Which Power Markets Are Most Vulnerable to Energy Storage and Why?

admin
Print Friendly, PDF & Email

______________________________________________________________________________

Eric Selmon Hugh Wynne

Office: +1-646-843-7200 Office: +1-917-999-8556

Email: eselmon@ssrllc.com Email: hwynne@ssrllc.com

SEE LAST PAGE OF THIS REPORT FOR IMPORTANT DISCLOSURES

______________________________________________________________________________

August 14, 2017

Which Power Markets Are Most Vulnerable to Energy Storage and Why?

In our note of March 22, 20% Price Declines for Wholesale Power? The Compelling Social Case for Electric Energy Storage and Why Regulated Utilities Are Likely to Roll It Out First, we pointed out that the impact of grid storage on market electricity revenues reflects two facts: the first is that power supply curves tend to be less steep at low levels of demand, allowing batteries to be charged with little impact on price, but quite steep at high levels of demand, allowing the discharge of batteries to have a significant impact on peak hour prices. The second is that power demand during the four highest price hours tends to be ~40% higher than demand during the four lowest price hours, implying that the reduction in peak hour prices brought by the discharge of batteries affects ~40% more MWh of electricity than the increase in off-peak prices brought about by their charging. We also found that small amounts of grid storage can materially reduce market electricity revenues. We believe the substantial savings to electricity consumers from the deployment of even limited amounts of storage will motivate state utility regulators to encourage investment in storage by regulated utilities; storage may also be attractive to municipal and co-operative utilities, and to electricity retailers.

Given the potential impact of storage on market electricity revenues, this note seeks to identify those characteristics of wholesale power markets that may render them particularly vulnerable to storage. Our analysis suggests that markets such as ERCOT, with a very high concentration of market electricity revenues in a limited number of hours, will be most vulnerable. These markets tend to have characteristics that drive power prices significantly higher during hours of peak demand, such as tight reserve margins, very steep supply curves or a tendency toward extreme hot or cold weather. Importantly, the impact of these market characteristics on power prices can be alleviated by the deployment of storage. In markets where reserve margins are tight, demand approaches the limit of available generation capacity during a higher proportion of hours than in markets where reserve margins are ample, forcing the dispatch of the least economic units on the system. The cost to start up and operate these units tends to result in critical peak pricing that is absent during the other hours of the year. In such markets, the discharge of storage during these highest demand hours adds to the supply of low cost electricity, eliminating the need to dispatch the most costly generating units on the system, and reducing critical peak prices during the highest demand hours of the year — and with them market electricity revenues.

How can our findings be used by investors to identify those markets whose electricity revenues are most at risk from the deployment of storage? While many of our quantitative analyses rely on data not easily available to investors, two readily available metrics can serve as rough indicators of the impact of storage on market electricity revenues. The first is the average price dispersion, or the gap between a market’s average peak and average off-peak power prices, expressed as a percentage of the average price of power, and the second is the market’s reserve margin. Looking ahead using forward power price curves and our forecasted reserve margins, ERCOT appears to continue to have the highest sensitivity to storage and it is increasing over time, while ISO New England’s sensitivity to storage appears to be decreasing over time.

