Mead Wholesale Energy Market and Natural Gas Prices

Transcription

1 Date: October 1, 2015 To: From: Subject: Russ Callejo, Ron Smith and Kevin Black U.S. Bureau of Reclamation Tian Tian, David Hurlbut, and Jeff Cook National Renewable Energy Laboratory Mead Wholesale Energy Market and Natural Gas Prices Introduction This technical memorandum summarizes the results of the wholesale electricity price analysis conducted by the National Renewable Energy Laboratory (NREL) related to the Navajo Generating Station (NGS). The Mead Hub, to which NGS connects via its northern transmission line, is one of the most active wholesale power trading points in the Southwest. 1 This memo includes a market analysis of Mead Hub prices to assess how the all-in cost of a newly built natural gas combined cycle (NGCC) plant compares to recent and future wholesale market prices at Mead. This memo specifically addresses two questions related to Mead wholesale energy prices and the future of NGS. Have Mead Hub prices been high enough to support investment in a new NGCC generator? What wholesale energy prices can be reasonably expected at the Mead Hub through 2030, and how might this influence the economics of a new NGCC plant as it relates to a NGS glide path? We note at the outset that the decision to build new generating capacity in a regulated market environment is not necessarily price-responsive. A utility may decide to pursue building a new NGCC even if wholesale power prices are low. If a regulated utility anticipates that load growth and plant retirements will cause its physical reserve margin to shrink in the foreseeable future, it may seek regulatory authorization to build a new plant before wholesale prices rise to signal the future shortage. Thus, it is not necessarily true that wholesale power prices alone will arbitrate the balance between demand and the supply of generation in a classic economic sense. 1 Mead is specified as the switchyard adjacent to the Hoover Dam in southern Nevada. Mead is described as trading hub in SNL Financial Inc. data and a node in the CAISO system.

2 Utility-specific decisions for adding a new NGCC are not represented in this analysis. The aim here is to assess how observable prices can shape current and future market conditions, recognizing that these modeled conditions will affect but not necessarily determine actual capacity expansion decisions. We note further that this analysis is not intended to present a formal forecast of energy prices in the Southwest through Rather, the information presented here is a projection of trends and relationships that are observable today. The major findings are as follows. Mead day-ahead prices have increased from an annual average of about $38/MWh in 2009 to $47/MWh in Given these historical prices, neither the standard nor the clean spark spread indicate operating revenue from energy prices at Mead that are sufficient to cover the capital costs of a new NGCC. Mead Hub prices are likely to increase through 2030, but the standard spark spread does not indicate that operating revenue from energy prices at Mead would sufficiently cover capital costs of a new NGCC throughout much of the projection period. 2

3 Historical Mead Wholesale Prices The time series used in this analysis begins April 1, 2009, the go-live date of the California Independent System Operator s (CAISO s) current market design. From April 1, 2009 to October 31, 2014, Mead day-ahead prices have increased from an annual average of about $38/MWh in 2009 to $47/MWh in 2014 (Figure 1). 2 The decline in Mead prices from 2011 to 2012 corresponds to the decline in natural gas prices over that same period, while the subsequent rise can be attributed to a combination of natural gas price increases and the launch of California s greenhouse gas cap-and-trade program on January 1, $50 $40 $30 $20 On-Peak Off-Peak $10 $ Figure 1: Mead Hub Historical Day-ahead On-and Off-peak price ($/MWh) A spark spread analysis can offer insight into the economic viability of a new NGCC plant in a particular price environment. The spark spread is a measure of the theoretical operating profits of a power plant in the electricity market considering fuel costs as the only operating expense. 4 The spark spread measures the headroom for recovering all other variable costs, fixed costs, and capital costs. The formula for the spark spread is: SS! = P! (HR NG! ) 2 Day-ahead prices from each source are employed for this analysis because the vast majority of energy is transacted in the day-ahead market (as opposed to 15-minute- or 5-minute-ahead markets). According to the CAISO s Department of Market Monitoring, 97% of physical system load was transacted in the day-ahead market in CAISO (2015) Annual Report on Market Issues & Performance. p CAISO (2015) Annual Report on Market Issues & Performance. p For more information on spark spreads, see 3

