Pressure-based forecasting of next-day solar energy availability using evolutionary fuzzy rules

Prediction of solar energy availability is an important part of a number of applications, from weather and climate modeling, to forecasting power output of photovoltaic cells. This work proposes and evaluates the use of evolutionary fuzzy rules, discovered by genetic programming, for accurate and inexpensive forecasting of next day available solar energy, based solely on atmospheric pressure. The use of only one meteorological variable simplifies the collection, storage and prediction tasks, making this approach suitable for a number of hardware-limited applications. The proposed forecasting system uses a time series of pressure values for the hours leading up to the sunrise of the day for which predictions are made. The pressure values and their differences are submitted to the evolutionary optimizer that combines these values into simple fuzzy rules. The resulting rule set provides predictions that are a substantial improvement over simple analytical estimates. The results of pressure-based forecasters also compare favorably with other simple estimation systems presented in the literature.