Research On Urban Spatial Load Forecasting Method Based On Data Mining

With the improvement of people’s living standards,the demand of various types of power users for electricity is also increasing,which puts forward higher requirements for the space load forecasting work in the early stage of distribution network planning.Moreover,as the variety of power loads is increasing,the power load in cities is also more complicated.The emergence of electric vehicles makes some loads has a movable nature.Therefore,in the space load forecasting,the spatial distribution of the charging load needs to be additionally carried out,which makes the spatial load forecasting results more accurate.And with the development of Internet,the massive data has provided new ideas for spatial load forecasting.Data mining technology can deeply analyze the correlation of data and build models to better predict the development of power load.Based on the above situation,it is of great significance for grid planning to carry out urban spatial load forecasting based on data mining technology.Based on the general method of data mining,this paper constructs a multi-source data mining architecture for power data.This paper analyzes the meaning of power multi-source data,the correlation between the source and the data and the power load.This paper introduces in detail the data collection process of power grid load forecasting based on open source data,and constructs a multi-data set to provide data support for urban spatial load forecasting.Based on clustering algorithm and neural network algorithm,a grid spatial distribution prediction method is proposed in this paper.The total load forecasting method is integrated to obtain the total load forecasting results of the development situation in the appropriate areas.The distribution of various load types of power users was obtained by cluster analysis of power users,and the load development characteristics of power users were analyzed based on the s-shaped development curve.Finally,the neural network algorithm is used to analyze the influence weight of the number of each type of power users on the existing load data,and the load prediction results of each power grid are obtained by combining the total load prediction results.The advantage of this method is that the urban total load prediction is used to correct the spatial load prediction of power grid units,and the result is smaller than that when only the grid development is considered.In addition,considering that the characteristics of EV charging loads in cities are quite different from traditional loads,this paper makes a planning prediction based on the space-time characteristics of charging loads.Firstly,the Bass model is used to predict the EV ownership,and the charging power characteristics are analyzed.At the spatial level,the parking generation rate of electric vehicles in various types of land is considered in this paper.Since the peak time of charging load in different types of land is different,the simultaneous rate is introduced at the time level to quantify the charging demand of electric vehicles at the space-time level.Finally,based on the improved density peak fast clustering algorithm,the charging demand is clustering analyzed,and the spatial distribution of charging load is obtained.Considering the simultaneous rate between the traditional load and the charging load,the prediction results of EV charging load and the traditional load are organically combined to obtain the forecast situation of urban spatial load distribution.