Research On New Parameter Identification Methods And Failure Prediction Algorithms Of Power Electronic Circuits

With the proposal of the concept of most electric aircraft and all electric aircraft, the aero powersystem is greatly impelled. The power electronic circuits change the three-phase alternating-currentsupply into other current supplies which can support the use of other equipments on the plane,becoming an important segment of the aero power system. Thus, ensuring the realibility of the powerelectronic circuits, predicting and estimating the residual life of it are the prompt demanding problemswhich share great engineering application significance.This paper mainly focuses on the research of failure predition methods of power electroniccircuits, the research content contains the following three aspects,(1) The failure mechanism and failure model of key components of power electronic circuitssuch as capacitor, power MOSFET, power diode, lnductance are studied and the failure characteristicparameters which can reflect the failure mode of the components are analysed and the thresholds ofthe parameters are comfirmed.Through analysing the changing of the output parameters of the circuitsduring the degradation of the components and circuits, the off-set rate of the output voltage is definedas the characteristic parameter which reflects the degration rate of the circuit.(2) Two parameter identification methods are researched, that is, the method based on RBFdistance and the method based on BP neural network. The principle and identification steps of themethods are presented. The physical experiment based on Boost circuit are done in different workingconditions and the factors which impact the identification precision are analysed.(3)The failure prediction methods of power electronic circuits are studied. The open-loop Boostcircuit is taken as an example for the failure prediction and residual life estimation using LS, LS-SVMand grey system algorithms with the degration rules of the characteristic parameters of the circuits.The failure prediction methods of close-loop Boost circuits are pilotly studied, combing thedegradation rules of capacitors, LS and LS-SVM algorithms are used to failure prediction and residuallife estimation of the capacitors. The physical experiment shows the efficiency and accuracy of themethods used in the paper. Accelerated test is conducted on the key components of power electroniccircuits. Through the research, the degradation curves of the components and circuits are obtainedwhich provide a basis for the physical validation and diagositic assessment of parameter identificationand failure prediction methods of power electronic circuits.