Fault Diagnosis Of Three-Level Inverter Based On QLSTM Neural Network

As a conversion device to convert direct current into alternating current,inverter brings great convenience to people’s daily production and life.Multilevel inverters are widely used in medium and high voltage fields because they can achieve high voltage output from low withstand voltage devices,but their special structure also increases the possibility of fault.As the inverter fault will have an impact on the inverter itself and the whole power generation system,so the research of effective fault diagnosis method is of far-reaching significance to the stability and safety of the whole system.In this paper,the current status of domestic and international research on multilevel inverter fault diagnosis methods and quantum neural networks is reviewed.This is followed by an analysis of the NPC three-level inverter power tube IGBT open-circuit fault and the three-phase output voltage waveform.Then,we embed Variational Quantum Circuits(VQC)on the basis of Long Short-Term Memory(LSTM)network to form an inverter fault diagnosis model of Quantum Long Short-Term Memory(QLSTM)network.And in order to extract the deep features in the NPC three-level inverter fault data,the CNN-LSTM network is combined with VQC into a QCNNLSTM network model,which further improves the inverter fault classification capability.The details are divided into the following points:(1)This paper firstly introduces the structure and working principle of NPC threelevel inverter.Then the inverter IGBT open-circuit fault characteristics and the threephase output voltage waveforms of two or less IGBT open-circuits are analyzed.All the original data of fault diagnosis research in this paper are obtained through simulation model.(2)By inputting the LSTM network hidden layer output and the memory cell state into VQC separately,the QLSTM model is fused.In order to evaluate the classification effectiveness of the QLSTM network,the model is compared with the LSTM network.The experimental results show that the classification accuracy of the QLSTM network is 1.31% higher than that of the LSTM network,and it also shows higher accuracy under different noise conditions.Therefore,QLSTM network has better fault diagnosis effect and noise-resistant capability than LSTM network.(3)In order to make full use of the automatic extraction of deep features of Convolutional Neural Network(CNN)and the quantum advantage of VQC,the convolutional kernel in CNN is replaced by VQC to implement convolutional operations in this chapter,and the LSTM network is used for fault identification,thus combining as a QCNN-LSTM network.The QCNN part is used to extract fault features and data dimensionality reduction,and the resulting signal features are divided into training data and test data,which are then fed into the LSTM network for fault classification.According to the experimental analysis,Compared with CNN,LSTM and CNN-LSTM models,QCNN-LSTM network has better classification effect with99.73% classification accuracy and shows good noise-resistant capability under noisy data.Therefore,the QCNN-LSTM network can be applied to three-level inverter fault diagnosis to effectively improve the fault classification capability.