Mid-Long Term Forecasting For Power Load Based On BP Neural Networks

Mid-long term load forecasting in power system is a long-range layout.It provides detailed plans of power sources,power grid so as to reconstruct grid or expand grid,make arrangements for the annual maintenance and adjust power operation.The main factors of influencing the forecasting accuracy in electric power load are social and economic development,population,product unit consumption,proportional allocation of industrial structure et al.Due to the uncertainty and randomness of changes of the industrial and commercial structure for power demand and market operation,the mid-long term electric power load forecasting is increasing difficultyRecent achievements and the future trend of mid-long term load forecasting in the power industry and other industry all over the world is analyzed.The merits and drawbacks of the load forecasting methods and its application scope are comprehensively compared.Method of neural network on the structure of back propagation multilayer feed-forward(BP)neural network is studied.It includes learning parameters selection,the algorithm of standard learning.BP neural network algorithm and its improved methods are also compared.Based on the local area economic development,population and the historically real data of the power load,neural network model predicting power load for future 5-10 years are established.To accelerate training and realize predicting results accurately in abnormal datas,time series prediction based on standardized processing method is proposed for load forecasting.It is applied to medium and long-term power load forecasting in Suzhou Grid,testing the accuracy and forecasting time.The testing results show that the proposed model reach the load forecasting requirement in local power area.