Probabilistic Forecasting of Hourly Electricity Price by Generalization of ELM for Usage in Improved Wavelet Neural Network

In restructured markets where transactions process is competitive, forecasting of electricity price is inevitably an important available tool for market participants. Due to the sensitivity of forecasting issues in market's performance, and high prediction error resulted from the behavior of price series, nowadays probabilistic forecasting highly attracted participants' attention. In this paper, a probabilistic approach for the hourly electricity price forecasting is presented. In the proposed method, the uncertainty of predictor model is considered as the uncertainty factor. The bootstrapping technique is used to implement the uncertainty and since the method is needed to be fast and of low computational cost in the daily forecasting, a generalized learning method is applied, which has high accuracy and speed. This newly presented learning method is based on generalized extreme learning machine approach to be used for improved wavelet neural networks. Also in order to reach more accommodation, the predictor model with the changes of price time series, the wavelet preprocessing is used. Effective performance of the proposed model is validated by testing on data of Ontario and Australian electricity markets.