Application of BP algorithm in short-term load forecasting

This paper firstly analyzed the traditional short-term power load forecasting theory and methods and made a detailed research and analysis on the application of BP neural network in short-term power load forecasting, pointing out its deficiencies. Then the refined research was conducted on the prediction model in this paper and eventually the short-term power load forecasting model based on improved BP algorithm was established. Simultaneously, accorded with the established model structure, this paper adopted the traditional BP algorithm, variable learning rate (adaptive) BP algorithm and additional momentum - adaptive BP algorithm to predict and simulate the power load condition in a region, and the actual results were compared and analyzed, which illustrated the difference between the different algorithms. The additional momentum - adaptive BP algorithm can achieve excellent effect of prediction, and fully meet the accuracy requirement of 0.001, which simultaneously reduced the number of iterations by more than 93.5% and greatly cut down the predicted time and improved the prediction efficiency.