Research And Application Of Medium And Long Term Load Forecasting Technology

Power load forecasting is one of the important work in power system.It not only provides index data support for the planning,transformation,expansion and development layout of power grid,but also provides data auxiliary function for the national strategy layout of economic and energy development.For a long time,the work of power load forecasting is divided into short-term scenarios and medium and long-term scenarios.In the scenario of short-term power load forecasting,the forecasting period is short,the influencing factors are few,and the forecasting accuracy is relatively high.The frontier workers have produced a lot of exploratory work achievements in this field,the technology research is quite mature,and the technology iteration speed is fast.Comparatively,in the medium and long-term scenario,compared with the short-term load forecasting,the medium and long-term load forecasting is affected by more factors,the forecasting cycle is longer,the model accuracy is difficult to guarantee,the exploration of cutting-edge technology is relatively less,and the technology change speed is slower.Therefore,in this research work,the main research core of this study is to explore all kinds of technical solutions of power load forecasting scenarios,aiming to design an optimization and improvement method suitable for medium and long-term power load scenarios.The specific research contents mainly include:(1)the characteristics and influencing factors of load data are studied through historical technical data,and the real power data is used to analyze the power load forecasting scenarios At the same time,it puts forward corresponding solutions for some problems in data preprocessing.(2)This study explores the mature technical scheme of short-term load forecasting scenario,through the application of tree model in short-term load forecasting scenario,carries out the corresponding simulation experiment and result analysis.(3)Aiming at the existing research content,this study analyzes XGBoost and DNN in the medium and long-term scenarios.Then,based on the analysis results,this study proposes a combination model method: XGBoost-DNN,which combines short-term scenarios with medium and long-term scenarios.Finally,based on the real load data,the simulation experiment and result analysis further verify the scientificity and rationality of the combined model method.