reduces the amount of wind power that cannot be absorbed. Index Termsâ storage, wind energy, spinning reserve, standing reserve, CO2 emissions, fuel costs, ...
1
The Role of Storage in Integrating Wind Energy Mary Black, Goran Strbac, Member, IEEE, Vera Silva, Member IEEE,
Abstract—This paper presents a new methodology to quantify the value of storage in the integration of intermittent resources, mainly wind. The main goal of the work is to provide quantified estimates of the potential value of storage, in managing the intermittency of wind generation, in the context of the future United Kingdom (UK) electricity system. In an electricity grid with large wind penetration additional system balancing costs are incurred due to the intermittent nature of wind. We developed studies to evaluate the benefits of using storage for providing standing reserve, as part of the overall reserve needs, in terms of savings in fuel cost, CO2 emissions and conventional energy. These studies were conduced considering a number of generation systems characterized by different mixes of generation technologies, representative of the size of the UK system with different levels of wind penetration. From these studies we were able to conclude that providing a greater part of the increased reserves needed, from standing reserve, in the form of pumped hydro storage, increases efficiency of system operation and reduces the amount of wind power that cannot be absorbed. Index Terms— storage, wind energy, spinning reserve, standing reserve, CO2 emissions, fuel costs, conventional energy.
I. INTRODUCTION Due to environmental concern over CO2 emissions, it is expected that penetration of intermittent renewable resources into the electricity grid will increase in future years. This has raised concerns over system costs, focused on whether these new generation technologies will be able to replace the capacity and flexibility of conventional generating plant. As intermittency and non-controllability are inherent characteristics of renewable energy based electricity generation systems, the ability to maintain the balance between demand and supply has been a major concern. In [1] an analysis of the breakdown of the total additional system costs incurred when extending renewable generation to 20% or 30% of demand is presented. This analysis included costs of balancing and capacity, transmission, and distribution. This work demonstrated that balancing and capacity costs, principally the cost of maintaining system security, dominate all other costs. The concern about these costs arises because the intermittency of wind creates a large wind forecast error which leads to imbalances between the scheduled generation supply and the electricity demand that need to be met in real time. In [2] we can find a pioneer study of the use of storage in
the reduction of fuel cost savings. The general approach of some of the work presented in this paper is in line with this study. Bulk energy storage systems such as large-scale pumped storage appear to be an obvious solution to deal with the intermittency of renewable sources and the unpredictability of their output. During the periods when intermittent generation exceeds the demand, the surplus could be stored and then used to cover periods when the load is greater than the generation. The use of storage as an enabling technology to increase wind integration on power systems is being object of recent research. In [3] we can find a detailed study of the future value of storage in the integration of intermittent resources in the UK system. Barton & Infield, in [4] presents a study based on probability methods to predict the availability of energy storage to accommodate the intermittency of wind power. This is based on the use of slam storage facilities associated with wind farms in order to smooth the wind power output and store surplus wind. In [5], Doherty & O’Malley present a methodology to quantify the reserve needed on a system due to the increase of uncertainty with the increase of wind penetration. Generator outages rates and load and wind power forecasts are considered. The reliability of the system is used as a measure the effect of increasing the wind power penetration. The purpose of this work is quantifying the value of energy storage in the context of electricity system security, under alternative generation development scenarios. In particular it is concerned with the application of storage technology in enhancing the value of intermittent energy resources. This includes evaluating its contribution to reducing the cost of balancing supply and demand in operational timescales, given the increased uncertainty caused by the unpredictable nature of wind generation; This paper is organized in the following sections: Section II presents the components of the model behind the developed methodology, Section III presents some of the more relevant results obtained and finally some conclusions are outlined in section IV. II. METHODOLOGY The developed methodology is based on a detailed simulation of the operation of the system. This is concerned with the evaluation of underlying costs associated with the operation of a system with a considerable contribution of intermittent generation, and so focuses only on the question of the management of intermittency by providing standing reserve. The evaluation model used applies a simulation approach using year round evaluation of system operation, for