Portfolio Manager’s Summary

  • We have modeled the impact on market electricity revenues[1] of deploying batteries dedicated solely to the arbitrage of peak and off-peak energy prices in each of the principal competitive power markets in the United States (CAISO, ERCOT, ISO New England, the New York ISO and PJM).
  • Our analysis suggests that by far the largest percentage decline in market electricity revenues from the deployment of storage would likely occur in ERCOT, where we estimate that the deployment of 500 MW of storage would have reduced market electricity revenues by 5.9% over 2012-2016.
  • By contrast, in CAISO, ISO New England, the New York ISO and PJM, we estimate the net reduction in market electricity revenues from the deployment of storage would have ranged between 1.4% and 2.3%, on average over 2012-2016. (See Exhibits 4 and 5.)
  • Regression analysis suggests that markets with a high concentration of market electricity revenues in a small number of the hours of the year, such as ERCOT, will feel a greater impact from the deployment of storage than those where market electricity revenues are more evenly spread across the hours of the year. (See Exhibits 7).
    • In ERCOT, an average of 16% of market electricity revenues over 2011-2016 were realized during the top 1% of hours; for PJM the comparable figure is 8% and for CAISO only 3%. (See Exhibit 6).
  • Markets with a high degree of revenue concentration are often characterized by a high degree of dispersion in power prices around the mean, with critical peak pricing occurring during more hours of each year than in other markets.
    • These features of vulnerable power markets in turn reflect characteristics such as tight reserve margins, steep power supply curves, and extreme hot or cold weather.
  • The sensitivity of market revenues to the deployment of storage depends upon:
    • the extent to which the decrease in peak hour prices, due to the discharge of storage, exceeds the increase in off-peak prices, when the batteries charge; and
    • the difference in the amount of electricity consumed during the highest price hours of the day relative to the amount consumed during the lowest price hours.
  • These factors are of different importance in identifying markets vulnerable to energy storage:
    • The variance in the gap between peak and off-peak load is limited across RTOs, and is thus of lower value in predicting differences across RTOs in the sensitivity of market revenues to storage. (See Exhibit 9.)
    • By contrast, RTOs vary widely in the extent to which the decrease in peak hour prices, due to the discharge of storage, exceeds the increase in off-peak prices, when the batteries charge. (See Exhibit 3.)
  • The power markets where the deployment of storage is likely to have the largest impact on market electricity revenues are those with two structural features: (i) tight reserve margins, contributing to a higher proportion of hours when demand approaches the limit of available generation capacity, and (ii) steep supply curves at high levels of demand, raising the price of power when these hours occur.
  • Temporary dislocations may also drive spikes in peak hour prices. These include: (i) extreme winter or summer weather that pushes demand to unusually high levels, and (ii) weather related constraints on the availability of generating capacity, such as poor hydrological conditions or inadequate natural gas transmission capacity to serve both heating and generating loads. Such dislocations tend to increase the impact of storage on market electricity revenues.
  • Storage is also more likely to achieve a material reduction in market electricity revenues in RTOs whose architecture and regulators favor a significant role for energy price volatility (such as ERCOT’s all-energy market) than in RTOs with capacity markets designed to mitigate energy price volatility (e.g., PJM).
  • Critically, the manifestations of these characteristics — frequent high power prices and the concentration of revenues in a limited number of hours as a result — can be mitigated by the discharge of storage during the highest price hours of each day.
  • To help investors monitor the vulnerability of markets to storage, two readily available metrics can serve as rough indicators of the impact of storage on market electricity revenues: the peak/off-peak price gap and the reserve margin.
    • A regression analysis of (i) the estimated decline in market electricity revenues due to the deployment of storage on (ii) the difference between a market’s average peak and average off-peak power prices, expressed as a percentage of the average price of power, results in an r-squared of 53%. (See Exhibit 15.)
    • Combined with a second readily available metric, the market’s reserve margin, in a two variable regression equation, the r-squared rises to 57%.
  • Looking ahead using forward power price curves and our forecasts of reserve margins, ERCOT should remain the most sensitive to storage deployment, while the New England ISO’s sensitivity should decrease. (See Exhibit 17)
  • An increasingly important factor in determining the sensitivity of market electricity revenues to the deployment of storage is the solar generation capacity on the system.
    • Solar generation is coincident with much of the daytime peak in power demand, and thus acts to suppress peak hour prices, flattening the supply curve and diminishing the potential impact of storage on prices and revenues. (See Exhibit 17.)
    • Solar power also tends to reduce the volume of demand prevailing when power prices are highest. In most U.S. power markets, peak power prices are coincident with afternoon peaks in demand. Yet in markets with substantial solar capacity, daytime power prices have tended to fall relative to prices when solar generation is not available. This can push the highest priced hours of the day from the afternoon to the early evening, when demand begins to taper off – again diminishing the potential impact of storage on market electricity revenues. (See Exhibits 18 through 21.)

Exhibit 1: Heat Map: Preferences Among Utilities, IPP and Clean Technology

Source: SSR analysis

Details

In our note of March 22, 20% Price Declines for Wholesale Power? The Compelling Social Case for Electric Energy Storage and Why Regulated Utilities Are Likely to Roll It Out First, we argued that relatively small amounts of electric energy storage on the bulk power system can significantly reduce peak power prices while only slightly increasing off-peak prices. The negative impact on market electricity revenues of much lower peak hour prices, and only slightly higher off-peak prices, is compounded by the fact that the volume of electricity sold during peak hours is some 40% higher than that sold off-peak. We believe the substantial savings to electricity consumers from the deployment of even limited amounts of storage will motivate state utility regulators to encourage investment in storage by regulated utilities; storage may also be attractive to municipal and co-operative utilities, and to electricity retailers.

Given the potential for storage to erode market electricity revenues, thus reducing both electricity costs for retail customers and the revenues of competitive generators, we focus in this note on identifying those characteristics of wholesale power markets that may render market electricity revenues particularly vulnerable to storage. We hope our analysis will help investors to anticipate company impacts as power markets change over time.

Methodology

To conduct our analysis, we have modeled the impact on market electricity revenues of deploying batteries dedicated solely to the arbitrage of peak and off-peak energy prices in each of the principal competitive power markets in the United States.[2] Allowing for 10% electricity losses through the charge/discharge cycle, we assumed that batteries deployed on the grid would charge for 4.4 hours each night during the hours when power prices are lowest, and discharge during the 4.0 hours each day when power prices were highest. To estimate how the daily charging and discharging of each 100 of MW battery capacity would affect peak and off-peak prices, we assumed that power prices would fall during the hours that batteries were being discharged to the average level prevailing when power demand was 100 MW lower, and that power prices would rise during the hours that the batteries were being charged to the average level prevailing when demand was 100 MW higher. (Given that power prices reflect numerous market factors in addition to demand, including the prevailing prices of coal and gas, seasonal swings in the availability of hydroelectric and other renewable resources, and the seasonal maintenance schedule of large thermal power stations, we estimated the price impact of 100 MW of storage on a month by month basis.)

Given the wide variation in size of the five RTOs studied, we have assumed the deployment of different amounts of storage in different markets. In each RTO, we have assessed the optimal amount of storage capacity in light of its impact on market revenues. For example, in simulating the impact of deploying battery storage in ERCOT over the last five years, we found that 500 MW of storage, in a market where average daily peak demand is ~44,000 MW, would reduce the average daily peak hour price of electricity by 37%. Increasing the amount of storage above 500 MW had very limited incremental impact; deploying 1000 MW of storage, for example, would have reduced the average daily peak hour price by just 39%.

