Harnessing the Renewable Generation Potential

Transcription

1 Harnessing the Renewable Generation Potential Alberto J. Lamadrid, Tim D. Mount, Ray D. Zimmerman and Carlos E. Murillo-Sánchez Abstract States in the U.S. and developed nations around the world have established mandates on what the amounts of socially desirable renewable generation should be. With further adoption of zero marginal cost resources, the payments for energy ($/MWh) are being reduced for conventional generation, which then needs to be compensated in a capacity market ($/MW). Due to the inherent uncertainty associated with these resources, other operational considerations, like ramping costs for conventional generation, may play a significant role in the financial viability of these generators. In this context, new unconventional providers of ancillary services can help in the operation of the system. The objective of this paper is to analyze the role of deferrable demand and Energy Storage Services (ESS) for planning over a year, taking the load duration curve observed in historical patterns. The objective of this paper is to study the effect of two main changes aimed at operating a secured system, while maintaining the sustainability of the system: 1) the role that ramping costs and can have in the operation of the system, to counteract the unpredictable nature of Renewable Energy Sources (RES) and 2) The optimal management of Deferrable (or controllable) loads, given the inter-temporal constraints they face, to be coupled with RES. This will extend the concept of controllable loads to include thermal storage, and in particular, the use of ice batteries to replace standard forms of air-conditioning. 1 Introduction For agents participating in the wholesale market, the decision to provide ancillary services requires proper incentives to match the socially desirable outcomes. Previous research by this and other groups has shown the complementary roles that electric vehicles and thermal storage can play in flattening daily patterns of demand and mitigating the variability in generation from Renewable Energy Sources (RES) ([Kintner-Meyer et al.(2007)]). This is especially important for regulators who need to play the role of a social planner by maximizing the welfare of all market participants. The availability of Energy Storage Systems (ESS) allows the owners to 1) lower the cost of purchasing energy through price arbitrage across the day, and 2) get paid for mitigating the variability RES by selling ramping services. While the exact mechanism by which deferrable loads and ramping markets are established are not discussed, we believe these are imminent developments for electricity markets. In the case of the U.S., the electrification of transportation is happening in many places in response to major developments by automakers (e.g. Nissan, GM, Ford) and subsidies from the government. In addition, the control of air-conditioners to supply short-run A.J. Lamadrid, T.D. Mount and R.D. Zimmerman are with Cornell University. C.E. Murillo-Sánchez is with Universidad Nacional de Colombia. Corresponding author ajl259@cornell.edu 1

2 demand reductions in response to system contingencies is widespread. Using ice batteries instead of air conditioning to move demand from peak to off-peak periods is a natural extension of these capabilities. The objective of this paper is to analyze the effect of adding deferrable loads and ancillary services to the electric system, and and to determine the impact for all agents involved. Recent research at Cornell has demonstrated why the current policy debate on the benefits of reductions in the wholesale price of electricity associated with additional sources of renewable energy misses the hidden system costs [Mount et al.(2011)]. The basic argument is that a greater dependence on wind (or solar) generation will lower average wholesale prices, and by doing so, will undermine the financial viability of some of the conventional generators that are essential for maintaining reliability. This missing money is generally paid to generators in some form of capacity market, such as the Forward Capacity Market in New England. The basic economic principle is that all customers purchasing electricity should pay for both energy ($/MWh) and ademand charge for capacity ($/MW). At the present time, most residential and small commercial customers do not pay a demand charge, but in an efficient smart grid, all customers or their aggregators should pay for the missing money associated with their demand for capacity during peak system periods. Although markets provide real-time" locational prices, these prices are really determined in discrete time steps, usually five minutes. Nevertheless, supporting system reliability may require much more rapid response to provide, for example, ramping services to mitigate the variability of generation from renewable sources such as clouds passing over solar panels on a sunny day. Customers providing these ramping services should be rewarded appropriately [Mount and Lamadrid(2010)]. The National Electricity Market (NEM) in Australia, for example, is designed to use market prices to provide incentives for responses that take longer than five minutes and automatic controls for more rapid responses. For example, some appliances are switched off in response to a drop in frequency and others switch off in response to a wireless signal when curtailment of demand is needed to maintain system reliability. An important additional feature of the NEM is that real-time prices are projected for a number of hours into the future based on current information about bids and offers in the market and the forecasts of wind speeds and solar intensity. Although these projected prices are not binding for transactions, they do provide incentives for customers to delay demand if lower prices are projected or for generators to enter the market if higher prices are projected. Regulated and deregulated markets in the USA do not provide the full range of incentives/signals that are provided in the NEM. An efficient smart grid will, however, have to provide these types of incentives if customers and aggregators are going to be effective participants in a market for electricity and contribute to maintaining system reliability. A primary goal of this paper is to evaluate how financial incentives for customers can be structured to reflect the true system costs/benefits and encourage investment in new technology such as controllable demand. This is an essential for regulating a smart grid if it is to support a hierarchical structure of control for real time operations and an economically efficient two-sided market. The paper is organized in 6 sections, with a literature review and introduction to the model in sections 2 and 3, a description and calibration of the model in section 4, presentation and discussion of results in section 5, and conclusions in section 6. 2 Literature Review The issue of strategic behavior and self-commitment has been analyzed from different perspectives. [Zhang et al.(2000)] study the provision of energy services, taking into account ramping constraints, using Lagrangian relaxation, with reaction models for both individual agents and the social planner 2