2 forecast net profile and realisation ne t (de m a nd - w ind) in MW
an hourly time series of wind and demand. This has the advantage over analytical models that there can be a more accurate allocation of spinning / standing reserve and that it can deal with chronology. The model essentially quantifies the benefits of energy storage for providing short-term demandsupply balancing capability in generation systems with high wind penetration. The major components of the model include: a) Mixed Integer Programming (MIP) formulation or, alternatively, a priority ranking method, for committing generating units on a day-ahead basis, based on wind and demand forecasts. This is based on the forecast net demand and wind profile (Figure1); b) Linear Programming (LP) formulation for dispatching power among generators, wind, storage and OCGT standing plants in real time. The economic dispatch is carried out based on the actual net demand and wind profile (Figure1). This allocates the hourly power output necessary to meet demand among the committed conventional generators, wind generators, storage facilities and/or OCGT. This is done in order to minimize the overall fuel cost over a year whilst observing power system constraints. Reserve comes from the committed generators, storage and/or OCGT and is responsible for supplying the imbalances between generations and demand. Whenever the imbalances correspond to a lack of generation capacity in the system storage in discharged, OCGT plant is started or load shed is applied. c) Random walk method for generating imbalances between the forecast wind and the realized wind in real time. This is one of the core components of the methodology. The random walk is generated assuming that the frequency with which different sizes of random imbalances occur follows a normal distribution, for an annual set of hourly values. A simulation approach, based on the work presented in [6] involves representing the hourly imbalances by generating random numbers drawn from a normal distribution and using a reserve cost function to cost each imbalance. The optimal (least cost) allocation between spinning and standing reserve can be found by running the simulation repeatedly for different allocations and using a grid search method; d) Statistical methods, based on wind time series data analysis and wind forecast error, for determining reserve levels needed. Persistency based wind forecast techniques are used considering a four hours lead time, considering this as the typical time to start a new plant. Based on this it is reasonable to look at the output changes in wind power over a four hour period. This is detailed in [3]. The forecast profile presented in Figure 1 represents the uncertainty in wind forecasting and is generated combining random walk displacements with the original historical profile. This study is in line with previous work assessing storage value from forecast and realized profiles [7].
45000 40000 35000 30000 25000 20000 15000 10000 5000 0
forecast realised
1
3
5
7
9
11 13 15 17 19 21 23 hour
Fig. 1. Forecasted and actual net profiles
The selection of case studies modelled is based on the application of this methodology for reducing spinning reserve (SR) in order to increase the amount of wind power absorbed by the system and finding the optimal allocation of reserve between spinning and standing reserve. The inputs of the model, used in the performed case studies are: conventional generation, wind generation, storage power rating and OCGT output power. The allocation of SR in terms of how many standard deviations of wind forecast uncertainty is also pre-established. The main outputs of the model, used in the performed case studies are: the annual energy produced by conventional plants, annual generation costs (including SR cost), annual energy not supplied, annual wind curtailed, annual storage charge and discharge energies, annual energy produced by OCGT and annual CO2 emissions. III. CASE STUDIES In addition to SR, which is provided by part-loaded synchronised plants, the balancing task can be supported by so called standing reserve, which is supplied by higher fuel cost plant, such as OCGTs and storage facilities. Application of standing reserve can improve the system performance through reduction of the fuel cost associated with system balancing. This reduction in the amount of synchronised reserve committed leads to: (i) an increase in the efficiency of system operation, (ii) an increase in the ability of the system to absorb wind. The allocation of reserve between synchronised and standing plant is a trade-off between the cost of efficiency losses of part-loaded synchronised plant (plant with relatively low marginal cost) and the cost of running standing plant relatively high marginal cost [7]. The cost of using energy storage facilities for this task is influenced by their efficiency. The balance between synchronised and standing reserve can be optimised to achieve a minimum overall reserve cost of system management. Studies have been carried out for four SR reduction categories considering different levels of wind penetration. The first area that is covered by these studies is a comparative analysis of results to identify the drivers for Storage in providing standing reserve, in conjunction with CCGT providing SR, as part of the overall balancing services. The second area explores how the value of storage changes with different wind penetration levels, including an examination of the influence of key drivers such as the flexibility of the generation system. The third area is concerned with what happens when the standing reserve is made up of a
