Probabilistic Forecast of Electricity Price based on Adaboost_RBF method

Accurate and reliable electricity price forecasting is essential for market participants to make various decisions in the deregulated electricity market. However, due to the time-variant and nonstationary of price, which is related to change of market competitors' strategies, predicting price accurately in advance is rather difficult. Therefore, probabilistic interval forecast instead of traditional point forecast can be of great significance to make bidding strategies. In this paper, a hybrid approach for probabilistic forecast is proposed with two-stage formulation: 1) An improved RBF NNs based on Adaboost algorithm (Adaboost_RBF) is proposed for point forecast of price. 2) Prediction interval can be obtained according to the statistical distribution of price forecast error. Effectiveness and reliability of proposed model is validated through case studies from Australian electricity market by comparing with existing methods such as RBF neural network and ARMA.