3 (ISO). [Shrestha et al.(2004)] studies strategic behavior of generators and considerations on how to use ramping rates in energy bidding operating under market conditions. The model used assumes cost minimization by the social planner, and the bidding agent uses the ramping capability of the generating units as a source of additional revenue. In the modeling, the physical characteristics of the units is taken into account (e.g. economic valuation of rotor life). In [Bouffard et al.(2005)], a security-constrained OPF (SC-OPF) model for energy is proposed. The system is constrained to have a single contingency for the optimization, and a probability is associated to the occurrence of the contingency in the system. Reserve prices are determined by the cost of load shedding in post-contingency states. In [Chen et al.(2005)], the objective function is modified to include the cost of energy and reserves in a co-optimization framework (CO-OPT), with a base case and a set of credible contingencies with associated probabilities of occurrence for each one. The use of RES in the system, specifically wind, is studied in [Karki et al.(2006)] with a wind model that allows to asses the reliability contribution of a wind farm. In [Condren et al.(2006)], the problem of dealing with expected post-contingency flows in including ramping is studied in what the authors call an Expected Security-Cost OPF (ESCOPF). The transition from pre to post contingency states is valued with functions reflecting the ramping cost per generator to deal with the change in operating points, for a single time period. The framework for that we called the SuperOPF was introduced in [Thomas et al.(2008)]. This is stochastic, security constrained Alternating Current (AC) OPF with endogenous reserves. Following CO-OPT, energy and reserves (positive and negative) are solved simultaneously, taking into account the social cost of Load Not Served (LNS), priced at the Value of Lost Load (VOLL). In this context, the contingencies are included as economic constraints, allowing for a better accommodation stress induced in the network (e.g. the inclusion of stochastic sources). 3 Formulation The overall framework is a Security Constrained Alternating Current Optimal Power Flow (AC- OPF). It includes the behavior of energy storage for each time period, making this a nonlinear problem [Wood and Wollenberg(1996)]. The operator wants to minimize the total system operation costs, subject to optimal loading constraints and unit limits. Under the co-optimization framework proposed by [Chen et al.(2005)], the reserves are determined simultaneously as the dispatches, adjusting to the operational changes that the system may experience. Additionally, as suggested by [Thomas et al.(2008)], the amount of Load Not Served (LNS) is an additional variable to be solved, with compensation to affected consumers at the Value of Lost Load (VOLL). The nature of storage usage decisions spans a horizon that involves several time periods. For each inter-period decision, the ramping capabilities of the storage unit need to be explicitly modeled as part of the optimization. This is implemented in the framework of the SuperOPF. 3

4 min Gitsk,Ritsk,LNSjtsk πtsk CGi (Gitsk ) t T s S t k K i I Gitc ) Dec its (Gitc Gitsk ) VOLLj LNS(Gtsk, Rtsk )jtsk Inc its (Gitsk j J t T ρt [CR (Rit ) it CR (Rit ) it (1) CL (L it ) it i I CL (L it )] it t T ρt s2 S t s1 S t 1 i I ts2 0 Rpit (Gits2 Gits1 ) Rp it (Gits2 Gits1 ) Subject to meeting Load and all of the nonlinear AC constraints of the network. 4 Case Study The framework above was applied to a case study on the utilization of storage, controllable demand and the effect of ramping costs. The network model used is a reduction of the North Eastern Power Coordination Council, by [Allen et al.(2008)]. This is a reduction focused on the nodes in New York and New England, with 19 buses in New York, eight buses in New England and a total of 36 buses. The one-line-diagram of the test network is included in Fig Test Network Figure 1: A One-Line-Diagram of the 36-Bus Test Network. The information about the generation mix is taken from [Allen et al.(2008)], and the generation cost data comes from Energy Visuals/PowerWorld. A description of the generation fleet is included in Table MW of Wind in NYISO and 55 MW of refuse capacity in Ontario are excluded from the summary 4

