Abstract:

The new power system's high-voltage distribution grid faces challenges posed by the large-scale and diversified connection of loads. Substation load clustering is a core method for accurately identifying user electricity consumption patterns and optimizing grid resource allocation. Its analysis results can directly support grid planning, demand-side management, and the formulation of renewable energy integration strategies. Therefore, it is urgent to conduct substation load curve clustering analysis to precisely analyze differentiated load patterns and their dynamic evolution patterns, thereby providing data support for intelligent distribution grid operation decisions. Addressing the limitations of the iterative self-organizing data analysis techniques algorithm (ISODATA), such as slow convergence speed and difficulty in capturing high-dimensional data features—particularly the insufficient capture of load data's dynamic characteristics—this study enhances the algorithm's ability to analyze high-dimensional features of substation load curves by optimizing the initial cluster center selection strategy and introducing a kernel function mapping mechanism. After completing missing value filling and data standardization preprocessing, this algorithm first optimizes the selection of initial clustering centers based on the maximum distance criterion to maximize the heterogeneity between initial centers and improve clustering stability. Second, it introduces a kernel function mapping mechanism to map load curves to high-dimensional space clustering, achieving explicit decoupling and clustering analysis of high-dimensional features. Simulation results indicate that in terms of feature extraction capability, the principal component analysis (PCA) feature space generated by the improved algorithm exhibits significant differences in the seasonal load characteristics of substations, enabling better capture of high-dimensional load features; In terms of algorithm performance, the improved algorithm reduces execution time by 32.8%, lowers the Davies-Bouldin Index (DBI) by 29.1%, and increases the Dunn Index (DI) by 42.9%, validating the effectiveness and superiority of the proposed algorithm.

Key words: substation, load clustering, clustering effect indicator