Our analysis of the impact of storage on peak power prices in PJM produced a similar result. Over the last five years in PJM, by our estimate, the deployment of 500 MW of storage would have reduced the average daily peak hour price of electricity over by 8%; with 1000 MW of storage, the reduction in the average daily peak hour price increases to just 10%.

In the smaller ISOs, such as ISO New England and the New York ISO Zone J (New York City), our analysis suggests that 500 MW of storage is more than is required to achieve the bulk of the potential reduction in on-peak prices in these markets; rather, the deployment of as little as 100 MW of storage may be most cost effective.

The discussion that follows, therefore, is based on our estimate of the net impact on market electricity revenues of 500 MW of storage deployed in each of ERCOT and PJM, and 100 MW of storage deployed in CAISO, ISO New England and the New York ISO.

The Impact of Energy Storage on Power Prices and Market Electricity Revenues

A common feature of power supply curves across the five RTOs we examined is their tendency to slope steeply upward at very high levels of demand. (See Exhibit 2, which plots the supply curves of the two largest U.S. RTOs, PJM and ERCOT; in both these markets, the last 15% to 25% of generation capacity is materially more expensive to operate than the first 75% to 85%.) As a result, small increases in demand during the highest demand hours can result in disproportionate increases in the marginal cost of supply and thus in the prevailing market price of power. The deployment of relatively small amounts of storage reverses this dynamic; the discharge of storage during the hours of highest demand is reflected in a small reduction in generation but a large decrease in marginal cost and thus in the market price.

Conversely, during off-peak hours, when demand is low and low cost generation resources are abundant, power supply curves tend to be quite flat. As a result, it is possible to charge batteries during these hours with little upward pressure on off-peak power prices.

Exhibit 2: Power Supply Curves in ERCOT and PJM (Variable Cost of Generation in $/MWh Plotted Against % of Generation Resources)

ERCOT PJM

___________________

Source: SNL, SSR analysis

Grid storage thus tends to have a much greater impact on peak hour prices than on off peak power prices. The difference is illustrated in Exhibit 3, which presents the estimated impact of grid storage on peak and off-peak electricity prices by RTO over the period 2012-2016. The red columns represent the estimated decrease in average prices during the four highest price hours of each day, and the blue columns the estimated increase in average prices during the four lowest price hours of each day.

Exhibit 3: Change in Cost of Procuring Peak and Off-Peak Power to Serve

System Load Following Deployment of Battery Storage, per RTO ($/MWh) (1)

______________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis

Our next step was to examine the impact of storage on market electricity revenues. Here we found that the depressing effect of storage on peak power prices, and its far more limited upward pressure on off-peak prices, is accentuated by the fact that the volume of electricity sold during the highest demand hours of the day is on average some 40% greater than the volume sold during the lowest demand hours. Storage thus exerts significant downward pressure on market electricity revenues, depressing prices significantly during those hours when the highest volume of electricity is sold, while putting far less upward pressure on off-peak prices, when sales volumes are also materially lower.

In Exhibit 4, we quantify what these reductions in peak hour prices, and increases in off-peak prices, imply for annual market electricity revenues, and therefore generator revenues and consumer costs. The red columns illustrate the average reduction in market electricity revenues from the discharge of storage during the four highest demand hours of each day, and the blue columns show the average increase in market electricity revenues caused by the charging of storage during the four lowest demand hours of each day. Our analysis suggests that the roll out of very modest amounts of storage – ranging from 100 to 500 MW per RTO — could achieve significant reductions in market electricity revenues in all five RTOs. The largest dollar impact would be felt in the two largest RTOs, PJM and ERCOT: assuming the deployment of 500 MW of storage in each of these markets, we calculate that the net savings to consumers over the five years 2012-2016 would have been ~$590 million annually in ERCOT, while consumers in PJM would have saved ~$560 million annually.

A more useful way to compare the relative vulnerability of specific markets to the deployment of storage is to express the resulting erosion of market electricity revenues as a percentage of total market electricity revenues over 2012-2016. Our analysis suggests that by far the largest percentage decline in market electricity revenues from the deployment of storage would likely occur in ERCOT, where

we estimate that the deployment of 500 MW of storage would have reduced net market electricity revenues by 5.9% over 2012-2016. By contrast, in the other four RTOs, we estimate the net reduction in market electricity revenues from the deployment of storage would have ranged between 1.4% and 2.3% of total market electricity revenues, over 2012-2016. (See Exhibit 5.)

Exhibit 4: Change in Cost of Procuring Peak and Off-Peak Power to Serve

System Load Following Deployment of Battery Storage, per RTO ($ Millions)

______________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis

Exhibit 5: Estimated Net Reduction in Market Electricity Revenues from the Deployment of Storage, Expressed as a % of Total Market Electricity Revenues (2012-2016)

______________________

Source: SNL, SSR analysis

Why Are Certain Power Market Particularly Vulnerable to Storage?

Why are some power markets materially more sensitive to the deployment of storage than others? Why should ERCOT, for example, suffer a percentage decline in market electricity revenues from the deployment of storage that is four times as large as that expected in ISO New England?

Our analysis suggests that markets with a high concentration of market electricity revenues in a small number of the hours of the year tend to feel a greater revenue impact from the deployment of storage than markets where electricity revenues are more evenly spread across the hours of the year. Markets with a high degree of revenue concentration are often characterized by a high degree of dispersion in power prices around the mean, with critical peak pricing occurring during more hours of each year than in other markets. The discharge of storage during the highest price hours of each day can mitigate these price spikes, and thus decrease market electricity revenues.

