Day-ahead Price Forecasting of Electricity Markets by a New Hybrid Forecast Method

Abstract

Energy price forecast is the key information for generating companies to prepare their bids in the electricity markets. However, this forecasting problem is complex due to nonlinear, non-stationary, and time variant behavior of electricity price time series. Accordingly, in this paper a new strategy is proposed for electricity price forecast. The forecast strategy includes Wavelet Transform (WT), Auto-Regressive Integrated Moving Average (ARIMA) and Radial Basis Function Neural Networks (RBFN). Also, an intelligent algorithm is applied to optimize the RBFN structure, which adapts it to the specified training set, reduce computational complexity and avoids overfitting. In the proposed forecast strategy, the WT provides a set of better-behaved constitutive series, ARIMA generates a linear forecast and RBFN is developed as a tool for nonlinear pattern recognition to correct the forecast error. The proposed strategy is applied for price forecasting of electricity market of mainland Spain and its results are compared with the results of several other price forecast methods. These comparisons confirm the validity of the developed approach.