Passive Island Detection of Microgrid by Grid Forming Inverter

Abstract: The supply reliability of a local grid will be improved if the local grid is managed as a microgrid, with the ability to operate in island mode when the main grid is temporarily unavailable. This thesis focuses on microgrids powered by inverter-based resources (IBRs), one or several of which need to be equipped with grid forming (GFM) capability for establishing and controlling the voltage and frequency of the microgrid during island operation. To achieve a stable island operation, the detection of an islanding event is crucial. The main aim of the thesis is to investigate how the different parameters of a virtual synchronous machine (VSM)-based GFM controller affect the performance of the existing passive island detection methods (IDMs). The analyses have been carried out using a medium-voltage microgrid with IBRs through theoretical evaluations, simulation studies, and laboratory verification. It is found that to achieve a small non-detection zone (NDZ) for a voltage vector shift (VVS) based passive IDM, a relatively large virtual reactance is desired. In contrast, a large virtual reactance may increase the risk of misdetection during a load switching event. However, a larger value of virtual reactance is preferred during the grid-connected condition to mitigate the impact of grid impedance estimation error on the design of the voltage controller. Furthermore, a fast synchronization loop, corresponding to a low value of virtual inertia constant and damping in the VSM structure, reduces not only the island detection time but also the risk of misdetection when facing a large variation in the grid frequency or angle jump in the grid voltage. Therefore, when tuning the VSM control parameters, it is necessary to weigh the importance of the need for inertia support and damping during grid-connected conditions and the need for a reliable and fast island detection method. Moreover, an analytical expression has also been derived on the relation between the island detection time and the virtual inertial constant and damping constant. The relation is further verified in a laboratory experiment. This makes it convenient to tune the virtual inertial constant and damping constant of the VSM controller for a given required island detection time without resorting to time domain simulation. Besides the foregoing analyses, this thesis has also proposed a PQ-based method for estimating the cycle-to-cycle load angle jump for the VVS-based IDM. The method has shown better accuracy than the typical dq-based method as the PQ-based method is less sensitive to the electrical transients within each electrical cycle and less affected by the harmonics in the grid.