Increasing Backup Generation Capacity and System Reliability by Selling Electricity during Periods of Peak Demand

Transcription

1 Increasing Backup Generation Capacity and System Reliability by Selling Electricity during Periods of Peak Demand Elisabeth A. Gilmore a & Lester B. Lave a,b a: Department of Engineering and Public Policy, b: Tepper School of Business Mailing address: Department of Engineering and Public Policy Carnegie Mellon University Baker Hall 129, 5 Forbes Ave Pittsburgh, PA, USA, TEL (412) ; FAX (412) ; eagilmor@andrew.cmu.edu Abstract Meeting electricity demand during the 2 hours of greatest use each year is costly and challenging, especially in transmission constrained urban centers. Allowing installed backup and emergency generators to sell electricity for profit during peak demand and participate in the installed capacity (ICAP) market could lower the price of electricity and increase the reliability for all customers, not just the owners. The additional revenue could also provide an incentive for more customers to buy backup generators, adding still more reliability to the system. Here, we evaluate the economic potential of using backup generator for meeting peak electricity demand and ICAP requirements in the New York City Long Island (NYC-LONGLI) region in New York (NY) State. First, we calculate the implicit value of unserved electricity as revealed by the investment in backup generator. Second, we evaluate the profit that would be earned by a backup generator available during the 2 peak hours each year, using the NYC and LONGLI zonal prices in 25 and in 26 as a case study. The profits from real electricity are calculated as the zonal prices minus the marginal cost of the generator. The profits from the ICAP market are equal to the clearing price minus the opportunity cost of being ready to operate. We then subtract the available profits from the costs of ownership (fixed costs) to estimate the reduced implicit value of reliability, or the adjusted reservation price for purchasing a generator. We then estimate the increase in backup capacity. Finally, we address the effect of having all backup generators, rather than a single one, provide electricity to the market. We find that the implicit value of unserved electricity is much greater than the estimates in literature, since owners of backup generator reveal themselves to value electricity during lost service at approximately 26$/kWh or 85.9 $/kw year. A single backup generator selling during the top 2 hours in NYC or LONGLI and participating in the summer and winter ICAP markets could generate sufficient revenue to lower its costs by approximately 8% with most of the profits from the ICAP market. If additional backup generators could earn this revenue, we estimate that this would result in adding 4 MW of backup generator capacity in NYC and 115 MW in Long Island. However, as additional backup generators produce power during the peak hours, they reduce the market clearing price for electricity. They also reduce the price in ICAP markets. If all of backup generation were bid into the market, we estimate that the market clearing price would equal the generation cost of these backup units, generating no revenue beyond variable cost. Similarly, the price in the capacity market would fall to zero. While making all the backup generators available during peak hours would not generate much additional revenue for the owners of the generators, it would save electricity customers a great deal of money, especially in the capacity market, as well as increase system reliability. We estimate a savings of $66 million and $75 million in electricity payments in NYC and LONGLI, respectively, and $1 billion and $5 million in the ICAP market in NYC and LONGLI, respectively. 1. Introduction Matching supply to demand when demand for electricity is very high presents a major challenge for the US electricity sector. Unlike other commodities, electricity supply and demand must be met at all times or at least some customers will face curtailments (e.g. rolling brownouts). Under severe conditions, there will be a blackout. The problem of meeting peak electricity demand is becoming more acute. Load is projected to grow, while investment in new generation and transmission is lagging behind demand. In addition, there are substantial difficulties in siting new facilities, especially in highly populated urban areas where most of the growth in demand is anticipated. For example, in New York City (NYC) 1