3 combination of storage and OCGT installed. The studies performed consider three different generating flexibility systems: Low Flexibility (LF), Medium Flexibility (MF) and High Flexibility (HF). Table I presents the description of the test systems used in the case studies presented in this paper. CHARACTERISTICS OF GENERATION SYSTEMS CONSIDERED
LF
MF
HF
Moderately flexible generation
Flexible Generation
8.4 GW installed, has to run at 100% of max capacity
26 GW installed, minimum stable generation 77% of max capacity
>25.6 GW installed, minimum stable generation 50% of max capacity
8.4 GW installed, has to run at 100% of max capacity
26 GW installed, minimum stable generation 62% of max capacity
>25.6 GW installed, minimum stable generation 50% of max capacity
None
>60 GW installed, minimum stable generation 45% of max capacity
Inflexible Generation
None
350
340.7
317.3
300
265.2
250
LF
213.4
200
204.4
189.7
MF HF
156.1
150
127.6
100
98.8
90.3
70.4
55.5
50 0
All the case studies were conduced considering a base case in which all reserve is SR. The results obtained are based on a comparative analysis of the advantages for the system of each class of SR reduction over the base case. In the next subsections the results of the different case studies are presented. We consider four different classes for SR reduction. For each reduction of SR we determine statically the amount of standing reserve needed so that the total reserve is able to cover all the imbalances sized up to λ*σ where λ is the number of standard deviation(std) of the wind forecast error and σ is the standard deviation for each wind penetration. The value of λ is determined experimentally considering a load shed avoidance criterion. A. Results for 26 GW wind penetration The first studies conduced were based on a fixed wind penetration of 26 GW and explore the influence of generation mix flexibility in the following parameters: reduction in fuel cost, CO2 emission and conventional energy.
2
3
4
5
Storage Rating (GW)
Fig. 2. Storage reducing fuel costs: all generating flexibilities and 26 GW of wind penetration
Figure 2 shows that the value of storage in the LF generating system is significantly higher than in the HF system. It also shows the value of storage increasing, as standing reserve provision increases at the expense of SR in cases other than inflexible generating cases, although there are no wind savings in corresponding situations. Reduction in Conventional Energy with 26 GW of Wind and Storage
14 12
12.3
11.9 10.7
10 TWh/pa
Generation System
400
m ilionp/pa
TABLE I
Reduction in Fuel Cost with 26 GW of Wind and Storage
8.9
8
LF MF
6 4
4.5
4.2
3.7
4.5
HF
2
0.1
0 -2
2
0.1 3
0.0 4
-0.2 5
Storage Rating (GW)
Fig. 3. Storage reducing conventional energy: all generating flexibilities and 26 GW of wind penetration
Figure 3 shows that the reduction of conventional energy in the low flexibility generating system is significantly higher than in the high flexibility system. This is a measure of how much wind has been saved in the system, that is, wind utilised which would otherwise be curtailed in an all SR system. The same trend, of higher CO2 emission savings in the inflexible generating system compared to the flexible system, can be seen in Figure 4. It’s possible to conclude that the system flexibility is one of the key factors of this study and was found to affect significantly the value of storage. The value of fuel cost savings can be capitalized in order to evaluate the total amount of savings in a predefined time horizon. They are useful for investment decisions.
4 Reduction in CO2 Emissions with 26 GW of Wind and Storage
standing reserve fuel cost reductions 400
8
6
5.3
5
4.1
3.8
MF HF
3.1
3
MF stor LF
4.3
4
2.6
2
1.4
1.1
1
2.0
1.8
300
HF stor
250
LF combi
200
MF combi
150
HF combi LF ocgt
100
MF ocgt
50
0 2
3
4
5
2
Fig.4. Storage reducing CO2 emissions: all generating flexibilities, 26 GW wind
The capital value of storage capacity as a function of the amount of installed capacity is expressed in terms of the value of storage per kW, and represents the additional value created by storage in performing balancing tasks. This value was found to reduce with the increase in storage capacity installed. In this work the Present Worth Analysis method is applied considering an interest rate of 10% and a time horizon of 25 years [3]. Figure 5 presents the capitalized value of storage capacity as a function of installed capacity.
