Shared Nearest Neighbor Density Peak Clustering based Methodology for Energy Consumer Behavior Classification

The classification of user’s energy consumption behavior is the basis of generating typical energy consumption scenarios and further systematic analysis. Firstly, aiming at the sensitivity of initial value of K-means algorithm, this paper proposes density peak clustering algorithm for the classification of user’s energy consumption behavior. Secondly, according to the temporal characteristics of users’ energy consumption data, the shared nearest neighbor clustering algorithm is used to improve the peak density clustering algorithm, which avoids the division error caused by the simple definition of distance in the traditional peak density algorithm. Finally, the energy consumption behaviors of users are divided based on the shared nearest neighbor density peak clustering method. Compared with other clustering algorithms, the clustering effectiveness and convergence of the improved algorithm are stronger.