2 and Long Island (LONGLI), load is projected to grow by more than two percent per year. A recent study by the New York Independent System Operator (NYISO) concluded that by 211, without major investments in generation and transmission capacity, the region will be in violation of its reliability criteria (1). In a competitive electricity market, electricity prices should provide a sufficient signal to stimulate investment in the quantity, the location and type of generation needed to meet demand. The primary price signal is from selling real power. In its simplest form, a supply curve is constructed from bids from the available generators who in a perfectly competitive environment have the incentive to bid their marginal cost. This supply curve is intersected by the expected load. All generators that are dispatched to meet this demand are paid the marginal cost of the final generator dispatched. This price is referred to as the market clearing price (MCP). For generators with marginal costs less than the MCP, the difference between the MCP and the marginal cost is profit which is used to recover fixed capital and operating & maintenance (O&M) costs. However, if bids are restricted the marginal cost, a plant that is dispatched only at peak demand can never receive more than its marginal cost since it sets the MCP. In reality, generators can bid above their marginal costs when supply is constrained, especially if there are pivotal suppliers in the market. To prevent price spikes, many markets have established price caps (no greater than 1 $/MWh) to limit this type of bidding. While price caps are politically convenient, they reduce the ability of generators to recover their costs. A generator incurs certain costs to available to provide electricity. If it cannot recover these costs in the electricity market, the generator should shutdown. This, however, has implications for reliability. To provide sufficient profits to ensure adequate reserves and promote the construction of new generation, installed capacity (ICAP) markets have been developed (2). Several competitive electricity regions, including the PJM regional transmission operator (RTO), the independent system operator (ISO) New England (ISO-NE) and the New York ISO (NYISO) have instituted these markets. The ICAP markets are very much in flux. The NYISO market is a hybrid system where loads can acquire the stipulated reserves from bilateral trades and self provision with the balance purchased in a market. Similar to the market for electricity, the capacity market clears where supply meets demand, although in this case, the NYISO constructs the demand curve. These markets, however, are controversial and in the NYISO, the payments are expensive (3). At the same time as the markets are experiencing difficulty meeting demand, electricity consumers are demanding higher reliability. A range of consumers, for example financial services, hospitals and universities, with a high value load or high economic value for reliability have installed substantial amounts of backup or emergency generators (at least for their critical services). It is estimated that there is approximately 1,32 MW of backup capacity in New York City and 5 MW in Long Island (4). Presently operating only when the grid fails (approximately 2-3 hours/year in the United States) (5), these generators represent a substantial source of underutilized capacity. Since the capital costs of the backup generators can be ascribed to the reliability application, installed backup generators could operate at their marginal costs for peak electricity, increasing grid reliability and supporting electricity delivery (6). Further, allowing these generators to run for economic profit could also increase the total backup generator population, increasing overall reliability. Operating these generators could be very valuable in the New York City area as it experiences very high peak electricity prices, is severely transmission constrained and has not experienced sufficient investment and construction of new generation. There are also potential benefits in the capacity market, dramatically increasing the system reserves. There are, however, barriers to this strategy. First, there is the need for retrofits to the interconnections to allow the generator to operate in parallel with the grid. Second, most of these generators are diesel fuelled with limited environmental emission controls. Previous work on using these generators for meeting peak electricity demand showed that emission controls, specifically diesel particulate filters (DPF), are required to limit air quality and human health effects (7). While this is a suitable retrofit for some of the installed backup generators, newer diesel generators conform to the Environmental Protection Agency (EPA) emission standards for off road generators. Ideally, allowing these generators to operate for peak power and reliability would generate sufficient profit to encourage the purchase of new cleaner generators. Finally, the profits must be sufficient to justify the expense of dedicating resources to operate the backup generators during periods of peak electricity. Presently, backup generators are permitted into two NYISO programs: 1) the special case capacity reserves (SCR) and emergency demand response program (EDRP). In both programs, the backup generators operate when called upon by NYISO under conditions of system stress. These programs, however, do not capture the full value that these generators could bring to the market in terms of lowering electricity prices at peak electricity demand, reducing capacity payments and improving system reliability. In addition, the NYISO has expressed its intent to investigate protocols for bringing these generators into the regular markets, possibly by aggregating them to create megawatt (MW) size capacity blocks and decrease concerns about generators failing to comply. In this work, we quantify the potential of using backup generators for meeting peak electricity demand in the New York City (NYC) and Long Island (LONGLI). Specifically, our objectives are as follows: 2

3 1. To examine the available profits for backup generators selling during periods of peak electricity demand and participating in the capacity markets, and investigate how these profits influence the quantity of backup generators capacity; 2. To explore the system effects (simultaneity problem) of how profits will change with the additional backup generators; and, 3. To evaluate the system (social) benefits, specifically in reduced capacity market payments and improve reliability, and to what extent inducements are justified for this strategy. 2. Literature Review 2.1 Economics of reliability and backup generation The economic decision to install backup generation is a function of the value that a firm places on electricity. The value a consumer puts on an unsupplied MWh of energy is frequently defined as the value of lost load (VOLL). Since there is no market for unsupplied electricity, the VOLL is determined by indirect methods, and literature reveals numerous techniques (8). The most common method is the use of customer surveys. Industrial, commercial and residential sector customers are asked identify their costs from power interruptions. While this technique has been widely employed, there are concerns that individuals may overestimate their costs. To overcome this issue, economists also employ indirect methods. Telson (1975) popularized a macroeconomic approach, where VOLL is estimated as the gross economic output divided by electricity consumed (9). In 24, the US GDP was 11.7 trillion dollars and electricity consumption was 3.7 trillion kwh leading to an implied VOLL of 3.15 $/kwh. This value, however, does not account for the damages observed from an unplanned outage. This criticism prompted a second approach which evaluates the direct and indirect financial damages from an actual outage. For example, an evaluation of the New York City outage in 1977 yielded a VOLL of 13 $/kwh (in 27 dollars) (1). In Table 1, we present a range of literature values for VOLL in 27 dollars for the United States by consumer 1. Some researchers have noted a recent increase in VOLL (e.g. (11)), attributing it to the growth in importance of communication networks. Table 1: Mean values of lost load (with low and high values) (data from (8, 12)) Consumer Value in $/kwh Industrial 9.38 ( ) Commercial 1.4 ( ) Residential.8 ( ) United States Average 6.86 ( ) A different approach to estimating the VOLL is to assume that the cost of self-generating during an outage is equal to the marginal cost of an outage. Bental and Ravid (1982) argue that a rational consumer of electricity who would like insurance against all or some of the damages caused by a power outage will acquire backup generators. Since these generators are relatively expensive, the firm has to choose the optimal amount. For a profit maximizing, risk neutral firm, the optimal amount is found when the expected cost of generating the marginal kwh is equal to the expected gain from that kwh, which is also equal to the expected loss from the marginal kwh not supplied by the utility. They calculated a VOLL of 1.16 $/kwh (1982) or 3.5 $/kwh (27) for an expected outage of 1 hours per year (13). Presently, the US grid operates at an availability approaching 4 nines or approximately 99.96%. This corresponds to about 3.5 hours of outage per year. LaCommare and Eto (26) estimate that the average duration of an outage was 16 minutes with a standard deviation of 54 minutes and that outages occurred at a frequency of 1.2 events per year with a standard deviation of.5 events per year (5). Using these figures, we estimate 2.12 hours of outage per year. Even 2 hours, however, can have serious consequences for a wide range of electricity consumers, especially those that rely on communication networks such as banks, internet service providers, on-line trading and research centers. A wide range of 1 While these are point estimates, we note that in reality VOLL is also a function whether the outage is planned or unplanned, the time of day and the time of year and the duration of the outage. 3