At a more granular level, markets whose revenues are vulnerable to storage tend to have characteristics that drive power prices significantly higher during hours of peak demand, such as tight reserve margins, very steep supply curves at high levels of demand, or a tendency toward extreme hot or cold weather. In markets where reserve margins are tight, demand approaches the limit of available generation capacity during a higher proportion of hours than in markets where reserve margins are ample, forcing the dispatch of the least economic units on the system. The cost to start up and operate these units tends to result in critical peak pricing that is absent during the other hours of the year. In such markets, the discharge of storage during these highest demand hours adds to the supply of low cost electricity, eliminating the need to dispatch the most costly generating units on the system, and reducing critical peak prices during the highest demand hours of the year — and with them market electricity revenues.

Conversely, RTOs where market electricity revenues are smoothly distributed over the hours of the year tend to be characterized by more ample reserve margins, lower cost peaking plants and thus flatter supply curves at high levels of demand. These conditions are more typical, say, of CAISO, which also enjoys the benefit of more moderate weather. In markets such as these, scarcity conditions will occur less frequently, and the deployment of storage therefore has a lesser impact on peak hour prices and market electricity revenues.

In Exhibits 6 through 8, we present the key data on which our conclusions are based. Exhibit 6 provides a comparison of the degree of revenue concentration across the five RTOs studied. In ERCOT, an average of 16% of market electricity revenues over 2010-2016 were realized during the top 1% of hours; for PJM the comparable figure is 8%, and for CAISO only 3%.

Exhibit 6: Share of Total Market Electricity Revenues Received During Highest Revenue Hours, 2010-2016

__________________

Source: SNL, SSR analysis

In Exhibit 7, we examine the relationship between revenue concentration and the sensitivity of market electricity revenues to the deployment of storage. Specifically, the exhibit presents the results of a regression analysis of (i) the estimated change in market electricity revenues from the deployment of storage across the five RTOs over the seven years 2010-2016 on (ii) the percentage of market electricity revenues produced in the top 1% of hours over the same period.

Exhibit 7: Estimated % Decline in Market Electricity Revenues from the Deployment of Grid Storage vs. % of Total Market Electricity Revenues Generated in Top 1% of Hours (1)

All Five RTOs, 2010-2016 Excluding ERCOT in 2011

__________________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis.

When taking as our sample all five RTOs across the years 2010-2016, the regression demonstrates a statistically significant, positive correlation between revenue concentration and the impact of storage on market electricity revenue, with an r-squared of 74% (see the left hand chart of Exhibit 7). The remarkably high r-squared of this regression, however, reflects the impact of an outlier, ERCOT in 2011, when the top 1% of hours accounted for 36% of revenues; in ERCOT in 2011, we estimate, the deployment of 500 MW of storage would have reduced market revenues by 16%. Excluding ERCOT 2011, we continue to see a positive, statistically significant correlation between the concentration of market revenues and the percentage decline in market revenues from the deployment of storage, but the r-squared falls to 29% (see the right hand chart in Exhibit 7). [3]

In Exhibit 8, we examine the relationship between revenue concentration and the dispersion of power prices around the mean. Specifically, the exhibit presents the results of a regression of (i) the percentage of market electricity revenues produced in the top 1% of hours on (ii) the difference between average peak and average off-peak power prices, divided by the average or around-the-clock price of power.

Exhibit 8: The % of Total Market Electricity Revenues Generated in Top 1% of Hours vs. the Dispersion of Average Peak and Off-Peak Power Prices Around the Mean (1)

__________________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis.

When taking as our sample all five RTOs across the years 2010-2016, the regression demonstrates a statistically significant, positive correlation between revenue concentration and the dispersion of average peak and off-peak power prices around the mean, with an r-squared of 71% (see the left hand chart of Exhibit 8). Again, however, the remarkably high r-squared of this regression reflects the impact of an outlier, ERCOT in 2011. Excluding ERCOT 2011, we continue to see a positive, statistically significant correlation between the concentration of market revenues and the dispersion of average peak and off-peak power prices around the mean, with an r-squared of 48% (see the right hand chart in Exhibit 8). 3

Which Characteristics Render Power Markets More Vulnerable to the Impact of Storage?

What lies behind the correlation between the negative impact of storage on market revenues and the degree of price dispersion and revenue concentration in those markets? Recall that in estimating the impact of storage on market revenues we assume that the storage capacity in each RTO would charge during the four hours of the day when power prices are lowest (generally during the small hours of the morning when demand is at its minimum) and discharge during the four hours when prices are highest (generally in the afternoon or evening when demand reaches its peak). The sensitivity of market revenues to the deployment of storage, therefore, will depend on the relative change in prices during peak and off-peak hours, or the extent to which the decrease in peak hour prices, due to discharge of storage, exceeds the increase in off-peak prices, when the batteries charge.

A second key factor in determining the impact of storage on market electricity revenues will be the difference in the amount of electricity consumed during the highest price hours of the day relative to the amount consumed during the lowest price hours. In markets where there is a much larger volume of electricity consumed on peak than off-peak, the changes in peak and off peak prices caused by the discharge and charging of storage will have a much greater effect.

