The Value of Day-Ahead Solar Forecasting in the United States: A Market Bidding Perspective

Abstract:

The recent expansion of the solar energy industry has motivated research worldwide on the forecasting of solar photovoltaic (PV) power generation. With a forecast of the next day’s production, solar plants can make generation commitments in day-ahead markets with imbalances settled in real-time markets, which can often provide greater value to electricity markets versus participating only in the real time-markets. Though solar forecasting is an active area of research with new refinements being continually developed, there are few studies that evaluate the value of solar forecasting techniques. Even fewer studies take an empirical approach based on market wholesale prices.

This study fills this gap by quantifying how the value of solar forecasting has varied with United States region and time (2012 – 2018) at 663 utility-scale photovoltaic plants. We compare the added value of perfect forecasting to two imperfect forecasting approaches. The first approach, ‘persistence forecasting’, assumes that tomorrow’s solar equals today’s. The second approach utilizes forecasted meteorology from the North American Mesoscale Forecast System (NAM). We explore what underlying factors drive the variation in forecasting value, and we gain insight about how the value might evolve in the future based on these historical trends.

For the most recent year in our study period, 2018, we found that participation in the day-ahead market with perfect forecasting increases revenue by a generation-weighted average of 5.40 $/MWh, compared to participation in only the real-time market. The NAM-based forecast predicts an average value of 5.04 $/MWh, higher than the value from the persistence forecast, which gives 3.84 $/MWh. It is notable that the publicly available NAM forecast captures ~90% of the value as perfect forecasting. Overall, the ~ $5/MWh value premium from participating in day-ahead markets represents an increase of roughly 15% over the real-time market value, in 2018. This premium is significant in light of recent long-term solar purchasing agreements which range from 20 to 40 $/MWh. Further results throughout 2012-2018 provide more insights into temporal trends, the geographical distribution of forecast values, and the relationship between forecast errors and market penalties.