Power Load Curve Clustering based on ISODATA

Abstract

Load clustering is the early basis of power grid system planning, load modeling, demand side management, load forecasting and other work. The traditional load classification method based on user types can not meet the needs of power grid services. Iterative Self-Organizing Data Analysis Algorithm (ISODATA) is an unsupervised learning dynamic clustering algorithm based on statistical pattern recognition. In view of the current problems that the initial clustering number of each algorithm is difficult to take and easy to fall into local optimum, the principle and implementation steps of ISODATA are introduced, and this algorithm is applied to the power load curve clustering. The clustering analysis is combined with specific power load curve samples, and the results prove that the clustering effect is better and the time improvement is larger. ISODATA is compared with the traditional clustering method to compare the clustering effect and the time loss of the algorithm. The results of the comparison experiments show that ISODATA has good clustering effect when applied to power load curve clustering.Isodata-based clustering of power load curves can fine distinguish users and provide decision support and scientific basis for the reliable operation of power system.