Health Diagnosis Strategy For Main Circuit Of Three-Level Four-Quadrant Inverter Based On Fuzzy Neural Network And D-S Evidence

With the vigorous development of controllable power electronic devices and related technologies,high-power multi-stage variable frequency speed control systems of various topologies have been widely used in the industry,reducing unit energy consumption and achieving significant energy saving effects.The frequency converter is a very complicated electronic system.The open circuit fault of the power switch tube is the fault with the highest probability of occurrence in the frequency converter.This paper uses the advantages of multi-sensor information fusion technology to diagnose the open circuit fault of the power tube of the inverter and break it.The limitations of a single sensor improve the accuracy and accuracy of fault diagnosis.This paper takes ABB's ACS6000 frequency converter as the research object.The frequency converter is a three-level four-quadrant frequency converter.The main circuit of the frequency converter is divided into three parts: rectifier circuit,DC circuit and inverter circuit.The three parts of the circuit are detailed.The basic working principle and various output waveforms of each part of the power tube open circuit fault are analyzed,and the fault features are extracted from the output waveform.Three sub-fuzzy neural networks are established for the three-part circuit,and the fault diagnosis of each part of the circuit is first carried out,which is also a preliminary partial diagnosis of the health status of the inverter.D-S evidence theory has a good ability to process uncertain information.According to the uncertainty in fault diagnosis,the D-S evidence theory is used to fuse the diagnosis results of three sub-fuzzy neural networks of rectifier circuit,DC circuit and inverter circuit to obtain the health status of the whole inverter.However,due to the unsatisfactory effect of D-S evidence theory on strong conflict data,this paper uses D-S evidence theory based on similarity coefficient to fuse strong conflict data and achieves good results.The fault characteristic signal and the output signal are nonlinear,the fuzzy neural network has good nonlinear adaptive ability,and it is easy to construct the basic probability function;while the D-S evidence theory constructs the basic probability function is difficult and has great subjectivity,but the decision comparison accurate.Fuzzy neural network and D-S evidence theory have the advantages of complementary advantages.This paper combines these two evaluation methods,and establishes the main circuit health diagnosis framework of frequency converter based on fuzzy neural network and D-S evidence theory,using three sub-fuzzy neural networks.The frequency converter performs local diagnosis,and then globally integrates through D-S evidence theory to obtain the health status of the whole frequency conversion.Finally,an example is given to verify that the method of comprehensive evaluation has higher accuracy than the method of using one method alone.