Locational marginal price forecasting in deregulated electricity markets using artificial intelligence

Abstract

Bidding competition is one of the main transaction approaches in deregulated electricity markets. Locational marginal prices (LMPs) resulting from bidding competition represent electricity values at nodes or in areas. A method using both neural networks (NNs) and fuzzy-c-means (FCM) is presented for forecasting LMPs. The recurrent neural network (RNN) was addressed and the traditional NN-based on a back-propogation algorithm was also investigated for comparison. The FCM helped classify the load levels into three clusters. Individual RNNs according to three load clusters were developed for forecasting LMPs. These RNNs were trained/validated and tested with historical data from the PJM (Pennsylvania, New Jersey, and Maryland) power system. It was found that the proposed neural networks were capable of forecasting LMP values efficiently.