Its clear that the combined solution outperforms OCGT only as spinning reduction is further reduced. The weight of the value of storage in the combined solution increases as the amount of OCGT in the combination increases. Observing Figure 7 we can conclude that the importance of the 1 GW of storage in the combination has increasing importance as SR reduction is increased.
600
619
580 473
431
400 252
200
213
371
205
MF HF
179
0 2
3
4
5
value (£m pa)
LF
721
5
Storage & OCGT shares of combi reserve additional value for 26 GW wind
970 803
4
Fig. 6. Combined standing reserve reduction in fuel costs for 26 GW of wind penetration
1200
800
3
standing rating (GW)
Capitalised value of reduced fuel cost by storage
1000
HF ocgt
0
Storage Rating (GW)
value (£/kW)
LF stor
350
6.8
6.3
reduction (£m pa)
m iliontonnes/pa
7
45 40 35 30 25 20 15 10 5 0
LF stor LF ocgt MF stor MF ocgt HF stor HF ocgt spinning reduction 1
spinning reduction 2
spinning reduction 3
spinning reduction 4
storage rating (GW)
Fig. 7. Storage share of combined additional value over OCGT Fig. 5. Capitalized value of reduction of fuel cost with energy storage
B. Combination of Storage and OCGT to provide Standing Reserve In reality we will expect standing reserve to be provided by a combination of different technologies. We performed case studies in order to find the value of a small amount of storage combined with OCGT. We replaced the storage only option presented in section A for a combination of 1GW of storage with OCGT in order to get the same SR reduction categories. We find that the more expensive OCGT will only be used when storage is already discharging at its full power rating. Many of the imbalances can be dealt with 1 GW of storage. The major contribution to the value of storage comes from the first GW of available power. This can be observed in Figure 6. Equivalent results are obtained for the reduction in CO2 emission and conventional energy. Several case studies were performed for different classes of SR reduction and different levels of wind penetration. Here we present the more relevant results, noting the key trends.
C. Results with different wind penetrations Studies have been carrier out for four different classes of SR reduction, considering different wind penetrations: 16 GW, 26 GW, 36 GW, 46 GW and 56 GW. These studies concern the impact of the change on the wind penetration in the following parameters: fuel cost reduction, CO2 emissions reduction and wind energy savings. The impact of system flexibility is also considered in the studies. In order to be able to compare the amount of standing reserve for different wind penetrations (and considering the same base case of λ= 3.5 std), we keep the following SR reduction classes: Category 1: λ =2.3 std; Category 2: λ=2 std; Category 3: λ=1.5 std; Category 4: λ=1.2 std. The size of σ will be different for each wind penetration so the amount of reserve needed is also different. Table II illustrates the reserve requirements for the different wind penetrations.
5 TABLE II RESERVE CAPACITIES FOR DIFFERENT RESERVE LEVELS Total Category 1 Category 1 Category 2 Reserve (2.3* σ) (2.3* σ) (2.0* σ) Wind (GW) (MW) (MW) (MW) (MW) (3.5*σ) Standing Standing Spinning 5202 8453 11704 14955 18206
1000 2000 3000 5000 6000
Category 2 (2.0* σ) (MW) Spinning
Category 3 (1.5* σ) (MW) Standing
Category 3 (1.5* σ) (MW) Spinning
3000 4000 6000 8000 10000
2973 4830 6688 8546 10404
3418 5555 7691 9828 11964
2229 3623 5016 6409 7803
Category 4 (1.2* σ) (MW) Standing
Reduction in Fuel Cost: HF various wind levels
2000 3000 4000 6000 8000 Category 4 (1.2* σ) (MW) Spinning
4000 5000 7000 9000 12000
600 500
495.79 455.34
£ m illion/pa
16 26 36 46 56
present the results for the LF system. Figure 9 presents the results for the HF systems we obtain equivalent results but lower values.
