2014/17 94651 k nKonw A A weldegdeg e ol n oNtoet e s eSrei r e ise s f ofro r p r&a c t hteh e nEenregryg y Etx itcrea c t i v e s G l o b a l P r a c t i c e The bottom line Incorporating Energy from Renewable Resources into As countries generate more energy from renewable sources, Power System Planning that energy must be taken into account in the power-system Why is this issue important? at minimum costs, while also meeting established technical, social, planning process. However, financial, political, geographical, and environmental constraints. The because renewable sources As countries generate more energy from renewable most common method is the least-cost planning approach, in which such as wind and solar are sources, that energy must become part of the power- the planning problem is formulated as a mathematical optimization variable and geographically system planning process problem that minimizes the costs of investments and power system dispersed, specific procedures operation over the planning horizon: Many countries are scaling up their investments in renewable energy. are needed to ensure that In 2004, investments in all forms of clean energy represented about the grids into which they are 20 percent of total global investment in generation capacity. By 2011, integrated can continue to meet that share had topped 40 percent (Bloomberg New Energy Finance demand reliably every minute of 2012). Energy from renewable sources is sought for various reasons— every day, now and in the future. chief among them to reduce climate-altering carbon emissions, to With appropriate adaptations, diversify generation portfolios, and to decrease dependence on and the variable output of renewable exposure to volatile fossil fuels. However, renewable sources—nota- energy sources can be bly wind and solar power—have unique characteristics that require incorporated into commonly specific operational and planning procedures to ensure that the used power planning models. grids within which they are integrated can continue to meet demand Existing planning methods and approaches, such as least-cost reliably every minute of every day, now and in the future. planning, must now be adapted to account for the distinctive charac- With appropriate modifications, the output of renewable energy teristics of RES, including variability, geographic dispersion, and a cost sources (RES), even wind and solar, can be incorporated into curve that differs from that of fossil-fuel generation. Variability means commonly used power planning models. that the output wind and solar of generators varies from hour to hour and from day to day. RES are geographically dispersed because What is the key challenge? generation must occur where the water flows, the wind blows, and Marcellino Madrigal is Resource variability complicates traditional the sun shines. Resource variability is of particular concern to power planners a senior energy least-cost planning specialist in the World in ensuring long-term supply adequacy.1 Thermal power plants can Bank’s Energy Practice. In power sector planning (often referred to as investment planning generally guarantee supply (they are “dispatchable”). Hydropower or capacity expansion planning), demand forecasts are compared Rhonda Lenai Jordan is an energy specialist in against the existing power infrastructure to determine the type, size, 1 The short-term movement of RES power output (intra-hourly variation) also affects oper- and timing of the additions to generation, transmission, and distri- ations. Dispatch studies have shown that additional supply is needed to compensate for the the same practice. frequent changes in RES output. The intra-hourly variation of RES output is not covered in this bution capacity that are required to meet the forecasted demand brief, but Troy, Flynn, and O’Malley (2012) provide additional information. 2 I n c o r p o r a t i n g E n e r g y f r o m Re n ewa b l e Re s o u r c e s i n t o P o we r S y s t e m P l a n n i n g plants are subject to seasonal variation, but otherwise are dispatch- Figure 1.  Determining the probability of RES output using able as well. Wind and solar power, by contrast, is generated only the generation duration curve when the wind blows or when the sun shines. The change in power 100 output from day to day (intra-daily variation) must be taken into Generation duration curve Power output (y) as % of installed capacity 90 Generation blocks account when planning the expansion of generation, because there 80 is no guarantee that RES will be available at any given time to help 70 “Several simple methods ensure reliability and adequacy of supply. 