Research On Medium And Long Term Power Load Forecasting Based On PSOEM-LSSVM And Its Application

Power load forecasting is an important task in the electricity sector, and correctand reliable forecasting results can maintain the stability of the power grid operation,to ensure the people’s normal life and social normal production, and it is effective forreducing the cost of producing electricity, and it is helpful to achieve continuousimprovement of society and economic efficiency. Therefore, a significant thatwhether a power enterprise management goes to modernization is the level of powerload forecasting, especially for our country that it has an unprecedented developmentof the power industry now, solving the problem of power load forecasting has becomean arduous and important task we face.Power load forecasting is generally divided into the ultrashort term, short term,medium term and long term power load forecasting. As an important work of powerplanning department, the medium and long term power load forecasting can help todecide the time, location, type, and size of the new generator sets or substation, helpdetermine the grid planning,and decide the construction and development of the grid.Forecasting method is the core issue of power load forecasting, with thecontinuous progress of modern society and the continuous development of scienceand technology, load forecasting technology develops gradually and is in-depth.Basedon the analysis of power load forecasting, this paper bulids a mudium and long termload forecasting model of particle swarm optimization with extended memory(PSOEM) optimizing least squares support vector machine (LSSVM). Support vectormachine (SVM) is a machine learning method based on statistical learning theory, forthe local minima, high-dimensional, non-linear and small sample problems in thetraditional algorithms,SVM method can be used to resolve them. SVM is consideredto be the best alternative method instead of neural network, it is good to avoid slowconvergence, easily trapped into local minimum point, poor generalization capability,and network structure selection difficulty shortcomings in neural network method.LSSVM can reduce the computational complexity of the standard support vectormachines, and has a faster solution speed and better anti-jamming capability,it is anextension of support vector machine. In allusion to the problems of weak directivityand purpose during searching optimal solution in PSO,this paper uses PSOEM tooptimize parameters in LSSVM,which avoids the blindness of parameters selection and realizes the automation of parameter optimization, so as to establishPSOEM-LSSVM forecasting model.PSOEM-LSSVM forecast model is verified on the feasibility and effectiveness ofthe medium and long term power load forecasting by example simulation analysis,andthe application of the medium and long term power load forecasting in urban powergrid planning are researched and analyzed in PSOEM.