4 other consumers also rely on highly reliable electricity. Examples of sources of backup generators include hospitals, data centers, high-rise buildings, communication centers, water and wastewater treatment plants, military facilities, and airports (14). In addition, a facility is likely unable to resume full operations precisely when the power resumes, leading to a further loss of productivity. To protect against the disruptions caused by power failures, many firms have installed backup generators. The Federal Energy Regulatory Commission (FERC) estimates that there are about 12 million DG units installed across the country, with a total capacity of about 2 GW and that most of these are backup or emergency generators (15). In the Northeast states, it is estimated that the population of diesel internal combustion engines is well over 3, units with a combined capacity greater than 1 GW (4). Approximately 8% of the engines or 74% of the total capacity are estimated to be installed for emergency usage. Normally, a facility sizes the generators to meet the critical load of the facility. For some facilities such as hospitals this is a result of the legislation. For other facilities, Beenstock (1991) explains this behaviour has a reflection of either the high cost of self-generated power or the low assessment of the output loss relative to the equipment loss (16). In most jurisdictions, generators permitted for emergency usage are limited to operate only when the grid fails and for a maximum annual usage. For example, in New York State, the generators are limited to 5 hours per year and are exempt from emission standards or controls. 2.2 Backup generators and participation in electricity markets Regardless of the value of reliability of individual firms, once the generator is installed its operation is frequently limited to less than 2 hours per year for outages and approximately 2 hours per month for routine maintenance. Given the cumulative capacity in backup generators, their location near the load and their infrequent usage, several independent system operators (ISOs) have investigated or implemented programs that harness the backup generators during periods of peak electricity demand. In a recent survey conducted by the Federal Energy Regulatory Commission (FERC), it is estimated that there are 61 programs run by ISOs that make use of backup generators of which 27 are emergency demand response program, 16 are capacity market program, and 18 are demand bidding/buyback programs (17). These generators, however, do not participate in regular markets and are paid non-market prices for their participation. The NYISO operates three demand response programs which allow a load (either a firm or an aggregation) to profit from reducing its electricity demand either through efficiency or by shifting to a backup generator: 1) an Emergency Demand Response Program (EDRP); 2) a Day-Ahead Demand Response Program (DADRP); and, 3) an Installed Capacity Special Case Resources (ICAP-SCR) (18). The EDRP provides incentive payments to customers for load reductions during periods when reserve shortfalls arise. The ISO notifies the participants who have voluntarily enrolled to curtail their load serviced by the grid when called upon. They are, however, not required to actually curtail and face no penalties. In the NYISO, they are paid the higher of 5 /kwh or the wholesale electricity price in the generators area (known as the locational-based marginal price, LBMP). Since they are not required to curtail, they are not eligible to receive capacity payments from the ICAP-SCR. By contrast, the DADRP and ICAP-SCR have mandatory curtailment requirements, and the loads are subject to a penalty if they fail to meet their commitments. In addition, participants in DADRP are not permitted to use on site generation, while only 11% of the ICAP-SCR are from onsite generation in 24 (19). Participants in DADRP bid their load reductions on a day-ahead basis into the wholesale electricity market. These bids then compete with generators offers to meet demand. If accepted, the bidder is scheduled to reduce load during specified hours the following day. The bidder may specify a minimum payment, called a curtailment initiation cost with the participant receiving the higher of the curtailment cost or the hourly LBMP times the scheduled load. The ICAP-SCR program is designed primarily to provide an incentive for businesses and large power users to shutdown. In exchange, they are paid a capacity payment specified by the ISO upon program registration in advance for agreeing to cut power usage upon request. Payments ranged from 1.25 $/kwh on average in New York State to $/kwh in New York City. All these programs have met some success, and the NYISO s ERDP program is credited with providing an important resource during periods of reserve shortage during the August 23 blackout (17). There is, however, some controversy surrounding these programs as they require a system subsidy to ensure availability and are tightly controlled by the ISO rather than allowing the generators respond directly to market signals. In addition, in 23, the ERDP events were called only after the August 23 blackout, and there were no events in 24 (19). This situation leaves these backup generators available both to moderate high electricity prices and peak demand as well as to enhance system reserves. Presently, the NYISO has announced its intent to try to allow these generators to operate in their energy, ancillary and capacity markets. 4

