The Research On Fault Detection And Diagnosis Of Inverter Module Of Ship Power System

With the development of new energy ships,inverter will play an increasingly important role in the electric system of new energy ships.The inverter module was the critical equipment of the ship's electric propulsion plant.However the inverter module was the weak link of the power system,and also the inverter module's power switching components which control on-off of the circuits were components with high failure rate.Therefore,it had great significance to study the fault diagnosis of inverter module for the safety of ship and its equipment.Based on summarizing the development status of inverter fault detection and diagnosis at home and abroad,meanwhile as the ship inverter module existed noise and had few eigenvalues,this paper proposed a fault diagnosis method based on Unscented kalman filter and Support vector machine according to the characteristics of ship inverter module.In this paper,the simulation model of the inverter system was established according to the typical inverter system on ships,in this model the PWM technology was the method of controlling and the asynchronous motor was load.With this simulation model,this paper analyzed the current under different failure of power switching components.With the simulation experiment,the conclusion that the current can be used for fault diagnosis was obtained.As far as the frequency measurement is concerned,the complex version Kalman filter is more efficient than real Kalman filter,the complex domain current model of inverter system was established.This paper used this model to compare the convergence performance of Extended Kalman filter algorithm and Unscented Kalman filter algorithm under different SNR,it was concluded that Unscented Kalman filter algorithm was better.Then the frequency?fundamental wave and partial harmonics of inverter system were estimated by complex domain Unscented kalman filter,simulation results showed that those estimated parameters could reflect different fault feature with good performance.Therefore,those estimated parameters could be used as eigenvalues for fault separation of inverter power switching components.Support vector machine have robustness and can classify with few samples.But the traditional Support vector machine need many classification times and had low classification accuracy.In order to improve the classification efficiency of Support vector machine,this paperdesigned a fault diagnosis model of multi-Support vector machines.In order to improve the classification accuracy of Support vector machine in this model,this paper normalized the training and testing data and selected appropriate Kernel function.Simulation results showed that model of multi-Support vector machines could separate the faults of inverter power switching components effectively.