On-load Tap Changer Fault Diagnosis Based on Improved Whale Algorithm Optimized LSTM Neural Network

In order to realize accurate fault diagnosis of on-load tap changer (OLTC), an OLTC fault diagnosis method based on improved WOA optimized long short term memory(LSTM) neural network was proposed. Firstly, the WOA algorithm was improved by hybrid initialization strategy, nonlinear dynamic update strategy and elite solution retention strategy to improve the convergence. Then, the number of hidden layer nodes and learning rate of LSTM were optimized by improving WOA algorithm to obtain the optimal LSTM model. Then, the feature vector was formed by extracting the characteristics of OLTC vibration signal in time domain, frequency domain and energy entropy. The feature vector was input into the optimal LSTM neural network for learning and fault diagnosis. The results show that the average accuracy of the method is 98.7%, which is suitable for OLTC fault diagnosis.