5 3. Method and Data In this section, we describe the method we employed to evaluate the potential benefits of allowing the backup generators located in NYC and LONGLI into the regular NYISO markets. First, we calculate the implicit value of reliability from a backup generator. Second, we evaluate the profits available for a single generator in the electricity and installed capacity (ICAP) markets and calculate a reduced value of reliability. Third, we estimate the increase in the backup capacity in the NYC and LONGLI if the reservation price for purchasing backup is reduced. Fourth, we consider the effect on profits if all the previously installed capacity and the new capacity installed at the reduced reservation price enter into the electricity and ICAP markets. Even if there is little or no private incentive for the generators to operate, there is still the potential to reduce electricity prices, capacity payments and improve reliability by increasing system reserves. We calculate the system wide benefits from bringing these generators into the market. Finally, we contrast the value of this strategy in a regulated and deregulated market. For this analysis, we use market data from 25 and 26 from NYC and LONGLI as a case study. 3.1 The implicit value of reliability from a backup generator Following the approach of Bental and Ravid (1982), we calculate an implicit value of unserved electricity as the cost of the backup generator with no electricity or capacity market sales. A customer who installs a generator must believe that their cost of unserved load is at least this high. The implicit value is the total unit cost of a backup generator (TC) which consists of the levelized capital and fixed operation costs (FC) and the marginal or variable unit costs (MC) of operation: TC = FC + VC. Eqn 1 CC CRF FOM FC = + HY HY MC = VOM + FUEL Where CC is the overnight capital cost (in $/kw) CRF is the capital recovery factor which includes interest and depreciation (~ 1 %) FOM is the fixed operating and maintenance (O&M) ($/kw-yr) FUEL is the fuel cost ($/kwh) HY is the number of hours the plant generates electricity (hours/year) VOM is the variable O&M ($/kwh) We calculate the value of reliability based on the cost of installing and operating emergency backup for the three primary backup generator options: a diesel fueled internal combustion engine (ICE), a natural gas fueled ICE, and a natural gas microturbine. In Table 2, we present the data for four different backup configurations: the three generator options and a diesel ICE retrofitted with a diesel particulate filter (DPF). Since the actual costs are a function of the specifics of the generator installation, these costs are meant to be indicative. Table 2: Costs of Backup Generators with and without Emission Controls (2, 21) Generator Type Efficiency (%) Capital Cost ($/kw) Fixed O&M ($/kw-yr) Variable O&M ( /kwh) Diesel ICE 35% Diesel ICE w DPF 34% Natural gas ICE 29% Natural gas microturbine 25% In any given year, the price of fuel will determine the variable cost of operating the engine with the fuel cost is calculated as follows: FUEL = P HR. Eqn 2. fuel Where P fuel is the price of fuel (in $/mmbtu); and HR is the heat rate (in mmbtu/kwh). 5

6 The heat rate equals 3,412 Btu/kWh or.3412 mmbtu/kwh (the conversion of energy (Btu) to electricity (kwh) at 1% efficiency) divided by the efficiency from Table 2. The price of natural gas and oil also influence the electricity prices, especially during peak hours where the marginal plant is fuelled by natural gas, residual distillate and oil distillate. The annual average natural gas price at the city gate was $/mmbtu on average in 25 and 26. For diesel, the price was approximately 14.9 $/mmbtu on average in 25 and 26, respectively (22). For the diesel ICE with a DPF, we add an incremental cost for utilizing low sulphur fuel (~ 1 /gallon). 3.2 The profits and increases in backup capacity Electricity markets Selling power into the electricity markets has the potential to be a profitable venture for the owner of a backup generator. We calculate the profits in $/kw year with equation 3: n Profits = ( MCP n MC backup ) Eqn 3. 1 Where MCP is the market clearing (in $/kwh); MC is the marginal cost of the backup generator (in $/kwh); and, n is the number of hours per year where MCP > MC. The NYISO has a two settlement process: a day-ahead market (DAM) and a real time market (RT). About 45% of the electricity is cleared through the DAM with any deviations from the projected demand met by the RT market. The amount of power traded on the real time market is less than 1 %. The remaining 5% is sold through bilateral contracts. In Appendix A, we show summary statistics for the DAM prices and RT prices. The average prices during the top 2 hours are two to three times the yearly average in both the NYC and LONGLI zones. The average peak prices in 25 and 26 are relatively consistent, but the average prices in 25 are higher than 26 due to higher natural gas prices. Similar other studies, we find that the RT market is more volatile than the DAM. This is due in part to the incentives for bidding into the DAM as well as the stronger market mitigation measures and price caps in the DAM compared to the RT market. Since this is an ex-post analysis, we assume that the owner of a generator has perfect information about the prices and duration of the spikes, and hence can maximize its profits. For simplicity, however, we assume that the generator chooses to participate in either the DAM or the RT market. We do not develop an optimal bidding strategy. All market data was extracted from the NYISO website (23). Generators also have the options of bidding to provide ancillary services, such as regulation and frequency response, 1 minute spinning reserves, 1 minute non-spinning reserves and 3 minute operating reserves. The engines would be ineligible for regulation and spinning reserve since they cannot respond for frequency response nor are they operating at all times. The generators, however, could potentially operate in the 1 minute non-spinning reserve and the 3 minute operating reserve market. While 1 minute non-spinning reserve would be an ideal application for backup generators, the most lucrative ancillary service market is for regulation. Previous work has shown that during the summer, real power MCPs are higher than 1 minute non-spinning and 3 minute operating reserves (24). This is generally the case with this data. As the number of new generators entering the market increases, however, the arbitrage opportunities will decrease and the backup generator may become the marginal plant. The NYISO estimates that there is approximately 1,36 MW of generation in the NYC zone and 5,32 MW of transmission capacity into city. Approximately, 4, MW arrives from upstate New York, 1, MW from New Jersey (PJM) and 3 MW from Long Island. During periods of peak loads, however, the transmission capacity is lower due to transmission constraints and local reliability rules (27). Generation capacity in Long Island is 5,767 MW with 1,731 MW of import capacity. This value does not include the new 66 MW PJM-Neptune Line which became operational in June July 27 (25). We take two complimentary approaches to evaluate whether the backup generators would become the marginal plant in the electricity market: 1) we construct theoretical zonal short run marginal cost curves (SRMC) using actual plant capacities and heat rates to estimate the room for arbitrage in the system, and 2) we compare the estimates from the system curves with estimates from real load-price relationships on peak days. To construct the SRMC, we use data on plant capacity and heat rates from the Emissions & Generation Resource Integrated Database (egrid) compiled by the Environmental Protection Agency (26). We also estimate that the 87.4% of the nameplate capacity is available 6

