PHOTON International The Photovoltaic Magazine

Thought experiment – how PV reduces wholesale power prices in New England

A thought experiment using detailed historical data shows that if 1 GW of PV had been installed in the New England states in 2002, average wholesale electricity prices would have been reduced by 2 to 5 percent.

	© Massachusetts Technology Collaborative (MTC)
Modeling: With only 500,000 residential PV rooftop systems each rated at 2 kW – equal to around 3 percent of New England's power capacity – between $129 and $281 million in power expenditures could have been saved in the six US states.

The biggest obstacle to large-scale adoption of PV is its high price. That's nothing new for people who follow the solar power sector. It's also common knowledge that the zenith of PV's electricity production curve – middle of the day – matches peak power demand in industrialized countries. A less-common discussion is the impact solar power could have on electricity prices if PV is adopted on a larger scale.

Consider solar power in the New England states. Some of the major US PV manufacturers (RWE Schott Solar, Evergreen Solar) and promising start ups such as Konarka are located in Massachusetts. But, the six US East-Coast states (Connecticut, Massachusetts, Maine, New Hampshire, Rhode Island, and Vermont) with a total population of around 14 million have only installed around 10 MW of PV capacity compared to 32 GW total capacity mostly from conventional plants. The New England states have a range of PV-incentive programs – tax incentives, subsidies, and renewable energy certificates – but all programs are relatively small, especially when compared to those in Japan and Germany. However, this analysis suggests that large-scale PV deployment could create considerable cost savings in New England.

In New England, at the Massachusetts Institute of Technology's Laboratory for Energy and the Environment (LFEE), historical data on the US power system is being used to create simple but detailed models of the effects of regulations and renewable generation on regional power systems. Over the last few years, one model of the regional US power markets was developed that includes electricity generation and load, emissions from fossil generation, and PV generation. Originally this model contained hourly historical unit-by-unit data on power system performance for every generating unit in the US above 25 MW. The model now includes some market-clearing price data, which provides a means to estimate the impact of PV generation on prices. In thinking through the question – »What could solar power's impact be on electricity prices?« – the model was applied to New England.

Solar power reduces daytime load

Because typical PV installations are small, distributed residential rooftop systems, a 2 kW DC system was used as the basis for this thought experiment. In New England, such a PV system generates power between 9 am and 6 pm in the winter, and 6 am and 9 pm in the summer, with peak output in the early afternoon (see graph, p. 40). It was assumed that the system does not break, although cloud cover does affect generation, which is about 1.9 MWh during a year (2.4 MWh DC – high conversion losses were assumed to ensure a conservative estimate). This yield is based on a PV system simulated by MIT LFEE using hourly solar resource data for a site in New England in 2002.

In this thought experiment PV capacity is scaled up to 1 GW, which would be like installing 500,000 small rooftop PV systems – a fairly significant penetration in an area with 14 million people. It is assumed that the larger penetration of PV does not impact the transmission or distribution of power. The generation from these PV sites would be about 940 GWh or about 0.7 percent of New England's annual generation (127 TWh in 2002).

Solar power reduces market-clearing price

The system of economic dispatch in competitive electric power systems dictates that, in general, lower demand means a lower market-clearing price. A typical hour's supply curve is shown in the graphic on page 39. The slope of the supply curve increases as demand increases (right side of supply curve). The system dispatches the most expensive generator last and this generator sets the clearing price for all generation. Conversely, the baseload plants (often nuclear or coal) that make up the left side of the supply curve are »price takers« that capture the market-clearing price regardless of their bid.

When demand is high and the supply curve steep, a small decrease in demand can mean a considerable reduction in clearing price. For example, a demand reduction of about 1 GWh in an hour from a demand of 25 GWh in the hour results in a drop in clearing price from over $900 per MWh to about $600 per MWh (see graph, p. 39).

© source: Kate Martin; graphic: PHOTON International

This relationship does not always hold, due to out-of-merit generation induced by transmission constraints, contingencies, or the provision of ancillary services like reserve capacity. For this thought-experiment, however, it was assumed that the relationship holds.

Solar power shaves price spikes

Price spikes occur when demand is abnormally high and the system accepts the high bids from expensive generating units on the right side of the supply curve. Such a situation took place in a series of hours between Aug. 11-19, 2002 in New England. While there were a number of price spikes in that week, on Aug. 14 at 3 pm, a very small increase in load compared to the day before let the price go through the roof – up to $1,000 per MWh. If 1 GW of additional PV were installed at that time, the model for this thought experiment estimates that the price would have been reduced to about $480 per MWh due to the lower demand in the hour. This is a savings of 52 percent in that one hour, or about $13 million.

