Fault Diagnosis Method of Smart Meters Based on DBN-CapsNet

Rapid and accurate fault diagnosis of smart meters can greatly improve the operational maintenance ability power systems. Focusing on historical data information meters, a model based an improved capsule network (CapsNet) is proposed. First, we count sample size each type, mixed sampling method combining undersampling oversampling used to solve problem distribution imbalance size. The one-hot encoding adopted samples containing more discrete disordered data. Then, strong adaptive feature extraction capability nonlinear mapping deep belief (DBN) are utilized single convolution layer part traditional network; DBN also address high dimensions sparse due encoding. important features key input extracted as primary layer, dynamic routing algorithm construct digital capsule. Finally, results experiments show that effectively accuracy shorten training time.