7 accounting for forced outages, maintenance schedules and reserve margins. We make the following assumptions for generators without heat rates in Table 3 and fuel costs in Table 4. The fuel costs are slightly lower for the electricity sector than the individual customers who own backup generators. We include imports equal to the transmission capacity and assume it is generated at 1 $/MWh. Since we do not know the actual transmission capacity at peak loads, we treat this as an adjustable parameter such we meet the peak loads with the marginal generators. Using the curve, we evaluate how many MW of capacity are dispatched with marginal costs above that of the backup generators. High prices, however, do not always correspond with high loads. This curve is a simplification of the actual dispatch. Frequently, on peak days, unexpected events such as outages or unexpected demand cause real time price to rise substantially even at lower loads. To provide a check on the modelled marginal cost curve, we isolate the load-price relationship for each day for days with the top 2 hourly loads. We fit each day to a curve and use this curve to estimate how much the price would reduced if all the available backup was available to be dispatched. Table 3: Marginal Costs for Generators (in $/MWh) Generation type Marginal Cost Hydro 1 Wind 1 Nuclear 2 Pumped Hydro 5 Biomass 5 Table 4: Fuel Costs (in $/mmbtu) Generation type Marginal Cost Coal 2.2 Natural Gas 7.5 Residual Fuel Oil (No. 6) 8.5 Distillate Fuel Oil (No. 2) Installed Capacity Markets Even if there is no profit in selling real power, there may, however, still be a profit for these generators in the installed capacity market (ICAP). For NYC and LONGLI, 8% and 99% of the regional peak load capacity must be procured from resources located in these regions, respectively. In Table 5, we show the monthly capacity payments in 25 and 26 for NYC and LI for summer (May October) and winter (Nov April). Table 5: Average seasonal ICAP clearing prices (in $/kw-month) (28) Season NYC LI NYC LI Winter Summer The NYISO ICAP market is a hybrid system combining a quantity based system and a demand curve system. In the quantity based system, load service entities (LSE) are required to procure a certain amount of generation capacity as calculated by the NYISO to meet their loads with a reserve margin. A part of this requirement is obtained through bilateral trades or self-supply (e.g. demand response and interruptible loads). This is combined with a market where the NYISO purchases capacity on behalf of the load based on a demand curve (2). In Figure 1, we present the demand curve for the ICAP market for 25 and 26. UCAP is known as unforced capacity which is equal to the installed capacity minus some allowance for reliability of the capacity. It is difficult to evaluate precisely how adding the backup generators would influence this market since the NYC market has been subject to gaming. We assume in our analysis, however, that this is a functioning market. 7

8 25 2 Figure 1: Demand curve for ICAP for summer of 25 and 26 NYC 25 NYC 26 LI 25 LI 26 Price (in $/kw-month) Unforced Capacity (in MW) 4. Results and Discussion 4.1 Economics of backup generators for reliability The economic argument for using backup generators for meeting peak electricity demand is that the capital cost is already attributed to the reliability application. The average outage lasts 1.77 hours and occurs 1.2 times per year for 2.12 hours per year of outage (8). However, the number of hours of outage per year is only an indication of how much operation time the firm would loss if it did not have backup; it takes time to restart equipment. Previous work has suggested that it takes from 1 hour to 4 hours to restart after an outage (29, 3). Therefore, an average firm would buy a generator to protect against 3.3 to 6.9 hours per year. In Table 6, we show the implicit value of reliability for the three backup generator options in $/kwh and $/kw year assuming the minimum 3.3 hours per year of outage. These values are at the very high end of the VOLL literature values for commercial customers and are more than double the average VOLL. These figures suggest that either the owners of these generators are acting irrationally or that the literature VOLLs are far too low at least for these owners. It is, however, hard to believe that all 1.3 GW of backup generators were installed by irrational owners. In Figure 2, we show the sensitivity of the implicit reliability value for a diesel ICE to the number of hours of outage per year. As expected, as the number of hours of outage increase, the value of reliability is reduced. Table 6: Implicit value of reliability for backup generators Backup generator $/kwh ($/MWh) $/kw year ($/MW year) Diesel ICE 25.9 (25,9) 85.3 (85,3) Natural gas ICE 27.4 (27,4) 9.3 (9,3) Natural gas microturbine (28,95) 95.4 (95,4) The slightly lower implicit value of reliability helps explain why diesel fuelled generators are the most popular form of emergency backup. Presently, most of the installed generators are diesel fuelled. The diesel generators also have characteristics that make them highly suited for the backup application: a very short start up time, high reliability and efficiency, and have low maintenance costs. By contrast, natural gas engines have lower efficiencies and have longer start up times. Additionally, depending on the reason for the outage, the natural gas fuel system may be shut off. Fuel cells are an alternative to combustion based systems, but are very expensive. Other options like photovoltaic cells and wind turbines cannot be dispatched on demand, and therefore, are not suitable for backup. Also, we do not consider 8