4 where SS t P t HR NG t = spark spread at time t = wholesale power price at time t = heat rate corresponding to the type of CCGT examined = price of natural gas at time t If the energy from a new NGCC plant is imported into California, the rules of the state s greenhouse gas (GHG) cap-and-trade program require the importer to acquire and retire carbon allowances to offset the CO 2 emissions from the plant s operations. In practice, many generators have incorporated the cost of allowances as a component of broader costs, which include both the expense of buying natural gas and of procuring allowances to cover emissions from its combustion. This gives rise to a clean spark spread: where SSC! = P! (HR NG! + A! E) SSC t A t E = clean spark spread at time t = price of greenhouse gas allowances at time t = emission factor of the CCGT examined A spark spread is calculated for an assumed indicative heat rate. This analysis uses 6.5 MMBtu/MWh as representative of a highly efficient, newly constructed NGCC. 5 Natural gas prices are taken from CAISO s Southern California natural gas price index. Power prices are the day-ahead Locational Marginal Prices (LMPs) for CAISO s Mead node. The clean spark spread uses GHG allowance price as reported by the CAISO s daily index. This analysis bases a new plant s revenue requirement for recovering capital costs that is, the margin it needs to earn above wholesale power prices on the plant running at a 70% capacity factor. A new NGCC is assumed to cost $1,200/kW for wet cooling and $1,300/kW for dry cooling, financed over 30 years at a weighted average capital cost of 8.12%, with no significant change in constant-dollar capital costs over the period examined. This works out to $24/MWh for wet-cooled installations, and $28/MWh for dry-cooled installations. 5 The EIA suggests that the most efficient NGCC plants have a heat rate somewhat below 7 MMBtu/MWh. See 4

5 Figure 2: Spark Spread and Clean Spark Spread at the Mead Node Figure 3: Selective Spark Spreads at Mead (70% of Hours with Highest Prices) Neither the standard nor the clean spark spread indicate operating revenue from energy prices at Mead that are sufficient to cover the capital costs of a new NGCC facility (see Figure 2). The margin has been improving since 2012 but at no time has it reached the $24 benchmark for a new wet-cooled NGCC. Some owners may be less averse to market risk than others are, and one way to systematically quantify a higher-risk analysis is to calculate the spark spread on the basis of a perfectly selective dispatch. In this calculation, the spark spread is derived from the 70% of hours during the year when spark spreads are the highest. (Recall that a facility with a 70% capacity factor would only run part of the time.) That is, the NGCC plant hypothetically operates selectively and only during those 70% of hours when wholesale power prices are high and natural gas prices are low. (Figure 3). Even using this optimistic assumption of perfectly selective hourly operation, the standard and clean spark spreads rarely reach into the revenue requirement range. This supports the conclusion that recent Mead Hub prices generally have not been favorable to the construction of a new NGCC plant. 5

6 A new NGCC facility that is part of an NGS glide path strategy could also provide power for Arizona. Unlike CAISO, wholesale energy prices for Arizona s three vertically integrated utilities Arizona Public Service (APS), Salt River Project (SRP), and Tucson Electric Power (TEP) are not set by day-ahead and real-time auctions based on a security-constrained economic dispatch. They also do not presently face directly assigned emissions costs. If approved by the utility s regulators for inclusion in rate base, the plant s capital costs would be paid through rate mechanisms that are largely independent of wholesale power prices. Avoided costs that have been determined for APS and TEP can approximate the utilities marginal cost, although without the day-to-day specificity that LMPs provide in CAISO. The avoided costs of SRP are indicated by the utility s power purchase prices. Some of these rates vary by season and peak period. In this analysis we use the highest avoided cost for all three utilities, representing a best-case estimate of operating revenues. All avoided cost values for the utilities were between $20 per MWh and $30 per MWh. Figure 4 illustrates the spark spreads for a new NGCC facility in this market based on the utilities current avoided costs. The spark spread for all three Arizona utilities falls short of the revenue requirements for a new NGCC facility $/MWh Rev Req. (Dry Cool) Rev Req. (Wet Cool) 10 - (10) (20) (30) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct APS Spark Spread SRP Spark Spread TEP Spark Spread Figure 4: Spark Spreads in Arizona and California Future Mead Hub Wholesale Energy Prices To investigate whether a NGCC might become sufficiently profitable in the future, this analysis combined Energy Information Administration (EIA) gas price forecasts with a statistical analysis of the historical relationship between natural gas prices and wholesale power prices. Natural gas prices are often correlated with power prices, and the statistical analysis examined the systematic strength of this relationship specific to the Mead Hub during the period from January 1, 2013 to June 29, The objective was to identify a plausible range of future power prices within which to test the cost of a newly constructed NGCC. 6