60 can help planners Although sophisticated new planning tools have been developed 50 incorporate resource to incorporate renewables (Welsch and others forthcoming), they 40 may not be readily available to planners, or planners may simply variability into conventional 30 want a first-order estimate of how a generation expansion plan long-term power sector may be affected by the introduction of renewables. Several simple 20 planning using the least- methods can help planners incorporate resource variability into 10 conventional long-term power sector planning using the least-cost- 0 cost-planning approach.” 0 10 20 30 40 50 60 70 80 90 100 planning approach. x% probability of power output > = y% of installed capacity What solutions are available? Source: Synthetic data generated by AWS Truewind LLC. Simple techniques can be used to account solar resources, present an additional level of uncertainty related to for variability hydrological, wind, or sun conditions, and typically energy storage Although limited in the level of detail they can capture, least-cost- is not available. The rationale for this representation, as contrasted planning models can be modified to incorporate RES variability into with the approach used with thermal plants, is that it can capture power sector planning. This can be achieved by (i) carefully repre- the seasonal variation of RES units in addition to representing senting supply or demand and (ii) modifying long-term adequacy their nondispatchable nature. Seasonal variation is represented by constraints. specifying the expected generation and available capacity of the unit Supply-side approaches. The first supply-side approach is a by period, and, if multiple hydrological conditions can be modeled simple one in which the renewable resource is represented as an within the framework, the probabilities of expected RES generation “unreliable” thermal unit (Koritarov 2010). The forced outage rate can be specified as well. of the unit is specified to be very large so that the unit’s generation Using energy production data over a given period, it is possible matches the expected generation of the renewable resource, and to determine a “generation duration curve” for RES generation the planned outage rate for maintenance is set to zero because RES that depicts the frequency with which RES power output exceeds units such as wind and solar typically require very little maintenance. a certain level versus RES power output (as a fraction of installed (This can be modified to suit each case.) Fuel costs are set to zero, capacity).2 Figure 1 corresponds to simulated wind generation in and operations and maintenance costs should capture the true costs of the RES unit. In the least-cost-planning model, dispatch is simu- 2 Extensive, high resolution data are needed to provide certainty when developing gen- lated for each load level of each period. Because the running costs of eration duration curves. For example, one year’s worth of data measurements at 15-minute the RES unit are low, the unit will be dispatched when available. intervals following international practices is needed. ERCOT provides this information for Texas (United States) at the following site: http://www.ercot.com/gridinfo/generation/. NREL provides The second approach treats the RES unit as a run-of-river an extensive dataset for the eastern United States at http://www.nrel.gov/electricity/transmis- hydropower plant (Koritarov 2010). Run-of-river hydro, like wind and sion/eastern_wind_methodology.html. 3 I n c o r p o r a t i n g E n e r g y f r o m Re n ewa b l e Re s o u r c e s i n t o P o we r S y s t e m P l a n n i n g Figure 2. Calculating the net load duration curve a. Demand and wind generation record b. Net load duration curve Wind output Demand 19 20 Net load 17 Wind output 18 Net load duration “The traditional least-cost- 16 15 planning framework can 14 13 be used with both of the 12 ‘000 MW 11 ‘000 MW 10 supply-side approaches 9 8 described above by 7 6 representing demand in 4 5 3 the normal way and RES 2 1 units as either thermal or 0 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700 hydropower units.” Hour Hour Source: Synthetic data generated by the authors. Ontario from 2002 to 2006.3 Based on this curve it is possible to of these methods are that (i) the stochastic nature of RES output is define three wind-generation blocks that correspond to high (with misrepresented (Koritarov 2010) because neither intra-hourly nor a probability of 10 percent), medium (80 percent), and low (10 intra-daily output variation is represented, (ii) chronological power percent) wind-power production. If the periods are monthly, it can output information is not captured, and, (iii) in the case of the first be assumed that high winds occur 10 percent of time, that is, for approach, it is