Owing to the rapid development of power electronic technology, the market of power electronic equipment has expanded continuously, and it has applied widely to many fields such as power systems, home appliances, transportation, navigation, commerce and industrial. With the stricter and stricter demands on systems' safety and reliability, the fault diagnosis technology has been emphasized increasingly. Acted as power supply or motor driver, the power electronic equipment has a great impact on the systems' reliability. So the studies on the fault diagnosis for power electronics circuit (PEC) have significant theory value and practical value.The hybrid system theory can be applied to study PEC fault diagnosis by viewed PEC as a type of typical hybrid system. Based on the theory, this dissertation investigates the hybrid model, fault diagnosis, topology identification and wavelet transform application for PEC. The main contributions of this dissertation are as follows.(1) An uniform hybrid model for PEC.An uniform hybrid model for PEC is founded on the basis of PEC essential hybrid characteristic. And some definitions which are correlative with the model such as basic circuit event, control transition, condition transition are made. In the hybrid model, the subsystem' parameter and the transition rules between two subsystems are the only difference among PEC. So the model can be used to investigate the universal fault diagnosis for PEC.(2) An universal fault diagnosis principle for PEC.The PEC's fundamental theory and control strategies are analyzed with the hybrid model. Then an universal fault diagnosis principle for PEC, which all PEC's structural failure behaved as error circuit event transition, is obtained. So the fault diagnosis is transformed into circuit topology identification for making clear the failure range and type. This principle can meet the demand of structural or parameter fault diagnosis simultaneously and the demand of universality. Finally, some typical PEC are analyzed to obtain their event transition rules and fault diagnosis rules respectively.(3) Three topology identification methods for PEC.The traditional circuit topology analysis method is introduced to PEC by equivalent transforming the PEC to noswitch circuits. Then some conception such as currentidenticaltopology, voltageidentical topology, topology linear relative, topology vector, leastamount topology vector are defined, and some theorem such as current variables topology linear relative theorem, current loop irrelative theorem, the leastamount topology vector selection theorem are obtained. With these theorems, one topology identification method is gained by current variable topology relation and the switchbranch current state. And another two identification methods are obtained by current or voltage variables which vary differently in each topology. Finally, some typical PEC are analyzed with the three topology identification methods.(4) An event identification means by residual analysis.A PEC event identification means by residual generators is put forwards. Then the feasibility is proved, and the design method is deduced. To decrease the calculation amount, a priority is set when these generators are executed by software. The fundamental theory and its existence of robust residual generator is studied from frequencydomain and timedomain to improve the generators' robustness. Finally, the robust residual generators are built using the unknown input observer, and the observers' parameters design method is deduced.(5) A fault diagnosis technique by wavelet transformation.The filtering problem for PEC fault signal is discussed. Then a conclusion, which the wavelet transformation acted as an element of the residual generator has better effect than acted as residual filter, is drawn. By summarizing the known wavelet technique applied to fault diagnosis, another conclusion, which the traditional wavelet technique can not be used for PEC fault diagnosis, is brought out. Finally, with the singularity analysis of PEC signal, a novel fault diagnosis technique for PEC is obtained.
