Research On Modeling Simulation And Fault Diagnosis Method Of Self-Drawing Power Supply System

A certain type of domestic self powered power supply system is a kind of field training support equipment equipped to the company level by the army.To ensure the safe and reliable operation of the system is the top priority of equipment management.This system has complex characteristics such as non-linear,strong coupling and so on,and works in a relatively harsh environment for a long time.Some of its components have different degrees of failure,which can not be avoided.If effective measures are not taken in time,it will endanger the life safety of personnel and the property safety of society.But at present,the self powered power supply system has been put into operation for a short time,and there are relatively few fault types and fault data.Therefore,the construction of the simulation platform of the self powered power supply system can provide the corresponding support for the follow-up fault diagnosis method research,and has accumulated technology for the development of its twin system.In this thesis,the self powered power supply system developed independently in China is taken as the research object,and the research is carried out around the construction of system simulation platform and intelligent fault diagnosis method,The main work is as follows:1)In order to build a reliable simulation platform of self powered power supply system,the establishment of accurate mathematical model is the basis.Therefore,this thesis firstly analyzes the working principle and structure of self powered power supply system,and divides it into asynchronous generator,generator control system and inverter system.Then,combining with the operation mechanism and the first mock exam strategy of each module,a mathematical model of the self powered power supply system is established.2)In order to verify the accuracy of the power supply system,based on the mathematical model of each module,with the help of Matlab/Simulink simulation platform,through the setting of component parameters,the simulation modules are connected to build the simulation platform of the power supply system.Then,according to the test outline and test data of the power supply system,the running data error between the simulation system and the real system in parking and driving state is analyzed quantitatively,and the reliability of the simulation platform is verified.3)In order to better identify the fault of self powered power supply system,considering the good performance of multi model for multi condition fault diagnosis,this thesis proposes a fault diagnosis method based on k-fold optimization voting ensemble learning.Combined with the optimized fault feature data,the k-fold cross validation method is used to analyze the data of logistic regression(LR),linear discriminant analysis(LDA),k-nearest neighbor(KNN),classification and regression tree(CART),support vector machines(SVM),and the k-nearest neighbor(KNN),Five kinds of primary learners,such as support vector machine(SVM),are trained,and the diagnostic accuracy and fitting degree are taken into account.K-fold optimization and k-fold weighted integration strategies are used to further optimize the primary learners,and finally voting integration is carried out.The results show that the proposed method has higher fault diagnosis accuracy than single diagnosis model and k-fold integration algorithm.4)In order to further effectively fuse the diagnosis results of heterogeneous learners,a fault diagnosis method based on Improved D-S evidence theory is proposed considering the unbalanced fault diagnosis results of different primary learners in decision fusion.Based on five primary learners(LR,LDA,KNN,CART,SVM)optimized in the previous chapter.The basic probability assignment value of the recognition framework is constructed by combining the fault diagnosis accuracy of the optimization set and the diagnosis category of the verification set in different states of each learner,and uses weight coefficient to modify the basic probability assignment value,which effectively improves the imbalance of the fusion evidence and improves the accuracy of the diagnosis decision.