difficult to specify the generating capacity available three days; medium winds (80 percent) for 24 days; and low winds during peak demand periods. Representing the RES unit as a thermal (10 percent) for three days. More importantly, the wind-generation unit may be more appropriate for concentrated solar power and duration curve is used to define the power that may be available biomass units; however, this approach is less appropriate for wind with some degree of probability. For instance, based on the figure, and solar photovoltaic units. approximately 5 percent or more of the installed capacity is available Demand-side approach. Typically, variable RES output must 95 percent of time. be absorbed by the power system once produced, with output The traditional least-cost-planning framework can be used with acting like a sort of negative demand. Thus, the final approach both of the supply-side approaches described above by representing commonly used to represent RES in least-cost-planning models demand in the normal way and RES units as either thermal or hydro- has been termed the “net load” approach (De Jonghe and others power units (specifying the technical and economic parameters of 2011). It requires modification of the so-called demand load duration the unit accordingly). The advantages of this approach are that the curve. In this approach, net load is determined as the hourly load RES units are included in dispatch simulations, while seasonal vari- less RES generation for a given period; the net load duration curve ability and some uncertainty are also captured. The disadvantages is then generated by graphing net load data in descending order of magnitude. (The largest value is plotted on the left; the smallest value 3 Wind output simulation by AWS Truewind LLC was commissioned by the Ontario Power on the right.) Authority. The data can be found at http://www.powerauthority.on.ca/integrated-power-sys- tem-plan/simulated-wind-generation-data. 4 I n c o r p o r a t i n g E n e r g y f r o m Re n ewa b l e Re s o u r c e s i n t o P o we r S y s t e m P l a n n i n g Figure 2 depicts this process. Figure 2(a) shows a month of actual power system will fail to provide uninterrupted service to customers. demand, wind power generation, and demand minus the wind power This approach uses the characteristics of system components to contribution. This last curve describes the load to be supplied by predict the likelihood that demand will be served and provides a sources other than wind. Figure 2(b) depicts the net load duration meaningful representation of the random events that can affect curve. The wind generation corresponding to each hour is also supply and demand. Another probabilistic metric is the loss of shown. load expectation (LOLE), which is similar to LOLP and specifies the “Unlike the supply-side This approach does not capture uncertainty and assumes expected amount of time during which demand will not be met over approaches described that the same chronological wind pattern will prevail in the future; a given period. above, the net load however, the approach is able to capture the variability of RES Ensuring that a system comprising RES units will be reliable power output. Therefore, when using this approach it is important requires the modification of the long-term adequacy constraints as approach does not include to incorporate greater time discrimination in the least-cost-planning described above. In particular, assuming that reliability constraints RES units in dispatch model to capture as much variability as possible. Consequently, the are implemented using reserve margins, it is important to update simulations—in other approach requires extensive data and the use of renewable resource and increase the reserve requirements beyond the 10–20 percent words, the contribution assessments. Additionally, this modeling method reflects how supply typically required in most power systems and to determine the firm of RES units to long-term interacts with the load pattern and sharpens estimations of the costs generating capacity of RES units offline (external to the model) to of the generation required to supply demand when RES generation ensure that sufficient capacity is available to meet demand. For adequacy requirements changes. further detail on this topic, see Matos and others (2009). is neglected, resulting Unlike the supply-side approaches described above, the net load Reserve requirements. Integrating renewables in the grid in somewhat “fictitious” approach does not include RES units in dispatch simulations—in necessitates an increased reserve margin to meet standard reliability results.” other words, the contribution of RES units to long-term adequacy criteria. One rule of thumb