5 4.2 Controllable Loads Modeling For loads in urban areas, a certain percentage was set as open for fulfillment during the horizon period. The overall objective was to maximize the social welfare (minimizing cost and finding the amount socially optimal of load shedding, valued at VOLL). This amount that was deferrable and could be served at any point during the optimization horizon depends on location, with urban loads having a larger percentage, due to air-conditioning (a-c) load. The implicit assumption is that the a-c load could be served by a mechanism that resembles an ESS (e.g. an ice battery). The exact percentages correspond to the average historical amount observed for New York and New England of load that is sensitive to temperature. The deferrable loads were place in four urban buses: 74316, Dunwodie, 74327, Farragut, 71797, Millbury and 79800, Rochester. 4.3 Wind Modeling A total capacity of 16GW was added to the system - additional to the original installed wind capacity. The wind farms were located in New York (NY) and New England (NE) The capacity was divided in 8GW, placed in Volney (NY, bus 77406), and two wind farms of 4GW each, placed in Orrington and Northfield (NE, buses and respectively). The capacity added corresponds to around 25% of the total generation capacity in NY-NE, a penetration level that is expected to be achieved in coming years. The wind behavior was modeled with two scenarios that reflect a high overall available capacity. In average, the availability factor for the high wind scenario is 96%, while for the lower wind scenario is 73%. However, in the lower wind scenario, there are three possible outages, dispersed temporally and spatially. In the case of Volney, the outage occurs at 5AM. For Northfield, the outage occurs at 10AM and for Orrington, it occurs at noon. This corresponds to a west-to-east moving weather system. Such system creates very high winds, and therefore the need to shut down the turbines for equipment protection (turbines are curtailed for wind speeds above 25m/s). Such modeling allows to study the units performing ramping in the system. 4.4 Cases implemented The purpose of the paper is to study the effects of deferrable loads and ancillary services for wind variability mitigation. For this purpose, the following cases were simulated: 1. No Wind 2. Wind in three locations, buses 72926, and Wind Ramping Costs (RC). 4. Wind Ramping Costs Deferrable loads (DL). 5. Wind DL, buses 74316, 74327, and Results The results in this section assume that the market is deregulated, with the wind and controllable loads as modeled in sections 4.2 and 4.3. The performance of the system is measured by looking at three main metrics: The amount of the renewable resource that actually gets dispatched; the 5

6 operating cost of servicing the load; and the total conventional generation capacity needed to cover the load, including a credible set of contingencies. The first row of Table 3 shows the sum of all operating costs for the 24-hour period analyzed. By adding wind into the system, the total fuel bill is reduce by 18%. The resources displaced by the zero-cost wind are mostly natural gas (25% reduction), oil (20%) and coal (6% reduction). By adding inflexibility into the system by economic constraints (ramping costs, comparing Case 2 to Case 3), There is a small increase in costs, due to optimal spillage of wind capacity. The slack is taken by mostly coal and natural gas, due to network effects. The inclusion of deferrable loads (Case 4) has a sizable impact on operational costs with respect to Case 3, with a 6% reduction. The added controllable loads into the system reverse the effect of adding ramping costs, with wind displacing natural gas and oil mostly. Not accounting for ramping costs but maintaining the deferrable loads in the system with wind (hence comparing Case 4 to Case 5) increases daily costs. This happens due to conventional generation that is used in peak hours (11AM to 3PM) to serve part of the deferrable loads. By removing ramping costs, it becomes, from the system point of view, optimal to use such generation to minimize overall costs. The payment for ramping costs (second row, Table 3), though small in magnitude, led to a change in which the generation units in the system are used. The generators net revenue and congestion rents behave qualitatively similar to operating costs with one exception: Case 5 has overall lower Locational Marginal Prices (LMP s), and virtually the same amount of energy used over the day than Case 4. This leads to lower generator revenue in this case, as well as lower congestion rents. With these tow effects combined, the amounts that loads pay are smaller, therefore accruing benefits to customers. 400,000,000 Composition of payments in the Wholesale Market 350,000, ,000,000 Daily Cost ($) 250,000, ,000, ,000, ,000,000 50,000,000 0 Case 1 Case 2 Case 3 Case 4 Case 5 Case Operating Costs Ramping Costs Generators Net Revenue Congestion Rents Figure 2: Payments in the Wholesale market Fig. 2 summarizes the payments in the wholesale market for each one of the cases illustrated. The generation capacity needed to cover the load including contingencies is reduced by adding deferrable loads, as expected. Ramping costs on the other hand have an ambiguous effect: adding them in a system with no deferrable loads (Case 2 to Case 3), leads to slightly higher capacity needed. But removing them in a system with deferrable loads (Case 4 to Case 5) slightly increases the amount of conventional capacity needed. This effect is locational, as in New york and New England (where the deferrable loads are placed), the generation capacity needed in the areas is reduced (from 46,199MW in Case 4 to 45,471MW in Case 5). Finally, the wind energy accommodated in the system is the highest in Case. It is notable that in all cases, it averages around 63% of the installed wind capacity, due to the wind model underneath. To analyze the operational effects of each one of the cases, Figs 3-7 have the hour-to-hour 6

