Probabilistic Statistical Model of Station Load Based on Improved Cluster Analysis and Kernel Density Estimation

In order to more accurately describe the load changes in the power system and establish a suitable load model, a probabilistic statistical model of desk load based on improved cluster analysis and nonparametric kernel density estimation is proposed. Firstly, the shortcomings of ISODATA clustering algorithm are analyzed, based on which this algorithm is improved by optimizing the selection of initial clustering centers to enhance the effect of clustering and accelerate the convergence speed of the algorithm, and the improved algorithm is used to cluster the power users and analyze the characteristics of electricity consumption of different users. Then, the probability distributions of different loads in each quarter are calculated by using non-parametric kernel density estimation method, and their probability density curves are extracted. Combined with the actual analysis of the load data of a number of users in a station area, compared with the traditional clustering algorithm in terms of clustering effect, the results verified the superiority of the algorithm proposed in this paper and the application value of the proposed model.