often used when incorporating renew- requirements is neglected, resulting in somewhat “fictitious” results. ables into planning models is to increase the reserve requirement When employing this approach, therefore, it is particularly important by the fraction of firm generating capacity supplied by RES units. For that long-term adequacy constraints, or reliability, should be modified example, if the reserve margin requirement was originally deter- as described below. mined as Modification of long-term adequacy constraints. In order to ensure that current and future demand is met despite resource variability, forecasting errors, and unanticipated events, long-term adequacy requirements are expressed using both deterministic Then the new reserve requirement should be calculated as and probabilistic metrics. The simplest deterministic approach to addressing this concern requires a planning reserve, whereby a pre-specified level of excess “firm” capacity must be available above and beyond what is required to meet peak demand levels. The where PG represents the firm generating capacity of conventional planning reserve accounts for both the operating reserve require- power sources (without variable renewables), PD indicates peak ments for peak periods and uncertainty in projections of load growth. demand, and CRES is the firm generating capacity supplied by RES In most systems, regulators require reserve margins to be 10–20 units (Madrigal 2012). percent in order to ensure that, in case of generator breakdowns or Firm capacity. Firm capacities are used to indicate the contri- sudden increases in demand, the power grid is still operational. bution of individual units to the supply of peak demand and to help A common probabilistic metric is the loss of load probability planners determine if the integrated supply mix meets the planning (LOLP), defined as the probability over some period of time that the reserve constraint. For thermal generation, firm capacity is expressed 5 I n c o r p o r a t i n g E n e r g y f r o m Re n ewa b l e Re s o u r c e s i n t o P o we r S y s t e m P l a n n i n g Figure 3. Determination of firm RES capacity using (a) average production and (b) power generated during peak demand 60 300 Wind output Wind output 50 250 Average production Demand 40 200 Firm capacity = average production MW MW “While deterministic 30 150 20 100 methods provide Firm capacity = power generated 10 50 simple approaches to during peak demand 0 0 approximating the firm rch ril e st r er r r rch ril e st r er r r ry ary y y ry ary y y e e e e e e Ma Jul Ma Jul Jun Jun gu gu Ap Ap ua ua mb tob mb mb mb tob mb mb bru bru Ma Ma capacity of RES units during Au Au Jan Jan pte ve ce pte ve ce Oc Oc Fe Fe No No De De Se Se peak demand, probabilistic Month Month methods better reflect Source: Madrigal 2012. the uncertainty of the contribution of the as the time (in hours) that the unit is available to produce energy assumes that firm RES capacity is equal to the increment in demand variable source to system divided by 8,760, the number of hours in a year, scaled by maxi- that can be met at the same level of LOLE after the renewable reliability.” mum output capacity.4 For hydro power, it is calculated based on resource is added to the power system. The steps used to determine the power that the plant can guarantee at peak time with a given firm capacity using the LOLE method (Madaeni and others 2012) are probability. For other RES, firm capacity is highly dependent on the outlined below: type of technology and is determined using deterministic (figure 3) • Determine LOLE0, that is, LOLE without RES operating in the or probabilistic analysis. system. The most common methods of deterministic and probabilistic • Forecast renewable production and determine the net load analysis used to compute firm capacity are the average pro- duration curve. duction, peak demand, and LOLE methods. Average production, the simplest deterministic approach, assumes that, for a given • Determine LOLEn based on the net load duration curve period, firm generating capacity is equal to the average power (that is, with RES in the system). production of the unit. Another deterministic approach, peak • Increase demand across all hours until LOLE0 equals LOLEn. demand assumes that the firm generating capacity is the available • Calculate firm RES capacity as maximum demand corresponding generating capacity of the RES unit at times of peak demand to LOLEn minus maximum demand corresponding to LOLE0. (Madrigal 2012). The LOLE method is a probabilistic approach that determines While deterministic methods provide simple approaches to what is commonly known as effective load carrying capability. It approximating the firm capacity of