9 uninterruptible power systems (UPS) such as batteries in this analysis. Many backup systems include UPS to protect against voltage fluctuations and to avoid interruption as the engines ramp up to full load. These systems, however, are designed to last no more than an hour and do not provide long term reliability. Figure 2: Implicit Reliability Cost for a Diesel ICE as a Function of Hours of Outage (in $/kwh) 35 3 Implicit Cost of Reliability ($/kwh) Hours of Outage 4.2 Economics of using backup generators for peak electricity demand and for installed capacity In Figure 3, we show the marginal and average (levelized) cost of operating the different backup configurations from Table 2. The levelized cost represents a technology s average costs per kwh over its life span. We assume 2 hours a year of peak power operation (31) and a discount rate of 1%. We show the average cost in two cases: 1) assuming that the capital cost of the diesel ICE is covered by the reliability application, and only the incremental cost above the diesel ICE is not covered for the natural gas options; and 2) assuming that the entire capital cost is covered in the levelized cost. A well maintained diesel generator has a lifetime on the order of thousands of hours. Adding 2 hours per year of operation will not shorten its useful life. In addition, these units are discarded long before the units is worn out (i.e., the life span of the unit is determined by age not by the hours of operation). Finally, using the engine for peak power could improve the probability that the engine will turn on in the case of an outage. There are two other costs that we show in the levelized cost that are not captured in the marginal costs of operation. First, most emergency generators will require a retrofit on the electrical interconnections to operate in parallel with the grid. An average cost estimate for these interconnections is approximately $2/kW, although the actual cost depends on the size of the engine, the configuration of the electrical network and the utility requirements (32). Secondly, we show a diesel retrofitted with a diesel particulate filter (DPF) to prevent severely harming human health. Over the past decade, the Environmental Protection Agency (EPA) and state environmental agencies have promulgated increasingly stringent emission standards for off-road diesel generators, known as Tier I through IV standards. Generally, emergency and standby generators are not subject to these standards, generating only when the grid fails. In New York, an emergency generator is permitted to operate up to 5 hours per year with emission controls. If these generators were to operate concurrently with the grid, the regulations would most likely be enforced. In the ICAP market, a generator has the incentive to bid the cost per kw of keeping the plant on standby rather than shutting down when the LBMPs do not cover their marginal and fixed costs of being available. Since a backup generator must always be ready to operate, the opportunity cost of being available is zero. 9

10 75 Figure 3: The Marginal and Average Cost of Generators (in /kwh) Diesel ICE Diesel ICE w DPF Natural Gas ICE Natural Gas MT Price (in cents/kwh) 5 25 Marginal Costs Average Cost w Reliability Total Average Costs 4.3 Profits from using backup generators to sell peak power Since 8% of the installed capacity is diesel generators, we will restrict the remainder of our analysis to diesel generators. In order to make a profit in the electricity market, the market clearing price (MCP) must be higher than marginal cost of using the generator (i.e., MC of an uncontrolled diesel = $/MWh or MC for a controlled diesel = $/MWh). In Tables 7 1, we present the number of hours of profitable operation for the controlled and uncontrolled diesel generator in the DAM and RT markets as well as the profits for the top 2 hours and for up to the top 5 hours. Table 7: Profits for an uncontrolled diesel in New York City (in $/kw year) DAM 25 RT 25 DAM 26 RT 26 # of h above MC All Profits 16,48 62,68 7,95 34,56 Profits (5 h) 16,4 59,73 7,95 34,56 Profits (2 h) 12,36 46,59 7,95 34,25 Table 8: Profits for a controlled diesel in New York City (in $/kw year) DAM 25 RT 25 DAM 26 RT 26 # of h above MC All Profits 7,55 47,54 5,93 29,52 Profits (5 h) 7,55 47,54 5,93 29,52 Profits (2 h) 7,48 41,71 5,93 29,52 Table 9: Profits for an uncontrolled diesel in Long Island (in $/kw year) DAM 25 RT 25 DAM 26 RT 26 # of h above MC All Profits 2,65 98,5 3,12 78,4 Profits (5 h) 19,2 88,28 3,12 78,4 Profits (2 h) 12,9 62,75 27,72 66,15 1

