Short-Term Load Forecasting System Based On Intelligent Optimization Model

Short-term load forecasting is a complicated system, and the intelligent technology has gained wide application due to its ability to combine various factors and realize non-linear flections through continious training. Being the most developed method, BP neural network is faced with two problems in practice. One is its slow convergence as well as easily trapped into local minimum while the other is the structure design problem. Therefore, we first used momentum BP algorithm with variable step, adopted PSO to train initial weights of BP network, and proposed a novel intelligent optimization model PSO-BPNN. Then we determined a suitable network structure and adopted special normalization method according to short-term load characteristics of Baoding power network. In addition, we applied the optimization model to forecast simulation and analized the simulation results. Finally, through VB 6.0 and ACCESS database the author programmed the PSO-BPNN based short-term load forecasting system.