400
w ind 26
319.92
300
294.26
312.27
159.51
171.62
90.32 55.6
98.78 60.2
246.98 128.3
108.21 55.48 31.8
0 1
70.35 42.7 2
w ind 36 w ind 46
213.29
200 100
1784 2898 4013 5128 6242
w ind 16
379.6
3
w ind 56
4
Spinning reduction category
Fig. 9. Reduction in fuel cost for increasing wind power penetrations
We studied the impact of the increase in wind power penetration on the reduction of fuel costs, associated with balancing the system in real time. The costs are effectively associated with holding and exercising the reserve necessary for managing the forecast uncertainty of demand and generation. Reduction in Fuel Cost: LF various wind levels
1000 900
874.48
£ m illion/pa
800
775.2 689.93
700 608.3 539.24
600 500
503.06 444.59
400
265.15
100
103.2
80.9
w ind 16 w ind 26 w ind 36
412.82
213.36
200
587.88
534.28
336.82
300
761.21
317.31
340.73
127.2
137.0
w ind 46 w ind 56
From these results we can conclude that storage can play an important role in reducing the fuel costs and CO2 emissions resulting from the additional balancing task due to wind intermittency. The value of storage increases, with the wind penetration increase, due to the greater uncertainty, introduced by the wind forecast error. But storage is not the only technology capable to provide standing reserve and competes with other technologies like OCGT plants. The next section presents a comparative study between these two technologies. D. Comparative analysis of Storage and OCGT with different wind penetrations For different wind penetrations and system flexibilities comparative analysis studies between Storage and OCGT were performed. In Figure 10 we present the results of the value of storage over OCGT for the LF system.
0 1
2
3
Storage v OCGT fuel cost reduction: LF various wind levels
4
Spinning reduction category
Fig. 8. Reduction in fuel costs for increasing wind power penetrations
180.000 160.000 140.000 £ million/pa
For the LF system the increase of fuel cost reduction is directly related to the increase in wind penetration. For the MF and the HF systems we obtained the same trends but the value is lower as the flexibility of the system increases. Of particular interest is the closeness of 56 GW of wind penetration values to those with 46 GW wind penetration. For the MF and HF cases we get trends that are more spread out. Thus the particular benefits from storage in LF systems become less obvious with higher wind penetrations. Equivalent trends are obtained for the reduction in CO2 emissions and in conventional energy. The advantage of storage over OCGT in based on the unique ability of storage to utilise surplus wind. For all wind penetration levels the value of storage increases as spinning reserve is reduced further. The value of storage also gets higher as wind penetration levels increase, indicating the increased balancing task with higher wind and thus larger forecast uncertainty. The same trends were obtained for the reduction in CO2 emissions and reduction in conventional energy. Figures 8
120.000
w ind 16
100.000
w ind 26
80.000
w ind 36
60.000
w ind 46
40.000
w ind 56
20.000 0.000 -20.000
1
2
3
4
Spinning reduction category
Fig. 10. Storage fuel cost value over OCGT for the LF system
The advantage of storage over OCGT is based on the unique ability of storage to utilise surplus wind. It is important to understand why the advantage of storage over OCGT is different for different wind penetrations and generation flexibility. The key issue here is the ability to discharge storage. This can be limited when we have a high wind penetration in a generation system with a high amount of “must run” generation (like our LF system). Independent of the level of wind power available this conventional generation
6 can’t be displaced. This will lead to the lack of opportunity for storage to discharge and consequently charge when we have surplus wind because it’s always fully charged. This can be observed in Figure 11. 56 GW wind: inflexible system 60000
Power (GW)
50000 wasted wind
40000
wind stored wind power
30000
balancing tasks, for all wind penetration levels. Value is greatest in inflexible generating system. These benefits come from reducing spinning reserve and replacing it with standing reserve, which simultaneously increases the amount of wind power that can be absorbed, and moves closer towards the optimal allocation between spinning and standing reserve with regards to a trade off between part load efficiency losses of spinning reserve and higher fuel costs of standing reserve. Any storage advantage over OCGT in providing standing reserve comes from its ability to store wind which might otherwise have had to be curtailed.