RES units during peak demand, probabilistic methods better reflect the uncertainty of the contribu- 4 For thermal units, firm capacity is determined as (100-FOR) multiplied by the rated tion of the variable source to system reliability. The final probabilistic capacity of the generator. For large hydro units, firm capacity is determine based on the method described above is also able to reflect how generation probability of water inflows and production during times of peak demand when peak demand is the critical condition. Other measures simply refer to the power output that can interacts with the load pattern. be guaranteed with a given probability (for example, above 90 percent) in a given year. 6 I n c o r p o r a t i n g E n e r g y f r o m Re n ewa b l e Re s o u r c e s i n t o P o we r S y s t e m P l a n n i n g Table 1.Taking RES units into account in least-cost planning Quality of Method Ease of use approximationa Data requirements Pros Cons Represent RES as Very simple Rough first order Expected generation of RES units are included in Does not take into “With the global increase in thermal units RES units; operations and dispatch simulations. account stochastic nature, maintenance costs of RES chronological power output, renewable energy targets, units or firm capacity of RES. power system planners must quickly learn to Represent RES as Simple Good first order Expected generation and RES units are included Neither stochastic nature of run-of-river hydro unit available capacity of RES unit in dispatch simulations; RES output nor chronological integrate RES into long- per period; the probabilities of seasonal variation and power output is taken into term generation expansion expected RES generation nondispatchable nature of account. RES are taken into account. plans without increasing Net load approach More complex Good first order Hourly demand; hourly, Captures intra-hourly and Does not capture uncertainty; costs or compromising chronological RES generation intra-daily RES output assumes the same reliability or environmental for each period variation; reflects how supply chronological pattern in the interacts with demand; allows future. values.” more accurate estimation of system costs. Source: Authors. a. The quality of the approximation depends on the quality of data available. Higher resolution data and more time-series data improve quality. What are the key lessons? indicating the data needed for implementation as well as the advantages and drawbacks of each method. Trade-offs exist between simplicity of method The methods described here and summarized in tables 1 and and accuracy of results 2 have been used in various cases. For example, the regulator in With the global increase in renewable energy targets, power system Mexico issued an ordinance that determines the amount of energy planners must quickly learn to integrate RES into long-term gener- from RES (specifically wind) that should be used for planning ation expansion plans without increasing costs or compromising purposes, referencing the contribution of these sources to peak reliability or environmental values. The variability of renewable demand. Argonne National Laboratory in the United States has used resources can be captured in long-term planning using classical the LOLE method as a way to incorporate wind power into the Wien least-cost-planning models. This is achieved through a two-pronged Automatic System Planning capacity-expansion model. The Argonne approach: First, RES units are represented as if they were thermal model is used in various developed and developing countries.5 or hydro plants, or the net load duration curve is used to represent Additionally, the net load approach to incorporating RES units and demand (excluding RES from dispatch). Second, long-term adequacy constraints are modified by taking into account the estimated firm 5 Presentation can be found at https://www.ferc.gov/EventCalendar/ generating capacity of renewables in one of several ways. Tables 1 Files/20100608141145-Koritarov,%20Argonne%20NL%20-%20Modeling%20Wind%20in%20Ex- and 2 provide a comparative overview of the approaches presented, pansion%20Planning.pdf. 7 I n c o r p o r a t i n g E n e r g y f r o m Re n ewa b l e Re s o u r c e s i n t o P o we r S y s t e m P l a n n i n g Table 2.Calculating the firm generating capacity of RES Quality of Method Ease of use approximationa Data requirements Pros Cons Average production Very simple Rough first order Average RES production for a Easy to calculate. Results in gross over- or “The methods described estimate given period underestimation of firm capacity; does not capture in this note provide a uncertainty; not suitable for first-order estimate of project closure