Fault Diagnosis Of Three-level Inverter Based On Joint Optimized Echo State Network

Inverter as a common power converter is widely used in rail transportation,AC/DC transmission and many other fields.Compared with the traditional two-level inverter,the multi-level inverter has advantages of more output levels and stable performance.But its topological structure is complex and there are many switching devices.Once the fault occurs,it will affect the normal operation of the system and even cause damage to the equipment.Therefore,it is necessary to timely fault diagnosis of inverter.This paper mainly studies the fault diagnosis of ANPC three-level inverter Active Neutral-Point Clamped(ANPC).The research contents are as follows:Firstly,the topological structure and working principle of ANPC three-level inverter are introduced,and the switching state and loss are discussed in combination with its commutation path.Then,12 types of faults that are most likely to occur,including the normal operating state,are analyzed,and the signal characteristics of different types of faults are studied.The ANPC three-level inverter model was built through MATLAB/Simulink simulation to simulate the working state of different faults.Secondly,in view of the complexity of calculation caused by too much fault information,wavelet packet analysis is used to extract the fault feature vector of bridge arm voltage signal,save the effective features of fault signal,and then linear discriminant analysis is used to reduce the dimension of fault feature vector.Secondly,aiming at the problem that traditional particle swarm optimization(PSO)is easy to fall into local optimum,a nonlinear adaptive method is proposed to improve PSO.Aiming at the defect of ill conditioned solution caused by the random generation of reserve pool parameters in traditional echo state network,an improved particle swarm optimization algorithm is used to optimize reserve pool parameters,and a joint optimized echo state network(JO-ESN)fault classification method for ANPC three-level inverter is proposed.Finally,an experimental platform based on DSP and FPGA controller is built to verify the fault classification method of ANPC three-level inverter.The feature vectors were divided into training data and test data.The feature vector is divided into training data and test data Experimental results show that,compared with other fault diagnosis methods,the proposed algorithm has more advantages in time and accuracy,and is suitable for fault diagnosis of three-level ANPC inverters.