7 140,000 Fuel Utilization per hour of day, Case 1 120,000 E[Dispatch] per fuel Type, MW 100,000 80,000 60,000 40,000 20, Hour of the day nuclear hydror coal ng oil wind ess Figure 3: Composition of Generation, Case 1 dispatch of the generation fleet, divided by fuel type. In the no wind case (Fig 3), the prevalence of dispatchable generation allows for a stable dispatch of the resource. The load following ramping is done in it s majority by natural gas and oil, and a very small amount by coal. The introduction of wind into the system (Fig 4) displaces the fuels that were doing ramping, mostly natural gas, and some coal. Due to the location of the oil units, there is almost no displacement of this fuel, that is necessary to cover for the instances in which the system is congested. 140,000 Fuel Utilization per hour of day, Case 2 120,000 E[Dispatch] per fuel Type, MW 100,000 80,000 60,000 40,000 20, Hour of the day nuclear hydror coal ng oil wind ess opt (Case 5) Figure 4: Composition of Generation, Case 2 The wind dispatch is relatively constant over the day, indicating some geographical averaging among the three wind farms modeled in the system ([Karki et al.(2006)]). Such spatial compensation therefore is an additional way in which variability is compensated in this simulation. The light-gray line labeled opt(case 5)" shows the total generation dispatch for Case 5. The reason to compare to this case is that the only difference added with this case is the addition of controllable loads. Therefore, the profile obtained is a measure of the potential impact that deferrable loads have on the generation fleet dispatch. As expected, the obtained generation exhibits valley filling" and peak shaving". Such change in the generation does not increase load shedding in the system, and therefore the reliability is maintained. Fig. 5 shows the effect of adding ramping costs for generators, with values consistent with [Mount and Lamadrid(2010)]. The overall dispatch does not change dramatically. There are marginal increases in the usage of coal (around 1.5%), and natural gas, covering for the amounts of wind energy spilled. The curve labeled opt (Case 4)" shows the total generation fleet dispatch for Case 4, when controllable loads and ramping costs are added in the system. The optimal dispatch flattens the overall dispatch of all units, with smaller gradients in the hour-to-hour change. 7

8 Fuel Utilization per hour of day, Case 3 140, ,000 E[Dispatch] per fuel Type, MW 100,000 80,000 60,000 40,000 20, Hour of the day nuclear hydror coal ng oil wind ess opt. (Case 4) Figure 5: Composition of Generation, Case 3 140,000 Fuel Utilization per hour of day, Case 4 120,000 E[Dispatch] per fuel Type, MW 100,000 80,000 60,000 40,000 20, Hour of the day nuclear hydror coal ng oil wind ess Figure 6: Composition of Generation, Case 4 The dispatch for the case with wind, raping costs and deferrable loads (Case 4) is shown in Fig. 6. The deferrable loads servicing is shown in the curve ess. In cases where the curve is below the total area curve (e.g. 1-8 AM), the controllable load is being served by the vertical difference between the sum of all generation fuels and the ess curve (i.e., the area under the ess curve shows the load served that is not controllable). 140,000 Fuel Utilization per hour of day, Case 5 120,000 E[Dispatch] per fuel Type, MW 100,000 80,000 60,000 40,000 20, Hour of the day nuclear hydror coal ng oil wind ess Figure 7: Composition of Generation, Case 5 Fig. 7 shows the detail of the dispatch for each fuel used for demand catering. Besides the mentioned increase in wind utilization due to use of deferrable loads - and subsequent move of natural gas and oil - the serving of the controllable load is shown in the curve ess again. The optimization then solves for the hours in which there are not big overall demand to cater for the 8

