Research On Electric Power System Short-Term Load Forecasting Using A Neural Network

The short-term power load forecasting method is the base of optimization operation for power system. Accurate load forecasting is advantageous to improving the secure and economic effect of power system and can reduce the cost of generating electricity. Finding an appropriate load forecasting method to improve the accuracy of precision has important application value. Short-term system load variation has apparently periodicity, the load is a variable that is influenced by many kind of factors, it can be subdivided into periodic basic load component and variable load component. According to the characteristics of electric short-term load, this paper firstly analyzed the main factors which influence the regular variation of the load, concretely analyzed the day-types and weather factors which affect the short-term load forecasting and pre-disposal the past load data which influence the precision of load forecasting. The initial load is forecasted by artificial neural networks , this method adopted the BP algorithm that was used widely, this paper analyzed the method and procedures of BP Model in the practical load forecasting, Convergence speed of ANN is improved by using momentous factors and alterable step BP arithmetic in ANN; Aim at the ANN can't input language information variables directly, this paper explains the general theory and meaning of fuzzy set theory, fuzzy the weather factors, seasons and day-types in the daily load curves forecasting using different method, update the elementary result by the fuzzy reasoning rules. During forecasting phase, the past load data sample space is reconfigured as 12 fuzzy clustering sub-spaces by fuzzy clustering analysis. Each group is modeled by a separate module based on neural networks to forecast hourly loads for the next hour to 24 hours out. Though the reasonable choice of the fuzzy reasoning rules update the elementary result to improve the accuracy of prediction. Practical examples indicate that the forecasting method is convenient and practical, this method is more accurate and fast than conventional method.