The Study And Application Of Neural Network Model Optimization For Short-term Load Forecasting

This thesis starts with the development and actuality of load forecasting research for power system. First, the research actuality of artificial neural network method used for short-term load forecasting is summarized, and then the characteristic of power load and the impact factors of forecasting precision are analyzed. Through the analysis of the history load of Xingtai and using the neural network based on PSO, the short-term load forecasting model which synthetically considers every kind of impact factor is created. The input load data and temperature are normalized, and date type and weather condition variable are quantitatively transacted. Simulation results show that, the neural network forecasting model based on PSO (PSO-NN) created in this thesis can improve forecasting precision and speed, and its forecasting capability is obviously better than the neural network model based on BP algorithm (BP-NN).