11 Table 1: Profits for a controlled diesel in Long Island (in $/kw year) DAM 25 RT 25 DAM 26 RT 26 # of h above MC All Profits 8,23 78,49 23,21 67,53 Profits (5 h) 8,23 76,8 23,21 67,53 Profits (2 h) 8,2 57,87 22,84 61,27 In Tables 11 and 12, we examine the profits for a generator in the ICAP market. For both an uncontrolled and controlled diesel engine, the opportunity cost of being prepared to turn on is $/MWh since the backup generator must be ready operate all times. An uncontrolled generator, however, cannot operate all year long in its present configuration and is limited to 5 hours of operation a year. To maximize profits, this generator would operate in the summer only. A diesel generator retrofitted with emission controls could potentially bid into the capacity market all year long. Finally, we assume that the engines will be downrated by 95% of their nameplate capacity to generate an unforced capacity (UCAP). Table 11: ICAP profits in New York City (in $/kw year) Uncontrolled diesel Controlled diesel Table 12: ICAP profits in Long Island (in $/kw year) Uncontrolled diesel Controlled diesel To generate peak power and act as installed capacity, the generator must be retrofit with the interconnections to operate in parallel with the grid at 2 $/kw year. In addition, it is possible, however, that a generator without emission controls would not be allowed to operate at all in the capacity market. The DPF retrofit costs 4 $/kw or 4 $/kw year at a 1 % discount rate. In Tables 13 and 14, we present the total profits available in the New York City and Long Island minus the retrofits and calculate the change in the implicit value of reliability. We show the profits as the average of the DAM and RT markets. On average, an uncontrolled diesel generator can make sufficient profit to reduce the implied cost of reliability from more than two times the average to below the average VOLL for commercial and industrial facilities. The VOLL for the uncontrolled generator, however, is still double the VOLL for residential customers. However, a diesel generator which has been retrofitted with a DPF can operate for more than 5 hours per year. The extra profits from the capacity market reduce the implied VOLL to $/kwh in New York City and 1.67 $/kwh in Long Island. Table 13: Average profits and reduced implicit reservation price in New York City Generator Peak Electricity ($/kw yr) ICAP ($/kw- yr) Total ($/kw-yr) Profits w Retrofits ($/kw yr) % of Cost of Reliability Reservation Price ($/kwh) Diesel ICE % 3.19 Diesel ICE w DPF % < Generator Table 14: Average profits and reduced implicit reservation price in Long Island Peak ICAP Total Profits w % of Cost of Electricity ($/kw- yr) ($/kw-yr) Retrofits Reliability ($/kw yr) ($/kw yr) Reservation Price ($/kwh) Diesel ICE % 5.9 Diesel ICE w DPF % Reductions in Price and Increases in Backup Power Capacity As the cost of reliability decreases, we should observe an increase in backup capacity. At the reservations prices that we calculate, we estimate that many more customers would be willing to pay to protect themselves from an outage. Determining the increase in capacity that would result from this decrease requires estimating the demand for reliability as 11

12 a function of price (e.g. an elasticity of demand for reliability). We know of no estimate of this demand, but there have been many studies of the demand for electricity as a function price. While the two concepts are different, the latter can give some insights into the former. Literature values for the elasticity of demand is generally between.1 and.4 (33). This implies that a 1% increase in price would results in a 1 4% reduction in electricity usage. Using these values as a rough guide of the demand for reliability, the 77% to greater than 1% decrease in the implicit cost of reliability should increase backup capacity by 8% in LONGLI for uncontrolled diesel generators to almost 5% for controlled diesel engines in NYC. With an installed backup capacity of 5 MW in LONGLI, we estimate an increase of 39 MW 154 MW of uncontrolled capacity or 46.8 MW to 187 MW of controlled capacity. For NYC, we estimate MW of uncontrolled capacity and MW of controlled capacity. On average, we assume 115 MW of new capacity in LONGLI and 4 MW of new capacity in NYC. We will use only controlled capacity since we assume that controls will be required to operate in the markets. Bring these additional generators into the market will increase the reliability in the system as more customers will be protected in the case of blackout. At the same time, since these individuals would be selling into the market during periods where there are presently reserve shortages, there would also be an increase in system level reliability. 4.5 The Simultaneity Problem Presently, the high prices observed in the NYC and LONGLI markets arise when old peaking units with high heat rates are brought on line to meet high demand. As the existing and additional backup capacity is introduced into the market, the backup generators may become the marginal unit. In this limiting situation, there would be no economic profits available for these generators. Hence, there would be no motivation for the backup generators to enter into the electricity market or to invest in new backup capacity Electricity Markets To evaluate this simultaneity problem in the electricity market, we construct marginal cost curves for the NYC and LONGLI zones. In competitive markets like the NYISO, units are run in merit order economic dispatch. This means that the marginal costs of the generation units are sorted from lowest to highest. The lowest cost units are dispatched until the demand is met. In some parts of the grid, the least expensive plant cannot be dispatched to the location of demand due to insufficient transmission capacity. A more expensive generator closer to the load must be used. This is referred to as transmission constrained economic dispatch and leads to locational based marginal prices (LBMPs). Based on the theoretical SRMC curves for New York City, there appears to be little room for arbitrage at the peak loads. In NYC, the curve is especially steep with only 143 MW of generation operating with a marginal cost higher than a diesel generator with a DPF. We do not expect the profits to remain for NYC and no new generation would enter the market. In Long Island, there is 64 MW of capacity with a high marginal cost, specifically distillate oil fuelled turbines. As a result, there is the potential that all 615 MW of new capacity could be dispatched and not become the marginal plant. We recognize, however, that the SRMCs do not capture all the system dynamics, and specifically, cannot capture the features of peak load days. In Appendix B, we show the relationship between load and price for the two zones in 25 and 26. There is significant scatter in the price-load relationship, especially for real-time prices. While we have not yet reached this point in the analysis, we plan to verify the above results by isolating the days with peak load and investigate how many MW of generation could be added before the marginal cost is that of diesel with a DPF. Preliminary results suggest that the addition of the existing population of backup generators into NYC reduces the arbitrage opportunities to zero. In Long Island, we find some evidence that adding 615 MW might still leave opportunities for arbitrage. This analysis, however, did not include the new 66 MW Neptune PJM Line. With this additional capacity, we expect the arbitrage opportunities to decrease significantly. Adding the backup generators to the NYC market, however, would save customers a lot of money by reducing the peak electricity prices. As a first estimate of the benefits, we assume that if the day ahead market (DAM) price is higher than the marginal price of the diesel, the marginal cost is now the marginal cost of a diesel with a DPF (e.g $/MWh). Customers in NYC would save $67 million in 25 and $65 million in 26. Similarly, if we assume that the backup generators are the marginal unit in LONGLI, customers would save $36 million in 25 and $12 million in

