Abstract
Electricity price forecasting is a potential challenge for market participants and managers due to the high volatility of electricity prices. Price forecasting is also considered as an important management goal for market participants since it forms the basis of maximizing profits. These markets are usually organized in power pools and administrated by the independent system operator (ISO). The aim of this study is to examine the performance of asymmetric neuro-fuzzy network models for day-ahead electricity price forecasting in the ISO New England market. The implemented model has been developed with two alternative defuzzification models. The first model follows the Takagi-Sugeno-Kang scheme, while the second follows the traditional centre of average method. A clustering scheme is utilised as a pre-processing step to derive the necessary number of clusters and eventually the number of fuzzy rules in the network. Simulation results corresponding to the minimum and maximum electricity price indicate that the proposed network architectures could provide a considerable improvement for the forecasting accuracy compared to alternative learning-based schemes.