Locational marginal pricing prediction in a competitve electrical market using computational intelligence

Deregulation has created a competitive market among power market participants, and the pricing system plays an important role. Locational marginal pricing (LMP) provides clear market signals that identify the locations where power market participants could make their decisions so as to maximize their profits. In this work, artificial neural networks (ANNs) models are used to predict hourly LMP. ANN is trained using the particle swarm optimization (PSO) algorithm. PSO aims to minimize the error function by adjusting neural networkpsilas weights and biases using a stochastic optimal search. Wavelet transformed data is fed into neural network as pre-processing stage in order to get a better price pattern that will be reliable for forecasting. The historical LMP and corresponding load demand and temperature are trained, validated and tested over a period of one season. The efficient generalization of proposed model is investigated using early stopping technique. The results were compared with neural models using conventional back-propagation (BP) algorithm and radial basis function (RBF) and yielded encouraging results.