While the gap between peak and off-peak load is a significant factor in determining the impact of storage on market electricity revenues, variance of the gap is limited across RTOs; it therefore is of limited value in predicting differences across RTOs in the sensitivity of market revenues to the deployment of storage. Across the five RTOs studied, the difference between the consumption of electricity during the four highest price hours and four lowest price hours of the year is surprisingly uniform; it averages 1.4x, and varies between 1.38x in PJM and 1.44x in the New York ISO (see Exhibit 9).

Exhibit 9: Ratio of Load During the 4 Highest Priced vs.

the 4 Lowest Priced Hours per Day

__________________

Source: SNL, SSR analysis

By contrast, as illustrated in Exhibit 10, the five RTOs show wide variations in the extent to which (i) the reduction in peak hour prices due to discharge of storage exceeds (ii) the increase in off-peak prices when batteries are charged. In ERCOT, the deployment of storage would, by our estimate, reduce peak hour prices by ~$11.42/MWh while increasing off-peak prices by ~$3.36/MWh, implying a net savings $8.06/MW. In CAISO, by contrast, we estimate that the deployment of storage would reduce peak hour prices by ~$4.82/MWh, while raising off-peak prices by ~$2.09/MWh, for a net savings of only $2.73/MWh – only a third of the net benefit estimated for ERCOT.

Exhibit 10: Change in Cost of Procuring Peak and Off-Peak Power to Serve

System Load Following Deployment of Battery Storage, per RTO ($/MWh) (1)

______________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis

The much greater importance of net price changes than peak to off-peak demand differentials in determining the impact of storage on RTO market revenues is illustrated in Exhibit 11. As can be seen there, net price changes explain 87% of the difference in market revenues losses across the five RTOs over 2010-2016 (see the left hand chart in Exhibit 11). If ERCOT 2011 is excluded, net price changes still explain 66% of the net revenues losses (see the right hand chart in Exhibit 11).

Importantly, the net price impact from the deployment of storage differs across the five RTOs primarily due to differences in the impact of storage on peak hour prices. The charging of batteries during off-peak hours is estimated to raise off-peak prices by zero to ~$3.36/MWh, depending on the RTO (see Exhibit 10); the impact of discharging storage on peak hour prices, by contrast, varies across a far wider range, from ~$3.51 to ~$11.42/MWh. Presumably because of the much wider range of changes in peak than in off-peak prices, the change in peak price from the deployment of storage is an even better predictor of the change in market revenue than the change in net prices. Across all the RTOs and years, peak price changes explain 91% of the change in market revenue from the deployment of storage; if ERCOT 2011 is excluded, peak price changes explain 77% of the change in market revenue (see Exhibit 12). A key factor in determining the vulnerability of market revenues to the deployment of storage, in other words, is differences in the sensitivity of peak hour prices to the discharge of storage capacity.

Exhibit 11: Estimated % Decline in Market Electricity Revenues from the Deployment of Grid Storage vs. Net Change in Peak and Off-peak Prices ($/MWh) (1)

All Five RTOs, 2010-2016 Excluding ERCOT in 2011

__________________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis

Exhibit 12: Estimated % Decline in Market Electricity Revenues from the Deployment of Grid Storage vs. Change in Peak Prices ($/MWh) (1)

All Five RTOs, 2010-2016 Excluding ERCOT in 2011

__________________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis

Why might the deployment of storage produce a larger change in on-peak power prices in some RTOs than in others? Several factors contribute to this difference, albeit in different ways, some tending to increase the sensitivity of peak hour prices to storage and some tending to diminish it.

First, as noted above, in markets where reserve margins are tight, demand approaches the limit of available generation capacity during a relatively higher proportion of hours than in markets where reserve margins are ample, forcing the dispatch of the least economic units on the system. The cost to start up and operate these units tends to result in critical peak pricing that is absent during the other hours of the year. In such markets, the discharge of storage increases the supply of electricity during these highest demand hours, eliminating the need to dispatch the most costly generating units

on the system, and reducing critical peak prices — and with them market electricity revenues. As illustrated in Exhibit 13, a regression analysis of all five RTOs over the years 2010-2016 suggests that differences in reserve margins can explain 18% of the variation in estimated revenue losses (see r-squared in the chart on the left), rising to 24% if ERCOT 2011 is excluded (see chart on the right).

Exhibit 13: Estimated % Decline in Market Electricity Revenues from the Deployment of Grid Storage vs. NERC Foremast of Summer Reserve Margin ($/MWh) (1)

All Five RTOs, 2010-2016 Excluding ERCOT in 2011

__________________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis

A second important factor in determining the sensitivity of market electricity revenue to the deployment of storage is the severity of the weather in the RTO. It is important to recall that reserve margins are estimated ex ante by independent system operators to assess the capability of a control area’s generation resources to meet projected levels of demand, with a particular focus on ensuring adequate levels of reliability during summer and winter peaks. In these forecasts, projected demand is based on the assumption of normal weather. Even in markets with apparently adequate reserve margins, therefore, higher than normal levels of demand, reflecting severe weather, can cause load to approach the limits of available generating capacity, forcing the dispatch of high cost peaking plants.

In the early months of 2011, for example, the reserve margin forecast by ERCOT for the coming summer was 14.3%, assuming normal weather, which was above the minimum reserve margin of 13.75% targeted by ERCOT at the time. In fact, however, the summer of 2011 was characterized by prolonged periods of extremely hot weather, resulting in levels of air conditioning load that placed ongoing pressure on ERCOT’s generating resources. During the 31 days of August 2011, the average power price prevailing during the four highest price hours of each day was $153/MWh; by contrast, over the next five years, the average power price prevailing during these hours in the month of August was $38/MWh, 25% of the 2011 level. The severity of the weather in ERCOT in 2011 causes this case to stand out as an outlier in the left hand chart of Exhibit 13.

