A Hierarchical Clustering Method for Power Load Curves Based on Data Partitioning

Abstract

The invention provides a hierarchical clustering method for power load curves based on data division, which relates to the field of power grids. Based on the clustering effectiveness evaluation function DBI, the invention uses a hierarchical clustering algorithm based on data division for power load curve data. Clustering, while using the method of setting thresholds to further reduce the clustering time; because the present invention adopts the method of dividing the overall data into several subsets and then clustering them separately, the clustering time of the power load curve is greatly shortened, and multiple times Experiments show that when the clustering quality is evaluated by the DBI index, the clustering quality of the hierarchical clustering algorithm based on data partition is generally improved by about 3% compared with the traditional hierarchical clustering algorithm.

Description

technical field [0001] The invention relates to the field of power grids, in particular to a clustering method for electric load curves. Background technique [0002] With the introduction of the concept of demand-side response, user-side resources have gradually attracted the attention of academia and industry. Whether user-side load resources can participate in energy Internet supply and demand regulation is of great significance to the safe and stable operation of the entire power system. The subdivision of users in the grid is crucial to the formulation of precise incentive policies to enable user-side resources to participate in the supply and demand regulation of the grid. The load curve is the most important characteristic of power users. Through the cluster analysis of the user load curve, the load pattern of the user's electricity consumption can be extracted, which is helpful for deeply grasping the law of the user's electricity consumption, evaluating the potenti...

AI Technical Summary

Problems solved by technology

At the beginning of the study, the research on the clustering of load curves focused on using basic clustering algorithms to cluster them. However, as the scale of data continues to increase, only using simple clustering algorithms will lead to too long clustering time, and the clustering time will be too long. The class quality is not high, which makes the load curve clustering problem difficult to solve; in the past two years, some scholars have combined the two basic clustering algorithms to improve the clustering quality and reduce the clustering time, and most of them will The research on the combination of the two clustering algorithms is to combine the k-means algorithm with another algorithm, but due to the randomness of the initial clustering centers of the k-means algorithm, the clustering results are not repeatable and unstable. At the same time, the combination of the two clustering algorithms will also increase the time cost; some scholars also apply dimensionality reduction technology to the clustering algorithm, but no matter which dimensionality reduction technique is used, it will reduce part of the information in the original data, resulting in clustering inaccuracy of

Images