9 controllable demand. Focusing now on the total costs (operating and capital) of the policies implemented with deferrable loads and ramping costs markets, Table 4 shows the calculation for each one of the cases studied. The operating costs for the representative day chosen are calculated by taking the total operating costs ($) and dividing by the total amount of MWh demanded over the day. The capital costs are annualized, assuming that all days of the year are similar to the representative day chosen. In this case, the annual capital cost per MW for each conventional generation fuel is calculated (using the values per capacity type included in [Mount et al.(2009)]). This amount is then divided by the total energy used per day, times 365 (to set the calculation in a comparable annual basis). The addition of wind in the system drive down both the operational and the capital costs (Case 1 to Case 2). The addition ramping costs marginally increases the operating costs, while capital costs are unchanged (the generation capacity composition is very close in these two cases). Finally, the use of deferrable loads keeps the same trend of decrease of both operational and capital costs. 6 Conclusion The management of the electricity network from the point of view of the social planner is a changing arena, with more challenging conditions due to stochastic resources, but also more widespread availability of resources that allow to store energy and smooth electricity outputs. This article proposes a methodology for simulating the effects of policies with Energy Energy Sources in electrical networks with high penetrations of wind energy and overall high levels of availability. The load is divided between deferrable and non-deferrable load. This controllability can be used to 1) flatten daily load patterns, 2) mitigate the inherent variability of wind generation and 3) help to cover system contingencies. An important feature of the second generation SuperOPF used for the analysis is that it determines the amount of generating capacity needed to meet load and cover contingencies endogenously over the optimization horizon (24 hours). These specified contingencies include the different realizations of wind generation and the transition probabilities from each realization in one hour to the the realizations in the next hour. A second feature of the analysis is that explicit ramping costs associated with changing the dispatch points of conventional generating units from one period to the next period are included in the objective function. Adding wind generation to a network tends to increase these ramping costs. The results from the analysis show that 1) including ramping costs leads to higher operating costs and more curtailment of potential wind generation, and 2) using controllable loads flattens the daily pattern of demand, helps to cover contingencies, mitigates wind variability and increases the amount of wind generation dispatched. As a result, controllable demand 1) lowers the ramping costs, 2) lowers the operating costs, and 3) reduces the amount of conventional generating capacity needed to maintain reliability. All three of these effects reduce the total system costs. If regulators adopt such measures, customers could benefit from lower net costs. In order to materialize these benefits, customers should get paid for providing services such as ramping and be compensated correctly for lowering the capital cost of the generating capacity needed to maintain System Adequacy. This is not the case at the present time, and as such, the regulatory structure should be modified. 9

10 Acknowledgment This research was supported by the US Department of Energy through the Consortium for Electric Reliability Technology Solutions (CERTS) and by the Power Systems Engineering Research Center (PSERC). The authors are responsible for all conclusions presented in the paper, and the views expressed have not been endorsed by the sponsoring agencies. References [Allen et al.(2008)] Allen, E., Lang, J., Ilic, M., Aug A combined equivalenced-electric, economic, and market representation of the northeastern power coordinating council u.s. electric power system. Power Systems, IEEE Transactions on 23 (3), [Bouffard et al.(2005)] Bouffard, F., Galiana, F., Conejo, A., Nov Market-clearing with stochastic security-part i: formulation. Power Systems, IEEE Transactions on 20 (4), [Chen et al.(2005)] Chen, J., Mount, T. D., Thorp, J. S., Thomas, R. J., Location-based scheduling and pricing for energy and reserves: a responsive reserve market proposal. Decis. Support Syst. 40 (3-4), [Condren et al.(2006)] Condren, J., Gedra, T., Damrongkulkamjorn, P., May Optimal power flow with expected security costs. Power Systems, IEEE Transactions on 21 (2), [Karki et al.(2006)] Karki, R., Hu, P., Billinton, R., A simplified wind power generation model for reliability evaluation. Energy Conversion, IEEE Transactions on 21 (2), [Kintner-Meyer et al.(2007)] Kintner-Meyer, M., Schneider, K., Pratt, R., Impacts assessment of plug-in hybrid vehicles on electric utilities and regional u.s. power grids part 1: Technical analysis. Tech. rep., Pacific Northwest National Laboratory. [Mount et al.(2011)] Mount, T., Lamadrid, A., Maneevitjit, S., Thomas, B., Zimmerman, R., The hidden system costs of wind generation in a deregulated electricity market. The Energy Journal 33 (1), [Mount and Lamadrid(2010)] Mount, T., Lamadrid, A. J., jul Are existing ancillary service markets adequate with high penetrations of variable generation? In: PES General Meetings. pp [Mount et al.(2009)] Mount, T. D., Lamadrid, A., Maneevitjit, S., Thomas, R. J., Zimmerman, R., A symbiotic role for plug-in hybrid electric vehicles in an electric delivery system. In: Proceedings of the 22nd Rutgers Western Conference. Monterey, CA. [Shrestha et al.(2004)] Shrestha, G., Song, K., Goel, L., Strategic self-dispatch considering ramping costs in deregulated power markets. In: Power Engineering Society General Meeting, IEEE. p Vol.1. [Thomas et al.(2008)] Thomas, R., Murillo-Sanchez, C., Zimmerman, R., July An advanced security constrained opf that produces correct market-based pricing. In: Power and Energy Society General Meeting - Conversion and Delivery of Electrical Energy in the 21st Century, 2008 IEEE. pp

