-merit
or base-load capacity. The fast tracking of emergency power
generation was also tested and shown to be effective when introduced (as it could provide mid-merit or base-load power as
required) but with the disadvantage of a lead time required (812 months at least). The novel usage of a neighbouring country
as an effective pumped storage station was also shown to be
able to manage peak demand as energy can be stored in the
neighbouring country during off-peak times and used during
times of system peak. Finally, a brief demonstration of the
effect of security constrained system operation was presented
where a group of generators’ generation was constrained in
order to cater for a the possibility of a system contingency in
a specific zone of the system.
R EFERENCES
[1] East African Power Pool (EAPP) - SNC Lavalin Inter-
Page 6 of 8
African Utility Week, 12-14 May 2015 (Cape Town), Technical Workshop
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
national Inc in association with Parsons Brinckerhoff,
“Regional Power System Master Plan and Grid Code
Study (EAPP),” East African Power Pool (EAPP), Tech.
Rep., 2011.
Economic Community of West African States
(ECOWAS) - Tractabel Engineering (GDF Suez),
“Update of the ECOWAS Revised Master Plan for the
generation and transmission of electrical energy,” Tech.
Rep., 2011. [Online]. Available: http://www.ecowapp.
org/?page id=136
South African Department of Energy, “INTEGRATED
RESOURCE PLAN FOR ELECTRICITY 2010-2030,”
Tech. Rep., 2011. [Online]. Available: http://www.
doe-irp.co.za/content/IRP2010 promulgated.pdf
Electricity Control Board (ECB), “National Integrated
Resource Plan (NIRP) Republic of Namibia,” Tech.
Rep., 2011. [Online]. Available: http://www.ecb.org.na/?
page id=321
Ministry of Energy and Minerals, “Power System Master
Plan: 2012 Update,” Tech. Rep., 2013. [Online]. Available: http://www.tanesco.co.tz/index.php?option=com
docman&task=cat view&gid=39&Itemid=172
Ghana Grid Company (GRIDCo), “Electricity
Supply Plan 2014,” Tech. Rep., 2014. [Online]. Available: http://www.gridcogh.com/media/photos/
forms/supplyplan/2014 Electricity Supply Plan.pdf
A. S. Aliyu, A. T. Ramlia, and M. A. Saleh, “Nigeria
electricity crisis: Power generation capacity expansion
and environmental ramifications,” Energy, vol. 61, 2013.
Botswana Daily News, “Botswana: BPC Readies for
Power Supply Challenges,” 2015. [Online]. Available:
http://allafrica.com/stories/201504203026.html
ESI-Africa,
“Namibia
faces
electricity
supply
challenge
until
2018,”
2014.
[Online].
Available:
http://www.esi-africa.com/
namibia-faces-electricity-supply-challenge-until-2018/
South African Government News Agency, “War
room to address electricity challenges,” 2014. [Online]. Available: http://www.sanews.gov.za/south-africa/
war-room-address-electricity-challenges
ESI-Africa, “Tanzania faces 18 months of power cuts,”
2013. [Online]. Available: http://www.esi-africa.com/
tanzania-faces-18-months-of-power-cuts/
AllAfrica.com, “Ethiopia: The Power Man Speaks of
Power Cuts,” 2014. [Online]. Available: http://allafrica.
com/stories/201408250278.html
Middle
East
Eye,
“Egypt’s
power
outages
continue
to
intensify,”
2014.
[Online]. Available: http://www.middleeasteye.net/news/
egypts-power-outages-compound-559103879
International Energy Agency Demand Side Management
Programme, “International Database on DemandSide Management Technologies and Programmes
(INDEEP) Analysis Report,” IEA, Tech. Rep., 2004.
[Online]. Available: http://www.ieadsm.org/Publications.
aspx?ID=26
Eskom Holdings SOC Limited, “Integrated Demand
Management (IDM),” 2015. [Online]. Available: http://
www.eskom.co.za/sites/idm/Pages/Home.aspx
[16] Namibia Power Corporation (Proprietary) Limited
(Nampower), “Demand Side Management (DSM),”
2015. [Online]. Available: http://www.nampower.com.
na/Page.aspx?p=168
[17] R. Lilley, “Power ships: a real solution to
South Africas shortterm energy needs,” 2015.
[Online].
Available:
http://www.ee.co.za/article/
power-ships-real-solution-south-africas-short-term-energy-needs
html
A PPENDIX A
M ODEL PARAMETERS
R
The parameters used in the PLEXOSmodel
developed are
shown for the supply and demand side in Table II and Table III
respectively.
Page 7 of 8
African Utility Week, 12-14 May 2015 (Cape Town), Technical Workshop
Table II
S UPPLY MODEL SUMMARY
Installed Capacity (MW)
Rated Capacity (MW)
Units
Unit sizes (MW)
MSLa (%)
MUTb (hrs)
MDTc (hrs)
Ramp rate (%/min)
Heat rate (USD/MWh)
Fuel price (USD/GJ)
VO&M (USD/MWh)
Planned Maintenancef (%)
Planned Maintenanceg (%)
Forced Outage Rate (%)
Hydro
Nuclear
Renewables
Thermal
(gas/diesel)
Coal
Overall
2 228
2 228
26
2.1-333
33
50
12.6-14.2
2.9-3.4
6.0-12.9
1 800
1 800
2
900
50
168
168
5
11.1
0.5-0.6
1
9.6
0.5
4.7
1 474
1 474
419
1.3-138
-
2 005
1 984
35
9.7-167.5
50
2
1
20
9.9-14.7
7.2-7.5d , 28.9-31.9e
2.9-27.9
0.9-3.4
2.2-2.3
7.4-10.3
39 204
36 857
116
30-800
40
12
16
2
9.5-13.8
1.4-1.5
0.3-1.9
7.2-13.0
2.5-4.1
7.2-15.2
46 711
43 119
598
1.3-900
-
a MSL = Minimum Stable Level
b MUT = Minimum Up Time
c MDT = Minimum Down Time
d Natural gas
e Diesel
f Long duration
g Short duration
Table III
D EMAND MODEL SUMMARY
Peak demand
(MW)
Minimum demand
(MW)
Industrial
Mining
Commercial
Agricultural
Residential
15
4
7
1
11
556
940
810
235
111
9 819
4 693
1 077
617
2 695
Overall
35 850
19 775
Page 8 of 8
Energy (GWh)
110
42
36
7
49
Energy share
(%)
550
010
775
888
917
44.7
17.0
14.9
3.2
20.2
247 141
100