Weather related surges in demand can be particularly difficult to manage if they occur in the shoulder months of the spring or fall, when cooler weather usually prevails and power demand is commensurately low. Many coal fired and nuclear power plants schedule outages during these months for major maintenance or refueling. Unseasonably hot weather coincident with these outages can place a severe strain on the remaining capacity on the system.

Third, extreme weather can limit the availability of system resources. The California energy crisis of 2000-2001 was precipitated in part by a drought in the Pacific Northwest that depressed the hydroelectric output of dams in Washington and Oregon, whose excess generation under normal hydrological conditions contributes materially to California’s supply of power. Similarly, during the winter of 2014, rendered bitterly cold by a polar vortex, the supply of natural gas in the Midwest and Northeast was diverted away from gas fired power plants to serve heating load, significantly curtailing the amount of gas fired generation capacity available in these regions. In both cases, the unavailability of some of the lowest cost resources on the system required the operation of high cost, oil fired steam turbine generators to meet peak demand.

Fourth, the slope of the power supply curve at high levels of demand can exacerbate or alleviate price pressures during periods of capacity scarcity. As noted above, during the winter of 2014 severely cold weather constrained the delivery of natural gas to gas fired power plant in the Midwest and Northeast, forcing the dispatch of higher cost oil fired steam turbine generators. Given the relative prices of natural gas and fuel oil that winter, the transition from gas to coal fired generation was extremely costly. During Q1 2014, wholesale fuel oil prices in the Midwest averaged ~$135 per barrel, or ~$23.50/MMBtu, more than 3x the price per Btu of natural gas, which averaged ~$7.25/MMBtu over the quarter. By contrast, relatively flat power supply curves at high levels of demand can have the opposite effect. In CAISO, for example, utility scale solar power plants now comprise 16% of all the generation capacity on the system. The high levels of output from these units during the summer months tends to lengthen and flatten the power supply curve during some of the highest demand hours of the year. The importance of California’s solar resource in alleviating price pressures on peak is illustrated in Exhibit 14: compare the blue and yellow supply curves, which exclude and include, respectively, CAISO’s utility scale solar generation capacity.

Exhibit 14: CAISO Power Supply Curve, with and without Utility Scale Solar Generation Capacity (Variable Cost, in $/MWh, Plotted Against % of Generation Resources)

_____________________

Source: SNL, SSR analysis

In summary, the power markets where the deployment of storage is likely to have the largest impact on market electricity revenues tend to be characterized by certain structural features, including (i) tight reserve margins, contributing to a higher proportion of hours when demand approaches the limit of available generation capacity and (ii) steep supply curves at high levels of demand. However, markets may also be vulnerable to temporary dislocations that cause intermittent spikes in peak hour prices. Drivers of such dislocations include (i) extreme winter or summer weather that pushes demand to unusually high levels, straining system resources, and (ii) weather related constraints on the availability of generating capacity, such as poor hydrological conditions or inadequate natural gas transmission capacity to serve heating and generating loads. Whatever their cause, the resulting price spikes allow the discharge of storage during the highest price hours of the day to have a material effect on market electricity revenues.

Conversely, structural factors tending to diminish the efficacy of storage in reducing market electricity revenues include ample reserve margins and flat supply curves at high levels of demand. The latter could reflect relatively homogenous prices, on a per Btu basis, for fossil fuels such as coal, gas and fuel oil, or ample supplies of solar generation capacity. Factors tending to limit the frequency of the temporary dislocations described above include moderate weather and ample generation capacity with on-site supplies of fuel (e.g., nuclear, coal and oil-fired power plants) or other forms of energy (e.g., wind or solar generating capacity). In markets such as these, price spikes are less frequent and the ability of storage to reduce electricity revenues commensurately less.

Finally, we believe that the regulatory framework in which a wholesale power market operates may exert a considerable influence on the sensitivity of market electricity revenues to the deployment of storage. RTOs that have chosen to rely on an energy-only model, such as ERCOT, and which therefore lack well established capacity markets, require prolonged periods of elevated energy prices to incent the construction of new generation capacity. By contrast, RTOs that have embraced instead the capacity market model, and implemented markets that produce significant capacity revenues to complement power plants’ generation margins, have done so in large part to minimize the volatility of energy prices and ensure a steadier and more predictable stream of revenues for generators. Storage is thus more likely to achieve a material reduction in market electricity revenues in RTOs whose architecture and regulators favor a significant role for energy price volatility.

Identifying Markets at Risk from Storage

How can our findings be used by investors to identify those markets whose electricity revenues are most at risk from the deployment of storage? Many of the quantitative analyses we have conducted rely on data not easily available to investors. Are there other metrics that are more readily available that might serve as rough indicators of the impact of storage on market electricity revenues?

At the beginning of our analysis of which markets are most vulnerable to storage we noted that:

  • markets with a high concentration of market electricity revenues in a small number of the hours of the year tend to suffer a greater decline in market electricity revenues from the deployment of storage than markets where electricity revenues are more evenly spread across the hours of the year; and
  • markets with a high degree of revenue concentration are often characterized by a high dispersion in power prices around the mean, with critical peak pricing occurring during more hours of each year than in other markets.

