Current Unbalance Cluster Analysis Based on Self-organizing Competitive Neural Network

In power system, current unbalance is a kind of common fault that seriously affects the safety and efficiency of power system. There are many reasons that may cause three-phase unbalance, and now rely on manpower to judge according to specific conditions. This paper proposes a new algorithm for clustering analysis of unbalanced three-phase current data. We define a serials of feature parameters, and then use self-organizing competitive neural network for clustering analysis to subdivide current unbalance into five categories. In the experiment, a large amount of historical current data is analyzed by the proposed algorithm. We get five categories with obvious features and differences. The clustering results are reasonable and interpretable. The algorithm makes full use of the large amount of unmarked historical data produced by power system, and is helpful for the early warning of current unbalance and pre-judgment of causes.