Abstract: Multi-classification of smart meter fault is of great significance for making reasonable and timely maintenance plan of smart meter. Aiming at the problem of multi-classification of smart meter fault, support vector machine（SVM） is used to build a multi-classification model. The established model extracts the output voltage, output current, output power, power factor error and other data of smart meter as the classification basis to build a multi-dimensional space for the smart meter pattern recognition fault classification, including error tolerance, DC current open circuit, DC voltage short circuit, and short line of control circuit. According to the limited sample information, the established model seeks the balance between complexity and learning ability, and makes the best classification of multi-dimensional operation information of smart meters between hyperplanes, so as to carry out fault classification. By introducing one class to improve the multi-class optimal classification plane set, thereby, it is suitable for multi class model. The chaotic particle swarm optimization algorithm is used to design the solution flow of the smart meter fault multi-classification method based on improved SVM. Finally, the proposed model is used to simulate the fault classification of smart meters in a distribution station area, and the rationality of the model is verified.