Research On Open-circuit Fault Diagnosis Of Inverters In Electric Drive Systems Based On Data-driven And Model Methods

The motor drive system,as an important power source,is widely used in production and daily life.For permanent magnet synchronous machine(PMSM)drive systems,three-phase voltage source inverters(VSIs)are important energy conversion devices.Ensuring their reliability and safety is a prerequisite for ensuring the stable operation of the motor drive system.Therefore,fault diagnosis technology is of great significance.This article will propose three new methods for diagnosing three-phase VSI power transistor open circuit faults in electric drive systems from the perspective of models and data,and build an experimental platform based on d SPACE and RT＿lab to verify the effectiveness of the method through simulation and experiments.The research content of this article mainly includes the following three aspects:(1)To address the issue of traditional fault diagnosis methods being susceptible to changes in load and speed,this paper proposes a fault diagnosis method that combines near-zero current and line voltage residual.Firstly,after an open-circuit fault occurs in the inverter,a near-zero current will occur,and the position of the faulty phase will be determined based on the difference in near-zero current status under different faults.Then,the residual of the line-voltage is obtained through a magnetic flux observer and a Mixed Logical Dynamic Model(MLDM),and the position of the faulty switch is determined based on its polarity change.The time required for this method to locate a specific faulty tube is between 16%-90% of the fundamental current cycle.(2)In response to the difficulty of threshold setting in model-based methods,this paper proposes an FFT-K-means++ fault diagnosis method.Using Fast Fourier transform(FFT)to extract the Fourier amplitude of three-phase current,the algorithm selects Fourier amplitudes with larger feature weights to simplify the feature quantity.Using cluster analysis of faults,determine the cluster center of each type of fault,and determine the fault location by calculating the distance between the experimental current Fourier amplitude and the cluster center point.(3)In response to the problem of high computational complexity and the need for a large amount of data for training in data-based methods,this paper proposes a fault diagnosis method that combines voltage models and data-driven approaches.Firstly,the residual of the line-voltage is obtained through a hybrid logic dynamic model and a magnetic flux observer,and the fault type is classified based on its changing characteristics.Then,for each type of fault,Principal Component Analysis(PCA)is used to convert the residual data of line voltage into two-dimensional vectors.Finally,Support Vector Machine(SVM)is used to partition different faults,achieving fault diagnosis and reducing the computational complexity of the algorithm.