feasibility analysis, as further the generating capacity assessment is required. needed to meet demand Contribution during Very simple Good first order Power generated by RES units Easy to calculate; assesses Does not capture uncertainty; over time. To improve the peak demand at peak demand RES output at time of peak not recommended when RES demand. make up a large share of total quality of the results and generation capacity. obtain a more holistic LOLE approach Complex More Hourly load and RES Better reflects the uncertainty Is data intensive and requires view of the system, least- sophisticated generation for each period of the variable source’s proper modeling tools. contribution to system cost-planning models reliability; reflects how must be complemented generation interacts with demand. by additional analyses to Source: Authors. capture the intra-hourly a. The quality of the approximation depends on the quality of data available. Higher resolution data and more time-series data improve quality. variation of RES output as well as the geographic dispersion and dynamic the LOLE approach to estimating firm capacity are industry standards References costs of the renewable and used by very skilled planners and utilities worldwide. sources.” Bloomberg New Energy Finance. 2012 “Global Trends in Clean Energy The methods described in this note provide a first-order estimate Investment, 2012.” Frankfurt School of Finance and Management. of the generating capacity needed to meet demand over time. To http://fs-unep-centre.org/sites/default/files/publications/global- improve the quality of the results and obtain a more holistic view trendsreport2012.pdf. of the system, least-cost-planning models must be complemented De Jonghe, C., E. Delarue, R. Belmans, and W. D’Haeseleer. 2011. by additional analyses to capture the intra-hourly variation of RES “Determining Optimal Electricity Technology Mix with High Level output as well as the geographic dispersion and dynamic costs of the of Wind Power Penetration.” Applied Energy 88: 2231–2238. renewable sources. While not the focus of this brief, these features ESMAP (Energy Sector Management Assistance Program). should be included in the planning process to ensure that a diverse Forthcoming. “Electricity System Planning Tools for Enhanced supply mix meets demand in the short, medium, and long term. Integration of Renewable Energy.” World Bank, Washington, DC. Executing multiple models and performing manual analysis can be Koritarov, V. 2010. “Modeling Wind Energy Resources in Generation tedious. Fortunately, new commercial tools have been designed to Expansion Models.” Presentation to FERC Technical Conference alleviate the planning burden (ESMAP forthcoming). These will be the on Planning Models and Software, June 9–10, Argonne National subject of a forthcoming Live Wire. Laboratory, Argonne, IL (United States). 8 I n c o r p o r a t i n g E n e r g y f r o m Re n ewa b l e Re s o u r c e s i n t o P o we r S y s t e m P l a n n i n g Madaeni, S. H., R. Sioshansi, and P . Denholm. 2012. “Comparison of Troy, N., D. Flynn, and M. O’Malley. 2012. “The Importance of Make further Capacity Value Methods for Photovoltaics in the Western United Subhourly Modeling with a High Penetration of Wind Generation,” connections States.” NREL Technical Report NREL/TP-6A20-54704. National Power and Energy Society General Meeting, 2012 IEEE, pp. 1–6. Renewable Energy Laboratory, Golden, CO. July. July 22–26. Live Wire 2014/1. “Transmitting Madrigal, M. 2012. “Technical and Economic Impacts of Variable Welsch, M., M. Howells, M. Hesamzadeh, B. Ó Gallachóir, P . Deane, Renewable Energy to the Grid,” Renewables in Operating and Planning Electricity Grids.” N. Strachan, M. Bazilian, D. Kammen, L. Jones, H. Rogner, and by Marcelino Madrigal and Presentation at Condensed Training in Energy Planning, Finance, G. Strbac. Forthcoming. “Supporting Security and Adequacy in Rhonda Lenai Jordan and Cross-border Trade, June 2, World Bank, Washington, DC. Future Energy Systems.” Submitted to Applied Energy, 2014. Matos, M., J. Peças Lopes, M. Rosa, R. Ferreira, A. Leite da Silva, W. Sales, L. Resende, L. Manso, P . Cabral, M. Ferreira, N. Martins, The peer reviewers for this note were Husam Mohamed Beides (lead energy specialist, Energy and Environment Sector Group, Middle C. Artaiz, F. Soto, and R. López. 2009. “Probabilistic Evaluation East and North Africa Region, World Bank) and Efstratios Tavoulareas of Reserve Requirements of Generating Systems with (senior operations officer, Sustainable Business Advisory, Clean Energy Renewable Power Sources: The Portuguese and Spanish Cases.” and Resource Efficiency, IFC). International Journal of Electrical Power & Energy Systems 31(9). October.