Neural Network Based Power System Ultra-Short Term Load Forecasting

Power system load forecasting is based on historical data such as social influences,political factors,and natural conditions,to study the impact of the law of historical data of electric load on future power load,to explore the relationship between power system load and various related factors,and finally to conduct scientific research on future power load.Intelligent,accurate and efficient power system load forecasting is one of the key methods to ensure the stable operation of power system.This paper briefly introduces the current research and development trend of power load forecasting,overviews a variety of common classic methods of power system load forecasting,clarifies the fundamental principles of BP neural network,Elman neural network and PSO optimization algorithm,then uses mathematical statistics to test the influential factors of each dimension of ultra-short term electrical load,and finally retains the most significant factor affecting ultra-short term power load,which is taken as the input variable of the prediction model,and the load at the next moment is taken as the output variable.Based on the shortcomings of BP and Elman algorithm,such as low precision,poor stability and easy to fall into local optimum,this paper studies the Elman neural network model optimized by PSO algorithm,compares the simulation experiment results among four neural network models(i.e.traditional BP,improved BP,Elman,and PSO＿Elman),and finds that PSO＿Elman neural network model has better performance in terms of speed and precision than the former two models.This conclusion has inspired a new idea for the research work of ultra-short term load forecasting.In the end,according to the actual application,this paper proposes the limitations of the model and the areas that need to be improved.