Load and locational marginal pricing prediction in competitive electrical power environment using computational intelligence

This paper presents a comprehensive study of forecasting a day-ahead of load and locational marginal pricing (LMP) using artificial intelligent systems. An artificial neural network (ANN) is trained with a stochastic optimization technique called particle swarm optimization (PSO). This training algorithm works to adjust the network weights and biases as to minimize the error function. Wavelet transformed data is fed into neural network as preprocessing stage in order to get a better price pattern that will be reliable for forecasting. The proposed models were trained and tested using real data consists of historical load and LMP and corresponding influence variables such as weather information and marginal losses cost (MLC). The data used is from NYISO and Weather Source Stations, Buffalo, New York over a period of three years (2001-2003). Simulation results are compared with that of conventional back-propagation (BP) neural network and radial basis function network (RBFN) and provided highly accurate generalization capability.