Abstract:With the development of smart grid construction, traditional low-voltage cause diagnosis based on detection technology has become a data big data classification technology based on data mining, and data classification research focusing on the cause of low voltage failure is still in its infancy in China. For this reason, a low-voltage diagnostic model of distribution network using improved clustering algorithm and support vector machine classification algorithm is proposed. The model first uses Canopy-Kmeans clustering algorithm to analyze the low voltage causes based on the historical operation data of the distribution network and obtain the possible low voltage causes. Then the particle swarm optimization algorithm is used to support the SVM data classification algorithm. Parameter optimization, finally using the support vector machine algorithm of the result parameter optimization to classify the real-time operation data of the distribution network collected by the smart meter for low-voltage causes and finally output the diagnosis result of the low-voltage fault. Experiments show that sampling based on particle swarm optimization based support vector machine diagnostic model can achieve 90% low-voltage cause diagnostic accuracy.