In Exhibit 8, we demonstrated the strong statistical relationship between power price dispersion and revenue concentration. Across all five RTOs over the years 2010-2016, regression analysis demonstrated a statistically significant, positive correlation between revenue concentration and the dispersion of average peak and off-peak power prices around the mean, with an r-squared of 71%. Excluding the outlier of ERCOT in 2011, the r-squared falls to 48%.

Given that markets with a high degree of revenue concentration are more likely to suffer a decline in market electricity revenues upon the deployment of storage, perhaps the dispersion of prices in power markets might serve as useful practical indicator of the vulnerability of power markets storage. While revenue concentration in power markets is not a readily available metric, the data required to estimate price dispersion – the around-the-clock price of power, and the average price of power during peak and off-peak hours — are widely available from sources such as SNL.

Exhibit 15 presents the results of a regression analysis of (i) the estimated change in market electricity revenues from the deployment of storage, on (ii) the difference between average peak and off-peak power prices, divided by the average or around-the-clock price of power, or average price dispersion. The regression demonstrates a statistically significant positive correlation between the gap in average peak and off-peak prices, expressed as a percentage of the average power price, and the estimated decline in market electricity revenues as a result of storage. Across the five RTOs over the seven years 2010-2016, the r-squared of the regression is 53% (see the left hand chart in Exhibit 15); if we exclude the outlier observation of ERCOT in 2011, the r-squared falls to 18% (see the chart on the right side of Exhibit 15).

Exhibit 15: Estimated % Decline in Market Electricity Revenues from the Deployment of Storage vs. the Gap Between Average Peak and Off-Peak Prices as % of Average Price (1)

All Five RTOs, 2010-2016 Excluding ERCOT in 2011

__________________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis

A second, widely available metric that can be used to assess the vulnerability of market electricity revenues to the deployment of storage is the reserve margin. As illustrated in Exhibit 16, a regression analysis of all five RTOs over the years 2010-2016 suggests that differences in reserve margins can explain 18% of the variation in estimated revenue losses (see r-squared in the chart on the left), rising to 24% if ERCOT 2011 is excluded (see chart on the right).

Exhibit 16: Estimated % Decline in Market Electricity Revenues from the Deployment of Grid Storage vs. NERC Foremast of Summer Reserve Margin ($/MWh) (1)

All Five RTOs, 2010-2016 Excluding ERCOT in 2011

__________________________

1. In CAISO, ISO New England and the NY ISO, we have modeled the impact of 100 MW of grid storage, while in the much larger ERCOT and PJM markets we have modeled the impact of deploying 500 MW of storage.

Source: SNL, SSR analysis

Combining the two metrics, price dispersion and reserve margin, in a two variable regression explains more of the estimated loss in market electricity revenues than either variable alone. Across all RTOs and all years, the two metrics together explain 57% of the variation in estimated market revenue losses due to the deployment of storage, as against 54% for price dispersion alone. If we exclude the outlier of ERCOT in 2011, the r-squared of the two variable equation is 33%, up from 18% for price dispersion alone.

Using these indicators, which markets will be most sensitive going forward? In Exhibit 17 we present the forward price dispersion, or the gap between the forward on-peak and off-peak prices divided by the forward around-the-clock price, and our forecast of the reserve margins for each market. On both measures ERCOT will remain the most sensitive to storage and that sensitivity should increase. While CAISO’s reserve margin is declining, probably reflecting the conservative assumptions regarding new capacity additions in our forecast post-2019, its forward price dispersion remains low. We therefore expect CAISO to remain one of the RTOs that is least sensitive to storage. Interestingly, the New England ISO should become less sensitive to storage over time, as price dispersion declines and reserve margins increase.

Exhibit 17: Forward Average Price Dispersion and Forecasted Reserve Margins

Forward Price Dispersion, 2018-23 Forecast Reserve Margins, 2018-22

 

__________________________

Source: SNL, SSR analysis

Does Solar Power Undermine the Effectiveness of Grid Storage?

In attempting to explain why storage should have a more limited impact on market electricity revenues in an RTO such as CAISO, we have noted that the output of CAISO’s abundant solar generation capacity is largely coincident with the hours of peak power demand. By extending and flattening the supply curve during these hours, solar acts to suppress the volatility of on-peak power prices. (Compare the yellow and blue CAISO supply curves, which exclude and include solar generation, respectively, in Exhibit 18).

We have also found evidence, however, that solar power acts to undermine storage economics in a less obvious way: by shifting the highest price hours to periods of lower demand volume. In most North American power markets, peak power prices are coincident with afternoon peaks in demand. In markets like CAISO, however, with its substantial solar capacity, daytime power prices have tended to fall relative to prices when the solar resource is less abundant. This can have the effect of pushing the highest priced hours of the day from the afternoon hours to the early evening, when demand begins to taper off. As solar generation capacity expands over time, the implication could be that the highest priced hours of the day no longer coincide with the hours of peak demand. Since the impact of grid storage on market revenues depends on both the difference between peak and off-peak power prices and the difference between peak and off-peak power demand, this trend will erode the economics of energy storage.

