Research On Load Forecasting Based On Distribution Network Planning

Load forecasting is a key factor in distribution network planning.The accuracy of forecasting results directly relates to the safe operation of the power system and the utilization of power resources.Therefore,the research of load forecasting method must advance with the times.With the rapid development of social economy,the development of modern electric power load is affected by more and more uncertainties.This paper mainly studies two kinds of load forecasting methods,namely total load forecasting and spatial distribution load forecasting,summarizes and proposes a more comprehensive forecasting model,and validates it through simulation experiments and planning examples.This paper introduces several common methods of total load forecasting,and analyzes their respective advantages and disadvantages as well as the scope of application to provide reference for the selection of load forecasting methods for distribution network planning.For the defects of a single prediction method,a combined prediction model based on genetic algorithm optimization is established.The model considers the advantages of various algorithms synthetically,and uses the excellent global optimization ability of the genetic algorithm to optimize the weight of the model.And the simulation proves the effectiveness of the method.Aiming at the non-linear characteristics of monthly power load data of modern power system,which has both the trend of growth and volatility,a load forecasting model of hybrid SVM based on wavelet transform is proposed.The wavelet transform is used to decompose the load sequence into sub-sequences of different scales.Considering the seasonal fluctuation of load,the temperature factor is taken as the input variable to construct the mixed kernel function LWPSO-LSSVM.The load subsequences obtained by the decomposition are placed in the basic membrane of the membrane system,and prediction models are established for parallel prediction.Then each subsequence prediction data is reconstructed to obtain the prediction result.The application research of power grid load data in a certain area of Sichuan Province shows that the proposed model has significantly improved the prediction accuracy and efficiency compared with the traditional kernel function support vector machine.Under the premise of researching and analyzing the spatial load forecasting method,by investigating the economic development status of the planning area,the characteristics of the load itself,and the land use planning of the relevant departments in the planning area,the spatial forecasting method suitable for this area has been selected: classification and partition prediction.The planning area was modeled by the zoning method,and a saturation load density index selection scheme was formulated in conjunction with local load development.With reference to the land use planning of the planning area,the distribution of the total load and the load distribution of the planning area are predicted.A set of planning network planning load forecasting plan was made and the feasibility of the plan was verified.