A Probabilistic Approach to Committing Solar Energy in Day-ahead Electricity Markets

Abstract Grid-tied solar is governed by a variety of complex regulations. Since a higher solar penetration imposes indirect costs on the grid, these regulations generally limit the aggregate amount of grid-tied solar, as well as the compensation its owners receive. These regulations are also increasingly limiting solar's natural growth by preventing users from connecting it to the grid. One way to address the problem is to partially deregulate solar by allowing some solar generators to participate in the electricity market. However, day-ahead electricity markets require participants to commit to selling energy one day in advance to ensure system stability and avoid price volatility. Thus, to operate in the day-ahead market, solar generators must solve a solar commitment problem by determining how much solar energy to commit to sell each hour of the next day that maximizes their revenue despite the uncertainty in next-day solar generation. We present a probabilistic approach to addressing the solar commitment problem that combines a solar performance model with an analysis of weather measurement and forecast data to determine a conditional probability distribution over next-day solar generation outcomes, which we use to determine solar energy commitments each hour that maximize expected revenue. We show that, as the deviation penalty for over-committing solar increases, our probabilistic approach enables increasingly more savings than a deterministic approach that simply trusts weather measurements and forecasts.