discharge
20000
REFERENCES
generation
10000
[1]
0 1
3
5
7
9
11 13 15 17 19 21 23
hour of day (18 June)
[2] Fig. 11. Case of High wind and LF system: storage does not discharge [3] 56 GW wind: flexible system
[4]
60000
[5]
Power (GW)
50000 wasted wind
40000
wind stored wind power
30000
[6]
discharge
20000
generation
10000
[7]
ILEX, Strbac, G, System Cost of Additional Renewables, study for DTI, October,2003:available:www.dti.gov.uk/energy/developep/080scar_repo rt_v2_0.pdf D.G. Ingfield, A study of electricity storage and central energy generation, Rutherford Appleton Laboratory, 1984. G. Strbac, M. Black, Future Value of Storage in the UK, study for DTI, Dec 2004 available online: www.cst.gov.uk/energy/sepn/goranstrbac.pdf Barton J., Ingfield, D., “Energy Storage and its Use with Wind Power”, in proc. of IEEE Power Engineering Society General Meeting, S. Francisco, USA, June, 2005. Doherty R., O’Malley, M., “ A New Approach to Quantify Reserve Demand in Systems With Significant Installed Capacity”, in IEEE Transactions in Power systems, Vol. 20, NO 2, May 2005. Black, M., Brint, A., T., Brailsford, J., R., “Comparing Probabilistic Methods for the Asset Management of Distributed Items”, ASCE Journal of Infrastructure Systems, June 2005. British Electricity International, Modern Power Station Practice 3rd Edition, Volume L, System Operation, Pergamon Press 199
23
21
19
17
15
13
9
11
7
5
3
1
0 hour of day (18 June)
Fig. 12. Case of High wind and HF system: storage is used
In the case of Figure 12 we have the HF system with a lower “must run” generation value, so the system is able to charge and discharge storage for high wind penetration. This explains why for high wind penetration the value of storage is higher in the HF system. In conclusion storage has more additional value over OCGT in the HF generation system with high wind. With lower wind penetration this additional value is higher in the LF systems. IV. CONCLUSION In this paper we presented a methodology based on a detailed simulation of the operation of the system. This is concerned with the evaluation of underlying costs associated with operation of the system with considerable contribution of intermittent generation, and so focuses only on the question of the management of intermittency by providing standing reserve. The methodology was applied in the development of several studies. These studies include the reduction in fuel costs, CO2 emissions and conventional energy due to the reduction of SR and the use of storage and OCGT as standing reserve. The results we present concern storage only and combined storage and OCCT solutions. Different system flexibilities and wind penetrations are explored. In con conclusion we can say that storage adds value in terms of saving fuel costs, reducing CO2 emissions and reduction in conventional energy associated with system
Mary Black graduated from the University of Newcastle Upon Tyne in 1995 and received the PhD degree from the University of Salford in 2003. She is currently a network investment engineer at CE Electric UK. She was previously engaged in post doctoral research in the Manchester Centre for Electrical Energy (MCEE) at the University of Manchester, with an interest in wind generation and electrical energy storage. Main research areas include applying probabilistic models, based on simulation approaches, to Electricity Network problems, in particular for the Asset Management of distributed network items. Vera Silva (‘M05) graduated in the University of Porto in 1999 and received her MSc degree from the same University in 2003. She is currently pursuing the PhD degree in Electrical Engineering at the School of Electrical and Electronics Engineering of the University of Manchester. Her research interests include competitive electricity markets; ancillary services, storage, demand side management and renewable energy. Goran Strbac (‘M95) is a professor of electrical power engineering at the School of Electrical and Electronic Engineering of the University of Manchester, Manchester, UK. His research interests are in the area of economics of power systems centered on pricing networks, ancillary services, regulation and the economics of disperse generation. He actively works in CIGRE Task Forces and IEEE Specialist groups in these areas.
