A Short-term Load Forecasting Method Based On Big Data And Topological Network

Short-term load forecasting plays an important role in the management and distribution of electric power system.Electric power load is affected by many factors,including weather,date type,natural disasters,large scale activities,economic development,holiday,and so on.The variation of electric load results from a variety of natural and social factors.In this paper,the inner mechanism of load variation has been discussed and a short-term load forecasting method based on big data and topological network has been presented.Factors,including weather data,seasonal attribute,calendar attribute,holiday attribute,and economic development rate,have been considered to cause the variation of load and selected as the feature.With the analytical method of big data,the complicated causality correlation between the factors and load variation has been explored so as to reveal the pattern and trend of load variation and improve the accuracy of load forecasting.Besides,for the purpose of increasing the efficiency of constructing training set,a method combining big data and topological network has been proposed,in which the random walk with restart(RWR)algorithm has been applied.Furthermore,based on a large number of experimental analyze,a feedforward correction method based on load growth rate has been proposed.Forecasting case studies show that the proposed method outperforms the classical support vector machine method and an intelligent method based on load variation mechanism,and decreases the mean absolute percent error by 29.0%and 15.2%.