Smart Meter Fault Diagnosis Model Based on DBN-LSSVM Feature Fusion

Improving the accuracy of smart meter fault diagnosis is of great significance to the reliable operation of power acquisition systems. Aiming at the problems of difficult feature extraction and the high dimension of multi-sensor data of smart meters, based on deep learning, this paper proposes a Deep Belief Network - Least Squares Support Vector Machine (DBN-LSSVM). The deep belief network is trained by unsupervised learning to realize the feature extraction of multi-sensor data. And the least squares support vector machine algorithm is used to obtain the multi-dimensional features in depth to achieve the goal of fault multi-classification diagnosis. Taking the fault data of smart meters collected in recent years as a practical example, the proposed model is compared with typical methods. The experimental results show that it has higher classification accuracy.