An adaptive BP-network approach to short term load forecasting

This paper proposes an adaptive BP-network approach to short term load forecasting (STLF) in a deregulated environment, which is to determine the BP-network structure using genetic algorithm (GA). The aim is to optimize the network structure and improve the accuracy of STLF. The realization process consists of three steps. In the first step, the number of hidden nodes of BP-network is calculated by use of GA. In the second step, by use of GA a fittest initial weight value is selected from the solution group of initial weight values to avoid the blindness in the selection of initial weight value. In the third step, combining the structure of the obtained BP-network and the fittest initial weight value, the STLF of power system can be performed by use of improved BP algorithm. Simulation results show that the percentage errors of mostly of 24 h forecasting load are less than 3%, and prove that the approach can meet the need of forecast accuracy and enhance the performance of the network.