Research On Fault Diagnosis And Health Evaluation In Spacecraft Electrical Power System

With the development of space technology,high reliability,long life and reusability become the main development direction of future spacecraft.Electrical power system(EPS)is the key subsystem of spacecraft.Improving the state monitoring level of EPS is of great significance to enhance the reliability,safety and maintainability of EPS and even the whole spacecraft.As an effective method to improve the level of system monitoring,there are still many problems in feature extraction,information processing,fault diagnosis and health evaluation of spacecraft power system fault diagnosis and health evaluation.Therefore,in order to improve the condition monitoring level of spacecraft comprehensively,and further improve the reliability,safety and maintainability of spacecraft,it is particularly important to carry out the research on fault diagnosis and health evaluation of EPS.In this thesis,the key technologies in fault diagnosis and health assessment of spacecraft power supply system are studied,including component-level fault feature extraction and sensor layout technology,system-level data enhancement and fault diagnosis technology and system-level health assessment technology.A quantitative fault diagnosis method for open-circuit and short-circuit faults in solar arrays using low frequency impedance characteristics is proposed.Compared with the diagnosis method using the external characteristics of solar cells as fault characteristics,the proposed method is not susceptible to the influence of external environmental factors,and the approximate linear fault feature change law is more conducive to the continuous monitoring of the feature change process.On this basis,combined with the fault characteristics of the other key components of the system,the D matrix for power system testability analysis is constructed by using the multi-signal flow graph model.According to the results of testability analysis,the system test is optimized to determine the sensor layout of the system and provide support for the data sources of fault diagnosis and health evaluation of the system.In order to solve the problems of low data processing efficiency and difficult transmission to the ground,this thesis innovatively proposes the dimension reduction technology of spacecraft data,so as to realize multi-sensor data enhancement,and lay the foundation for the follow-up system-level research work.Aiming at the problem of low effective information density of multi-sensor data in spacecraft EPS,combined with the practical application requirements of system monitoring and fault diagnosis,the dimensionality reduction research of multi-sensor data in power supply system fusing fault propagation characteristics is proposed.Based on the fault propagation characteristics of power system,the multivariable weight analysis matrix of power system is established,and the normalized expression of multi-sensor data association based on the fault propagation characteristics of power system is realized.Furthermore,the qualitative method based on knowledge and the quantitative method based on data are combined to eliminate the redundant information unrelated to fault in multisource data and reduce the dimension of multi-sensor data.In the research of fault diagnosis technology,this thesis introduces a deep learning model with autonomous learning ability,and studies the principle,application and training process of the model.By constructing stacked sparse auto encoder(SSAE)model based on deep learning,the multi-sensor data characteristics of EPS are studied,the fault diagnosis model is established,and the parameters affecting the model characteristics are analyzed.The research and experimental results show that,by using the deep learning model,the order of feature extraction is increased.Compared with the neural network model and the hierarchical model,the accuracy of EPS fault diagnosis is obviously improved.System-level health assessment technology is the "outer loop" of power system condition monitoring.Sustainable power system health assessment is conducive to the realization of the whole process monitoring of power system operation from the "macro" perspective.Firstly,this thesis analyses the insufficiency of AHP method in the hierarchical multi-criteria index system,and puts forward a more general non-linear evaluation measurement conversion method.Secondly,in the process of establishing the health evaluation index system of power system,it puts forward the design of integrated power system and the influence of environmental factors,and establishes a hierarchical multidimensional comprehensive health evaluation index system.The research of system-level integrated verification technology includes the design of fault diagnosis and health assessment software and the construction of integrated verification platform.Following the existing spacecraft operation management architecture,this thesis develops a comprehensive verification platform for fault diagnosis and health assessment of spacecraft EPS based on.NET technology without changing the original system perception network and system data management structure.The test results prove that the functions of the fault diagnosis and health evaluation system of spacecraft power supply system can meet the application requirements.On this basis,the application of the fault diagnosis and health evaluation system is explored and studied.