Fault diagnosis for grid-connected power converters

Abstract

The reliability and functionality of power electronic converter systems (PECS) are of critical importance for industrial, commercial, aerospace, and other applications. One of the key requirements to ensure reliable functions of PECS is the behavior of the PECS during fault conditions. Characterizing the behavior of PECS during the fault conditions can provide a perspective for improving the design, protection and fault tolerant control. In general, faults in switching elements of PECS can be classified as Short Circuit (S-C) faults, Open Circuit (O-C) faults, and degradation faults. S-C faults in most cases cause an overcurrent condition that is readily detected and acted upon by standard protection systems, such as over-current, under-voltage or over-voltage protection. However, the degradation faults, as well as O-C faults often do not produce high currents that can trigger fault protection; rather they cause system malfunction or performance degradation. Since the standard protection system may not detect these fault types, their diagnoses become critical for PECS. The main objective of this dissertation is to investigate and develop new fault diagnostic algorithms for a typical single-phase grid-connected power converter with its DC capacitor banks, as well as to identify the unbalance input voltage to the converter. The power converter in this research consists of three main subsystems: the three-phase uncontrolled rectifier, the boost chopper, and the single-phase inverter circuits. The diagnostic algorithms have been investigated, designed, and implemented as follows: detection of unbalance three-phase input voltage, detection of O-C faults in the rectifier circuit, detection of O-C in the boost chopper circuit, detection of O-C in the inverter circuit, and detection of O-C fault and capacitor aging in the DC capacitor banks. The constraint for designing the desired fault diagnostics algorithm is the existing number of sensors in the PECS under study. This constraint is incorporated in this work to allow maximum integration of the developed diagnostics with low and medium size PECS.