Fault Diagnosis And Identification Of Wave Energy Converter System

Throughout the global energy trends,clean and low carbon has become an important goal of the future energy system.Developing and increasing the penetration rate of renewable energy is promising to achieve this goal and relieve the pressure of energy resources and environment.Wave energy has become one of the most promising renewable energy sources because of its abundant reserves,high power and long generating time.Power converter is the core of wave energy generation system which features flexible transmission and high conversion efficiency,and it is the medium to realize the two-direction decoupling of the input and output of electric energy.However,power converter is prone to failures due to harsh environment and complex structure.The ratio of power device faults to those of the converter is up to 34%,which is the most vulnerable part in the converter system.Therefore,it is of great significance to further study the high reliability diagnosis method for power device faults to improve the economy and reliability of the whole wave energy generation system.This paper takes wave power generation back-to-back converters as the research object,focusing on the open circuit fault diagnosis of the power devices.In order to understand the fault mechanism,the mathematical models of two-level backto-back converter system and three-level back-to-back converter system are established respectively.In this paper,model predictive control is used as the control strategy of converter systems.By analyzing the relationship between three-phase current flow path and fault current waveform,this paper selects the relative standard deviation and mean value of three-phase current as fault characteristics to promote the sensitivity of fault characteristics.Simulation results show that the fault characteristics can clearly indicate the different tube faults of power devices.On the basis of model establishment and fault feature extraction,a support vector machine(SVM)model is established,and its working principle and training process are described in detail.Firstly,the relative standard deviation and mean value of three-phase current under different fault types are selected as input data.Then,the SVM model of back-to-back converter system was identified and trained.Finally,the above SVM model achieved high identification performance.In the simulation verification,the accuracy of fault type identification of four kernel functions is compared and analyzed.The simulation test results show that for two-level back-to-back converter system,all the four functions have achieved high identification accuracy.For three level back-to-back converter system,Gaussian kernel function can achieve the highest accuracy.In addition,a model based on artificial Neural Network(ANN)is established,and its working principle and training process are described in detail.In this method,the relative standard deviation and mean value of three-phase current under different fault types are also used as input data.The ANN model achieved high identification performance.In the simulation verification,the influence of the number of neurons in hidden layers on the accuracy of fault classification is compared and analyzed,so as to select the optimal number of neurons which can make the model with high accuracy and fewer iterations.The simulation test results show that the identification model can quickly and accurately identify different fault types of the converter,and has the advantages of strong robustness and low model dependence.