11 [Wood and Wollenberg(1996)] Wood, A., Wollenberg, B., Power Generation, Operation and Control, 2nd Edition. Wiley Interscience. [Zhang et al.(2000)] Zhang, D., Wang, Y., Luh, P., Aug Optimization based bidding strategies in the deregulated market. Power Systems, IEEE Transactions on 15 (3),

12 Table 1: Definition of Variables, simplified Formulation π tsk T Set of time periods considered, n t elements indexed by t. S t Set of scenarios in the system in period t, n s elements indexed by s. K Set of contingencies in the system, n c elements indexed by k. I Set of generators in the system, n g elements indexed by i. J Set of loads in the system, n l elements indexed by j. Probability of contingency k occurring, in scenario s, period t. ρ t Probability of reaching period t. G itsk Quantity of apparent power generated (MVA). G itc C G ( ) Inc its ( ) Dec it ( ) Optimal contracted apparent power generated (MVA). Cost of generating ( ) MVA of apparent power. Cost of increasing generation from contracted amount. Cost of decreasing generation from contracted amount. VOLL j Value of Lost Load, ($). LNS( ) jtsk Load Not Served (MWh). R it < Ramp i (max(g itsk ) G itc ), up reserves quantity (MW) in period t. C R ( ) Cost of providing ( ) MW of upward reserves. R it < Ramp i (G itc min(g itsk )), down reserves quantity (MW). C R ( ) Cost of providing ( ) MW of downward reserves. L it < Ramp i (max(g i,t1,s ) min(g its )), load follow up (MW) t to t 1. C L ( ) Cost of providing ( ) MW of load follow up. L it < Ramp i (max(g its ) min(g i,t1,s )), load follow down (MW). C L ( ) Cost of providing ( ) MW of load follow down. Rp it ( ) Cost of increasing generation from previous time period. Rp it ( ) Cost of decreasing generation from previous time period. 12

13 Table 2: Summary of Generation Capacity and Load Capacity per Fuel Type (MW) Total Cap. Load Location (RTO) coal ng oil hydro nuclear isone 7,020 12,813 4,051 2,657 17,947 44,488 23,847 maritimes 478 3,682 1, ,352 3,546 nyiso 3,397 4, ,146 2,973 18,495 38,274 ontario 3,106 11,464 3,282 1,441 2,383 21,731 21,158 pjm 14,561 14,611 9,191 2,604 12,500 53,467 51,588 quebec Total 28,562 46,681 18,530 14,048 35, , ,412 a Values shown are taken as peak values. Table 3: Summary of Key Results Case 1 Case 2 Case 3 Case 4 Case 5 Operating Costs ($1000/day) 102,266 83,015 83,527 78,981 79,272 Ramping Costs ($1000/day) 0 0 1, Gen. Net Revenue ($1000/day) 261, , , , ,732 Congestion Rents ($1000/day 10,618 31,010 30,397 18,275 17,245 GenCap All (MW) 138, , , , ,752 Wind Energy (MWh) NA 254, , , ,868 Table 4: Annualized Capital Costs Case 1 Case 2 Case 3 Case 4 Case 5 No Normal W.ramp W.rampDL W.DL Wind W. Operating Costs ($/MWh) Capital Cost ($/MWh) Total OperatingCapital Cost ($/MWh) 13