Failure Prediction Research Of IGBT Module For Wind Power Converter Based On Grey Algorithm

Wind power accounts for a large proportion in the field of new energy power generation in China.In wind power generation systems,converters are very important devices,and their reliability is directly related to the safe operation of the grid system.A large part of the faults in wind power systems are caused by converter faults.It is of great significance to study and analyze the fault mechanism of wind power converters,and to perform condition monitoring and prediction on them.Achieving intelligent maintenance is also important for protecting the converter.The power device has the highest failure rate in the converter system.Its failure will affect the operating status of the entire converter system and can be used as a monitoring object to reflect the working status of the converter.The IGBT module in the power device is the most expensive core device inside the converter.It has the advantages of high operating frequency,fast switching speed,high input impedance,large current carrying density,and low saturation voltage.Its failure can most affect the safe operation of the converter.The failure of IGBT module is the focus of research on converter failure,and it is also a very important direction in the research of wind power converter.Research on IGBT module failure prediction of wind power converters based on grey algorithm rolling prediction method can provide new ideas and new methods for predicting the health status of key components of wind power converters.In this way,we can plan ahead,reduce the cost of wind power generation,and improve the safety of system operation.Monitoring and predicting the condition of key components of wind power converters is conducive to improving the stability of wind power generation.Based on the existing research,the following studies have been carried out:The research status of converter condition monitoring and prediction is summarized and analyzed,and the IGBT modules of common plastic converters are mainly studied.Combined with the Infineon company’s IGBT module model FZ2400R17HP4＿B29,the theoretical analysis of the state parameters of the IGBT module was carried out.This provides a theoretical basis for analyzing the failure mechanism and characteristic parameters of IGBT modules.By analyzing the physical structure of the IGBT module and combining the commonly used power simulation software Saber,the failure mechanism of the IGBT module is studied.Research shows that bond wire damage is the main factor that causes IGBT module failure.By analyzing the related research on bond wire damage,it is pointed out that the bond resistance R_J is a characteristic quantity that characterizes the comprehensive state of bond wire damage,and has a complex correlation with the degree of bond wire damage.Due to the irreversible bond line damage process,R_Jhas a definite change trend.The relationship between the R_J increment and the number of bond line shedding can be fitted to a quasi-exponential model,and there is a failure threshold.Therefore,the degree of damage to the bonding wire can be predicted by predicting the R_J change trend.The factors that affect the bond line damage are complex and diverse,with a certain degree of randomness.Each factor acts on the bond line.After various factors act on the bonding wire,the final result is that the bonding wire is damaged,and the R_J is continuously increased.Therefore,the bond wire damage process has the characteristics of a gray system,which is consistent with the applicable conditions of the gray prediction algorithm.Based on this research,combined with the MATALB software,a gray prediction algorithm program was written.Then the compound Simpson quadrature formula is used to improve the original algorithm and improve the prediction accuracy.A grey rolling prediction method is proposed.In addition,in order to predict the aging state of the bonding wire in the same time period as the sampling interval in the future,combined with the equivalent resistance of the bonding wires,sampling and screening the data matrix at equal intervals.The analysis shows that the proposed method is effective and continuous.