Short Term Electricity Price Forecasting with Multistage Optimization Technique of LSSVM-GA

Abstract

Price prediction has now become an important task in the operation of electrical power system. In short term forecast, electricity price can be predicted for an hour-ahead or day-ahead. An hour-ahead prediction offers the market members with the pre-dispatch prices for the next hour. It is useful for an effective bidding strategy where the quantity of bids can be revised or changed prior to the dispatch hour. However, only a few studies have been conducted in the field of hour-ahead forecasting. This is due to most of the power markets apply two-settlement market structure (day-ahead and real time) or standard market design rather than singlesettlement system (real time). Therefore, a multistage optimization for hybrid Least Square Support Vector Machine (LSSVM) and Genetic Algorithm (GA) model is developed in this study to provide an accurate price forecast with optimized parameters and input features. So far, no literature has been found on multistage feature and parameter selections using the methods of LSSVM-GA for hour-ahead price prediction. All the models are examined on the Ontario power market; which is reported as among the most volatile market worldwide. A huge number of features are selected by three stages of optimization to avoid from missing any important features. The developed LSSVM-GA shows higher forecast accuracy with lower complexity than the existing models.