7 The model tested historical Mead wholesale power prices derived from SNL Energy s Spot Power Index (peak and off-peak prices modeled separately) against natural gas prices and total load. 6 Both these parameters as well as the model overall demonstrated strong explanatory power at high levels of confidence (see Appendix for technical details). The resulting equations were: P!"#$,! = gas! peak MW! where P!""#$%&,! = gas! offpeak MW! P peak,t = predicted on-peak power price for year t P offpeak,t = predicted off-peak power price for year t gas t = EIA projected natural gas price for year t peak MW t = projected average peak-hour load for year t offpeak MW t = projected average off-peak-hour load for year t Many factors influence wholesale energy prices in addition to natural gas price and load. These can include temperature, power generation resource mix, power plant availability, and other changes in system operations. A more complex model for more precisely predicting future power prices would be necessary to incorporate the effect of these factors. The goal here is more modest: to establish a plausible range of future wholesale energy prices. The assumption here is that load and the price of natural gas exert the strongest influences on power prices and that the effects of other exogenous factors occur largely during extreme weather events or other system shocks that are outside the scope of the questions examined here. Two sets of EIA natural gas price forecasts were applied to these equations: one from EIA s reference case in its 2015 Annual Energy Outlook; and a High Oil and Gas Resource sensitivity case in which EIA tested high natural gas supply assumptions and their resulting low natural gas prices. 7 Projected load was based on analysis by the California Energy Commission (CEC), which expects energy demand to increase 1.23% per year from 2013 to 2025 across California. 8 This analysis applies the CEC forecast to hourly load (averaging peak and offpeak hours for each day) observed from , projecting these averages yearly through Figure 5 shows the plausible ranges for on-peak and off-peak energy prices through The model predicts that day-ahead on-peak wholesale energy prices at Mead Hub may increase from between $37/MWh and 6 Historical gas prices were collected from the SNL Daily Henry Hub Spot Natural Gas Index, which is a volume-weighted average based on completed Henry Hub contracts transacted for the next business day. On February 26, 2014, due to severe temperatures in the U.S., Henry Hub natural gas price spiked to $8.12 while Mead spot market power prices spiked to $106/MWh and $131/MWh respectively. This data point has been excluded from the dataset. Historical load data was collected from CAISO s day-ahead load forecast for both on and off-peak. 7 In this case, EIA predicts higher technological development and recovery per well, with closer well spacing than the reference case, resulting in Henry Hub natural gas spot prices 36% below the Reference case in 2020 and 44% below the Reference case in See: 8 See the Mid Energy Demand Scenario in Kavalec, C. California Energy Commission (2014). California Energy Demand Updated Forecast, The observed average used to conduct the prediction for on-peak load from was GW. The observed average employed for off-peak load from , was GW. 7

8 $42/MWh in 2015, to between $47/MWh and $65/MWh in Off-peak prices trend to between $31/MWh and $34/MWh in 2015, and to between $37/MWh and $51/MWh in $70 $/MWh $60 $50 $40 $30 On-peak Off-peak $20 $10 $ Figure 5: Mead Hub Wholesale Energy Price Projections A spark spread calculation based on these projected prices suggests that a new CCNG s net revenues would still fall largely below critical requirements of $24/MWh and $28/MWh for much of the projection period. 11 High gas prices offer the most revenue to a new NGCC in 2030 about $28/MWh during on-peak hours, declining to about $14/MWh during off-peak times. In this favorable case, a new wet cooled plant might meet minimum revenue requirements during on-peak times as early as 2023, whereas a dry cooled plant might be economic at on-peak times in Summary The persistence of low prices at the Mead Hub is likely to be a source of economic inertia affecting new NGCC capacity in the Southwest. This analysis suggests that the inertia could continue well past 2020, which could slow down new additions. These findings corroborate capacity expansion modeling results that suggest market conditions would not favor new gas-fired generation in the Southwest until around The approximate 95% prediction interval of the predicted on-peak price with the given parameters is +/- 9.5, and the interval for offpeak price is +/- 7.8 $/MWh. 11 It is expected that capital costs will remain constant through 2030, suggesting that critical revenue requirements will also remain constant. See: 12 See Trieu, M., Hale, E., Hurlbut, D., Clayton B., and Lopez, A. (2014). Future Capacity Expansion Scenarios of the Western Electricity System: Implications of Navajo Generating Station Retirements. NREL/OT-6A

9 Appendix The two-tailed P-value for both independent variables are <0.0001, indicating significance at 95% confidence interval (Table 1). The large natural gas price coefficient indicates a strong positive correlation between explanatory variable natural gas price and dependent variable, on-peak and off-peak day-ahead spot market price. Table 1. Relationship between natural gas price, day-ahead load, and day-ahead spot market price On-Peak Off-Peak Number of observations: Adjusted R Intercept Natural gas price GAS GAS Coefficient t statistic P value < * < * Day-ahead load ONLOAD OFFLOAD Coefficient t statistic P value < * < * *Significant at 95% confidence interval 9