Simulating Solar Forecasting for Energy Market Decision Models

Today’s energy markets are increasingly challenged by the uncertainty of supply inherently associated with weatherdependent energy resources [1]. Market participants’ behaviour depends on the available forecasts. Current models of market participants’ behaviour are based either on perfect foresight assumptions, or on a single forecast, usually 24 hours ahead of time. In reality, consecutive, increasingly more reliable forecasts become available closer to real time. These improvements affect consecutive decisions of market participants. For energy market modelling, usually only historical data, not the preceding forecast are available. Limited amount of work currently exists on simulating consecutive, increasingly more reliable forecasts from the available historical data. This paper details and analyses a statistical approach to solar forecasting based on historical data, for multiple forecasts, up to several days in advance.