Study On The Methods Of Spatial Forecasting For Distribution Network Based On Load Density Method

Load forecasting whose accuracy affects the quality of distribution networks planning is not only an important component of distribution networks planning, but also the prerequisite and basis for the grid planning. Load forecasting can be subdivided into gross load forecasting and spatial load forecasting, which play an important guiding role in distribution network planning.This paper focuses on features and applications of two spatial load forecasting methods (land-use method and load density index method) which is currently mainly studied in domestic and foreign countries. The applicability and usefulness of load density index method in our country is described and its key and difficult in engineering practice is also summarized in the paper, based on detailed analysis of factors related to spatial load forecasting.The traditional methods to obtain load density are often based on experience or simple comparison,the results can hardly meet the accuracy requirements. So a novel spatial load forecasting (SLF) methodology based on self--adaptive Least Squares support vector machine (LS-SVM) for distribution network is proposed in the paper. During the process of calculating the load density index, FCM method is firstly used to cluster the nature of different land use load into several grades and establish reletively adaptable and elaborate load density index system, then in order to improve the generalization capability of the LS-SVM prediction model, the more similar training samples are selected for the LS-SVM prediction model according to the planning properties of the land to be predicted. Finally, Genetic Algorithm(GA) is used to optimize the parameters of LS-SVM automatically, which effectively further improves the adaptability and forecasting accuracy of the prediction model. The applicability and effectiveness of the method are demonstrated by using a real SLF case. During the process of calculating the total load of the planning area, this paper intelligent forecasting method is adopted to avoid seeking the simultaneous-rat, which even enhances the reasonability and scientificity of spatial load forecasting.Because the traditional methods to obtain load density do not focus to seek the nonlinear functional relation between load density and its impact factors, the various factors'weight and influence is also not taken into consideration during the prediction, So the paper futher proposes a novel and new method to obtain load density based on ANFIS for distribution network. The methodology of entropy weight coefficient is firstly used to treat the value of every influencing factor with its weight ,then ANFIS is brought to establish the forecasting model of ANFIS to forecast the load density with the Fletcher-Reeves learning algorithm that improved the conventional mixed algorithm,which overcomes the conventional motheds'faults such as measurelessness of prediction result and low forecasting accuracy of the prediction model. Finally, the applicability and effectiveness of the method are demonstrated by using a real case.