Fault Diagnosis Of Three-level Inverter Based On EEMD Sample Entropy And Improved SVM

Inverter is one of the common power converters in power electronic technology,which is widely used in many production processes.The failure of any component in the inverter will affect the performance of the inverter,even damage the equipment and affect the stability of the entire system.Therefore,it is necessary to study the problem of inverter fault diagnosis.The neutral point clamped(NPC)three-level inverter has the advantages of less output harmonic content and less electromagnetic interference,which is the most commonly used inverter topology.This article takes NPC three-level inverter as the research object.The main research contents are as follows:(1)The topology,working principle and fault reason of NPC three-level inverter are analyzed,and the fault types are divided into three categories and 11 types.This paper studies the characteristics of circuit signals in different types of faults,and determines the type and acquisition position of fault signal.The simulation model of NPC three-level inverter is built in Simulink to simulate the working state of different faults.The voltage waveforms of upper,middle and lower bridge arms under 11 kinds of faults are collected to verify whether the selected fault signals can completely distinguish all faults.(2)To solve the problem of invalid eigenmode components and excessive dimensionality of eigenvectors generated by ensemble empirical mode decomposition,according to the fault signal characteristics of NPC three-level inverter,a method of fault feature extraction based on collective empirical mode decomposition and sample entropy is proposed by calculating correlation coefficients and sample entropy.In order to solve the problem of slow processing speed of SVM when processing large amount of data,the adaptive delta(Ada Delta)method is used to optimize the calculation efficiency of the algorithm.(3)The experiment platform of NPC three-level inverter fault diagnosis based on DSP and FPGA is built,and the hardware circuit and software program are designed respectively.By applying pulse signals to control the opening and closing of the power tube,the fault waveform and data are collected.Using the PC program to complete the feature vector extraction,fault diagnosis model training and testing.Compared with other fault diagnosis methods in time and accuracy,and the comparison results verify the superiority of this scheme.