Research On Fault Diagnosis Method Of The Bidirectional DC-DC Converter For Ship DC Grid

The bidirectional DC-DC converter is an important energy conversion device for the ship DC grid.It is mainly used for energy control of ship energy storage units.If it fails,it may cause the equipment to operate abnormally,and endanger the safety of the ship’s power system,and then cause a major safety accident.This paper takes the typical faults of marine bidirectional full-bridge DC-DC converters as the main research content,constructs a probabilistic neural network fault diagnosis model,and uses the genetic algorithm and particle swarm optimization to optimize neural network parameters to achieve accurate fault diagnosis.The main research work of this paper is as follows:First,summarize and analyze the relevant literature,choose the converter type according to the actual requirements of the ship,and analyze the working principle of the bidirectional full-bridge DC-DC converter.Analyze the failure mechanism of the key components in the converter to determine the type of failure research.The MATLAB / Simulink software is used to build a simulation model of the ship’s power system and bidirectional full-bridge DC-DC converter,and collect fault parameter samples.Secondly,according to the working principle of the converter,the faults and parameters are compared and analyzed,and the fault parameters are selected.Wavelet packet decomposition is used to extract fault features of fault parameters.Then,the principal component analysis method is used to reduce the dimension of fault feature vectors and combine them to reduce the dimension of fault feature vectors.Because there is no precise mathematical model,the characteristics of nonlinear processing and distributed parallelism of neural networks can be used to implement simple and efficient fault diagnosis.Therefore,the BP neural network fault diagnosis method and the probabilistic neural network fault diagnosis method are designed separately,and MATLAB software is used to carry out fault diagnosis simulation experiments to compare the performance of fault diagnosis.Finally,for the parameter selection problem of probabilistic neural networks,this article uses the genetic algorithm,the particle swarm algorithm,the genetic algorithm and particle swarm optimization（GA-PSO）to optimize the parameter selection,and compares the performance of the three optimization algorithms.The genetic algorithm and particle swarm optimization takes into account the global search capability of the genetic algorithm and the fast convergence of the particle swarm algorithm,and is not easily trapped in local extremes and premature.Finally,the GA-PSO-PNN fault diagnosis method is designed,and the three fault diagnosis methods of BP neural network,probabilistic neural network,and GA-PSO-PNN are compared and analyzed for fault diagnosis performance of bidirectional full-bridge DC-DC converter.