Why such big savings? First, in the hours preceding the spike, the demand and clearing price are smaller. Thus it is logical to assume that if the demand had not increased in an hour, the clearing price would be the same as the previous hour's clearing price. Similarly, if the demand did not increase for two hours, the clearing price would be that of two hours previous. Second, because during spike hours the supply curve is steep, a comparatively small reduction in load can result in a large reduction in price.

© source: Kate Martin

A simulation of PV power generation in New England in 2002. While the black on the left and right side of the graph shows nighttime, the dark horizontal streaks during the day are from cloud cover.

New England saw 36 price spikes above $100 per MWh in August and July 2002 (see graphs, p. 40). This may seem like a lot – and indeed 2002 was a particularly warm summer, nevertheless only 0.6 percent of the total hours for New England in 2002 were price spikes above $100 per MWh. This analysis suggests that PV's ability to peak shave has a significant impact even if applied for this small percentage of hours.

The results – significant price impact

In addition to 1 GW of installed PV capacity – as assumed above – smaller amounts (50 MW, 500 MW) and large amounts (2 GW) were also added to the ISO-New England system for 2002 (see table & graph, p. 41). In the model, for each hour it is assumed that the new demand is the actual historical demand minus the PV generation. A simple model assigns a new clearing price to each hour if the PV generation in the hour is enough to reduce it below a previous hour's demand. The model extrapolates the new clearing price from the previous hours' clearing prices according to the new demand.

In a »conservative estimate« this calculation was performed for only the hours in 2002 in which the market-clearing price was above $100 per MWh. The result suggests that the savings from 1 GW of PV in New England would be $129 million annually. While the average wholesale price in New England for 2002 was 3.78¢ per kWh, 1GW of PV would have reduced the average wholesale price to 3.71¢ per kWh, a 2 percent savings.

Adding more PV generation shaves more price spikes and the savings increase accordingly, but even a small amount of PV can have significant impact: in the »conservative estimate« scenario 50 MW PV (0.15 percent of total capacity) would save $25 million or 0.5 percent in the power market, increasing to savings of $199 million or 4.1 percent for a 2 GW PV capacity (6 percent of total capacity).

While these savings are already impressive, they increase dramatically if PV's impact is not only limited to price peaks above $100 per MWh. In fact, large amounts of PV generation will influence the clearing price in all hours. If this »moderate estimate« is modeled, 1 GW of PV creates $281 million in savings (6 percent). This equates to an average wholesale price of 3.59¢ per kWh (5 percent reduction). For 2 GW of PV power, expenditure would decrease by $381 million (8 percent).

Conclusions

Although this thought experiment is not a rigorous analysis, it seems there are at least three interesting implications:

© source: Kate Martin; graphic: PHOTON International

New England saw 36 price spikes above $100 per MWh in July (left) and August 2002 (right). This may seem like a lot, but only 0.6 percent of the total hours in this area in 2002 showed price spikes above $100 per MWh, which means PV’s ability to peak shave has a big impact even if applied for these few hours.

1) For policymakers considering programs to support PV market development, the electricity price reductions are large enough to consider incorporating them as potential benefits of PV programs. The savings from 1 GW of PV over a 25-year period have a net present value of $1.50 to $3.30 per W. (This is just a simple net present value calculation over 25 years based on annual savings of $130 million and $280 million from the estimates for 2002. Assumed is a 7 percent discount rate, no degradation of system, and no change in electricity prices.) It is also worth noting that, although these price reductions will widely benefit all consumers, large industrial consumers will benefit the most from wholesale price spike reductions. This is true even though residential customers might bear the cost of installing rooftop PV systems.

2) For generating companies and grid operators, it is worth understanding how PV's growth could impact existing assets, operations, and earnings. This could lead to potential changes in investing and operating strategies.

© source: Kate Martin;
graphic: PHOTON
International

3) For Japan and Germany, it may be worth conducting a similar analysis to better understand the impact of PV on electricity prices. Interestingly, PV generation in Japan and Germany appears likely to approach levels near those simulated in this thought-experiment (e.g. 1 GW of PV in New England is 0.7 percent of generating hours) by the end of the decade.

These three implications suggest that it would be worthwhile to take this analysis beyond the thought experiment. This would include using data from real and simulated PV systems that are distributed across the region, constructing simplified models to test the calculations of this model, and modeling other power pools and other years.

Kate Martin is a Ph.D. candidate in the Engineering Systems Division at the Massachusetts Institute of Technology. She works in the Laboratory for Energy and the Environment on a number of issues relating to the use of historical data in improving the modeling and understanding of electric power systems.