Fault Diagnosis For Open-Circuit Faults Of Power Transistors InThree-Phase Voltage-Source Inverters

Three-phase voltage-source inverters are widely used in rail transit,power systems,aerospace,robotics,electric vehicles and industrial general-purpose converters,and their reliability has become a key technology for safe operation of these systems.The faults of three-phase voltage source inverters are mainly divided into open-circuit faults and short-circuit faults.At present,mature hardware circuits are generally used to deal with short-circuit faults,while open-circuit fault diagnosis technology is still under research and development.Aiming at the open-circuit fault of the three-phase voltage-source inverter,this thesis conducts an in-depth study on fault characteristic analysis,fault influence mechanism,fault detection and location technology from multiple aspects,and proposes four different fault diagnosis methods.The main research contents and achievements are as follows:(1)A current vector phase-based method for open-circuit faults of three-phase voltage-source inverters is proposed.The characteristics of current vector phase change of the three-phase voltage-source inverter under health state,single-switch fault,single-phase fault,different-phase same-side double-switch fault,different-phase different-side double-switch fault are analyzed in detail.The change rule of current vector phase under different fault conditions is summarized,which provides the basis for fault diagnosis by using current vector phase.The fault detection algorithm based on the standard deviation of the current vector phase and the fault location algorithm based on the eigenvalue of the current vector phase are designed to achieve fast online detection and location of open-circuit faults using only the current vector phase information.(2)A fault on-line simulation-based method for open-circuits of three-phase voltage-source inverters is proposed.The hybrid dynamic characteristics of power electronic circuits are analyzed,the general modeling steps of power electronic circuits based on hybrid automata are summarized,and the operating mechanism of the three-phase voltage-source inverter-motor system is analyzed.On this basis,the hybrid model of the inverter-motor system is established by using hybrid automata.The current estimator is constructed by using the health hybrid model to detect the faults,effectively eliminate the impact of speed and load mutations and improve the immunity to false alarms.The current vector phase information is used to realize fault judgment to reduce the computational burden and improve the diagnosis reliability.The fault location method based on fault online simulation technology is proposed,which uses the hybrid model to simulate possible faults online,and then implements fault location according to simulated current residuals.The characteristics of the current vector phase are applied to the fault class judgment of the inverter,which can effectively improve the diagnosis speed and reliability.(3)An output line voltage residual analysis-based method for open-circuit faults of three-phase voltage-source invertersis proposed.The mathematical description of output line voltage residuals under single-phase and double-phase open-circuit faults is established,and the characteristics of output line voltage residuals under different open-circuit faults are analyzed.The consistency relationship between the positive and negative polarity of line current residuals and the corresponding polarity of output line voltage residuals is deduced,then the fault detection and location method based on line current residual information is realized by analyzing the output line voltage residuals,which can avoid using additional sensors and improve the speed of fault diagnosis.(4)A convolutional neural network-based method for open-circuit faults of three-phase voltage-source inverters is proposed.The current vector phase is selected as the fault characteristic variable to reduce the fault data dimension,and a data set of 22 open-circuit fault types(including the healthy state)of the inverter is established.By introducing the first layer convolution of large convolution kernel,a ConvNet model suitable for one-dimensional signal feature extraction is constructed,so as to improve the accuracy of network diagnosis.Based on this model,the open-circuit fault classification diagnosis of inverters is realized,and visual analysis and verification are carried out.The ResNet model built by adding the maximum pooling layer to the improved residual block speeds up the network operation and strengthens the feature information screening.Based on this model,the open-circuit fault diagnosis of the inverter under disturbance is realized,and it is compared with the ConvNet and VGGNet models.The analysis shows the feasibility and effectiveness of the ResNet model in practical applications.This thesis explores the characteristics and influence mechanism of multiple open-circuit faults in the three-phase voltage-source inverter from different perspectives,and provides some new solutions and technical support for the power switch status monitoring and open-circuit fault diagnosis of three-phase voltage-source inverters.