Study On Mid-long Term Load Forecasting In Regional Power Grid

Mid-long term power load forecasting is an important reference for power grid construction and planning,and can act as an important prerequisite for promoting the safe,economic and stable operation of power grid.The single model with strong limitations has its own characteristics and scope of application in load forecasting,Combined forecasting can maximize the adaptability of each single model after combination for different load series,and can improve the accuracy of load forecasting.In the traditional combination forecasting practice,the forecasters only select a single model according to their own experience for different load series,and do not use scientific theory as a supplementary to select the forecasting decision.In this paper,combination forecasting can be put into effect by analyzing the characteristics of each single model in depth,and by taking the original error of each single prediction model as the benchmark.the method of odds matrix is used to quantify the weight of the selected single model.Thus,the model with high weight value can be selected to participate in combination forecasting.In the power load forecasting,the influence of different historical periods on combined weight value is seldom treated differently,the principle of "near big and far small" is considered in the modeling of the optimal fitting combined model,sub optimal fitting combined model and optimal forecasting combined model,which can ensure the accuracy of the combined weight value to the maximum extent.Through the analysis of examples,including the comparison of different combination models,combination model and single model,it is proved that the model proposed by odds matrix in this paper has a good prediction effect in the medium and long-term load forecasting.