13 1 75 Figure 4: Marginal Cost Curve for NYC with and without backup generation 1.7 GW 1.3 GW BASE Price (in $/MWh) Figure 5: Marginal Cost Curve for LONGLI with and without backup generation MW 5 MW BASE Price (in $/MWh) Capacity Market A major source of profit for the generators is the capacity market. To evaluate whether the generator will swamp the market, we combine the capacity market clearing price with the scheduled demand curve presented in Figure 1. In NYC, we find that there is room for 1,24 MW in the summer of 25 and 1,31 MW in the summer of 26 before the clearing price would drop to zero. This value is approximately the same as allowing all the already installed backup generator capacity into the market. In the Long Island market, we estimate that 512 MW could enter in 25 and 451 MW could enter in 26 before there is no opportunity for arbitrage. Again, this is approximately the same as the 13

14 installed backup capacity in Long Island. Although the individual owners cannot make money by bidding into the capacity market and no new backup capacity would be added by reducing the implicit cost of reliability, there is a social benefit to adding these generators to the capacity market. Reducing the clearing price to zero would save, approximately $1 billion dollars in New York City and approximately $5 million in Long Island. One issue that may limit the ability of backup generators to participate in ICAP markets is that it must be capable of providing 4 hours or more of capacity. Depending on the size of the generator, the New York City fire code may limit operation. Presently, no more than 275 gallons of fuel can be kept in a day tank per floor. For example, a 5 kw generator requires approximately US gallons per hour. For four hours of operation, gallons of diesel fuel needs to be available which falls within the regulations. While this is within the fire code, the fuel required for a 1 MW generator would exceed the fuel limit. To allow these generators to operate for extended periods of times for reliability, a more complicated arrangement is required. Properly constructed tanks are stored underground, and the fuel is pumped to the day tanks. There is no limit on how much fuel can be stored below grade, except that no single tank can exceed 2, gallons. Diesel fuel can be stored for up to 18 months. Assuming that the pumping does not dramatically change the economics, most backup generators could be classified as an energy-limited resource (ELR) 2 and would qualify to participate in the capacity markets. 4.6 Value of backup generation in a regulated market Using real market data, we have shown that there are significant social savings if properly controlled backup generators are allowed to operate for profit in real power and capacity markets. The majority of this savings arises from the ability to ascribe a portion of the costs of these generators to the reliability application. Although there are no additional profits (with the potential exception of some real power profits in Lomg Island), the existing backup capacity will reduce prices by operating. By contrast in a regulated electricity market, the utility faces a different set of trade-offs and incentives. A firm that has installed a diesel generator has already accounted for the capital costs as the value of reliability. These firms could operate at an average cost of less than.3 $/kwh even after the retrofit with a diesel particulate filter (DPF). This is significantly cheaper than a new natural gas peaker. For every kwh provided by the generator rather than a peaking plant, the system saves approximately.2 $/kwh. If the firm valued reliability at $/MWh, however, it would need to recoup all the capital costs at an average levelized cost of.71 $/kwh. This is significantly more expensive than a peaking gas turbine at approximately.5 $/kwh, but is cheaper than a lower end value of lost load (VOLL) of 1 $/kwh. Allowing these private owners to recover their average costs would clearly increase the amount of backup capacity in the system as it has the effect of reducing the reliability value to zero. In general, however, it is unclear that a regulated utility would choose the diesel ICE over the peaking plant as it depends largely on whether it can incite firms with backup capacity to generate without lowering the implicit value for reliability. 5. Conclusions, implementation issues and recommendations The US electricity grid provides electricity with a reliability approaching 99.99%. In past, some have argued based on the value of lost load (VOLL) that the electricity grid is too reliable (9). Using the actual cost of purchasing and operating backup generators for providing electricity during the approximately 3 hours of outage per year, we calculate an implicit cost of reliability of 26 $/kwh or 85.9 $/kw-year for customers who have backup generators. This value is significantly higher than literature VOLLs. With an estimated 1.3 GW of backup generation in New York City, approximately 1% of the load has a VOLL much greater than the values in the literature. In New York City and Long Island, it is difficult to site new generation and transmission which has led to high zonal prices and reserve shortages. Since these generators operate only when the grid fails, they represent a large source of underutilized generation capacity which could be used to alleviate these shortages. These generators would be especially valuable during periods of peak electricity demand of approximately 2 hours a year. Bringing these generators into the market, however, cannot be achieved by market forces. Using price data from 25 and 26 from the NYISO, we 2 An ELR is a resource that is unable to operate continuously on a daily basis, but is able to operate for at least four consecutive hours each day. Examples include a pumped hydroelectric plant or battery that must charge, and a generator with environmental restrictions. 14