Exhibt 18: CAISO Power Supply Curve, with and without Wind and Solar Capacity, Showing Average Demand During 4 Lowest Price and 4 Highest Price Hours Each Day

________________________________

Source: SNL, SSR estimates and analysis

There is considerable evidence that the shift of high power prices to lower demand hours is occurring already. Exhibit 19 plots the ratio of the demand prevailing during the four highest price hours of the day to the demand prevailing during the four lowest price hours of the day. This has declined over the years from 2010 through 2016 in each of the five RTOs studied. As illustrated in Exhibit 20, the decline in this ratio has been most marked in those RTOs that have added the largest amount of the solar capacity. This is particularly evident in CAISO, where the ratio of demand during high price hours to demand during low price hours has fallen by 16% from 2011 to 2016, even as the share of utility scale solar capacity to the total resources on the grid has increased by 9.1 percentage points (see Exhibit 20 and Exhibit 21).

Exhibit 19: Ratio of Load During the 4 Highest Exhibit 20: Change in Ratio of Load During the 4 Highest

Priced & 4 Lowest Priced Hours per Day, 2011-16 Priced vs. the 4 Lowest Priced Hours (%) vs. Change in

Solar Share of Total Generation (Percentage Points), 2011-16

_____________________ _____________________

Source: SNL, SSR analysis Source: SNL, SSR analysis

Exhibit 21: CAISO — Ratio of Load During 4 Highest Priced Hours to 4 Lowest Priced Hours vs. Solar as % of Total Generation

______________________

Source: SNL, SSR analysis

To test the hypothesis that the growth of utility scale solar capacity has compressed the gap in demand between the highest and lowest price hours of the day, we ran a regression analysis of (i) the ratio of power demand during the four highest and four lowest price hours of the day across the five RTOs studied (using the average ratio for each month over the years 2010 through 2016) against (ii) the percentage contribution of utility scale solar capacity to total generation resources on the grid. As can be seen in in Exhibit 21, this regression analysis demonstrates a marked negative relationship between the ratio of demand during the highest and lowest price hours with the share of utility scale solar in total generation capacity. The equation has an r-squared of 29%, with statistically significant t-statistics for both the coefficient and the intercept.

Exhibit 22: Regression Analysis of Ratio of Load During 4 Highest Priced Hours to 4 Lowest Priced Hours vs. Solar as % of Total Generation, 2010-2016

______________________

Source: SNL, SSR analysis

There are other of course other reasons to deploy storage even in the face of rising solar capacity. Storage is often used, for example, to defer the sale of solar generation to higher price hours in the evening. Storage can also be used to provide spinning reserves that deploy more rapidly than conventional generation in response to increases and decreases in solar output over the course of the day. Finally, storage can mitigate the need for major capital improvements on circuits where distributed solar generation is deployed. The growing deployment of solar generation, however, will likely diminish the benefits of storage in mitigating price spikes and thus reducing market electricity revenues.

Appendix 1: Data Points Used in the Regression Analysis of Estimated Net Impact of Grid Storage on Market electricity revenues ($ Millions) vs. (i) the Difference in Peak and Off-Peak Prices ($/MWh) and (ii) the Difference in Peak and Off-Peak Power Demand (MW)

Exhibit 23: Estimated Net Impact of Grid Storage on Market Electricity Revenues ($ Millions) (1)

____________________

1. Assumes 100 MW of grid storage in CAISO, ISO New England and New York ISO and 500 MW of grid storage in ERCOT and PJM

Source: SNL, SSR analysis

Exhibit 24: Average Difference in Price Between 4 Highest Priced & 4 Lowest Priced Hours Each Day ($/MWh) (1)

____________________

1. Assumes 100 MW of grid storage in CAISO, ISO New England and New York ISO and 500 MW of grid storage in ERCOT and PJM

Source: SNL, SSR analysis

Exhibit 25: Average Difference in Load Between 4 Highest Priced & 4 Lowest Priced Hours Each Day (Millions of MWh) (1)

____________________

1. Assumes 100 MW of grid storage in CAISO, ISO New England and New York ISO and 500 MW of grid storage in ERCOT and PJM

Source: SNL, SSR analysis

©2017, SSR LLC, 225 High Ridge Road, Stamford, CT 06905. All rights reserved. The information contained in this report has been obtained from sources believed to be reliable, and its accuracy and completeness is not guaranteed. No representation or warranty, express or implied, is made as to the fairness, accuracy, completeness or correctness of the information and opinions contained herein.  The views and other information provided are subject to change without notice.  This report is issued without regard to the specific investment objectives, financial situation or particular needs of any specific recipient and is not construed as a solicitation or an offer to buy or sell any securities or related financial instruments. Past performance is not necessarily a guide to future results.

  1. We use the term “market electricity revenue” to describe the theoretical cost of all power sold in a competitive wholesale power market if it were all priced at the hourly market price of power, ignoring the impacts of hedges and long-term contracts. Market electricity revenue should reflect the long-run cost of electric energy in the market in future years as contracts expire and hedges roll off, and new hedges and contracts are repriced reflecting expectations that incorporate the new reality of storage. Market electricity revenue also represents the long-run energy revenues available to generators and costs to consumers under the same conditions.
  2. California Independent System Operator (CAISO), Electric Reliability Council of Texas (ERCOT), ISO New England (ISO NE), New York Independent System Operator (NY ISO), and PJM Interconnection (PJM).
  3. In comparing these two results, we believe it is important to bear in mind that one of the benefits of storage could well be to avoid the spikes in peak prices and market revenues that result from extreme weather events, such as the summer of 2011 in Texas, when Dallas suffered 100 days of over 100 degree heat.
Print Friendly, PDF & Email