Xiangyu Ding with adviser A. Domínguez-García
With the ever growing use of power electronics in the power grid, the issue of their failure is becoming increasingly important. In this research, a study is presented on an observer-based piece-wise linear fault detection filter for power electronics systems health monitoring. Using a state-space modeling approach, we mathematically derived piece-wise linear fault detection filters of three-phase inverters and static synchronous compensators (STATCOMs), then studied the effectiveness of these filters at detecting and isolating the system component. The detection filters run side by side with the actual system. When there are no system faults, the filter output is zero; when faults occur, the filter generates unique signatures that allow easy fault identification. The key advantages of this method include fast detection of nearly all possible component faults and the ability to capture slow degradations in individual components. Figure 25 shows the simulated output of a sensor failure in the inverter. Only the filter residual associated with the faulted phase changed after a fault occurred in the system (In this case green is phase A, red is phase B, and blue is phase C). This signature is unique to a sensor fault and very simple to identify in an inverter protection scheme.
In our earlier work, we have simulated the filter behavior under various faults using a combination of software including MATLABSIMULINKPLECS. More recently, the research focus has been to experimentally implement the fault detection filters in hardware. In a joint research effort with a research lab at MIT that specializes in real-time simulation, we were able to demonstrate the effectiveness of this class of fault detection filters for three-phase inverters. Detailed results can be found in a journal paper, “Fault Detection and Isolation Filters for Three-Phase Power Electronics Systems,” by X. Ding, J. Poon, Ivan Celanovic, and A. D. Domínguez-García, to appear in IEEE Transactions on Circuits and Systems. This research was supported by the National Science Foundation.
This research is funded by the Grainger Center for Electric Machinery and Electromechanics.