Opportunities for enhancing the security and reliability of power supply as well as the utilization of renewable and efficient energy sources have generated major interest in Microgrids. A microgrid typically consists of interconnected loads, distributed generators (DG) and energy storages, and should be able to operate in parallel with the utility grid or as a power-island. The main focus of this thesis is on the transition between parallel and islanded operation of a microgrid. A literature review on existing microgrids was carried out. Based on the survey, a microgrid test system was implemented on PSCAD/EMTDC simulation program. The microgrid controls essential for the study and a load shedding scheme were designed and implemented. When the microgrid changes from parallel to islanded operation, its controls need to be changed. It was found that delays in microgrid control mode transition can impact the amount of load need to be shed to preserve the frequency stability and the power quality of the islanded microgrid. The importance of fast detection of islanding was therefore highlighted. The IEEE standard 1547.4-2011 recommends application of the existing DG synchronization criteria for microgrid synchronization. The adequacy of these criteria for synchronization of a microgrid with highly unbalanced loading was investigated. It was found that the required criteria can be met with the support of switched capacitors for voltage balancing, and a circuit breaker supervised by a synchro-check relay is sufficient to successfully reconnect an islanded microgrid back to the utility. In order to meet the requirement for fast detection of islanding of microgrids, new islanding detection technique was proposed. In the proposed scheme, Discrete Wavelet Transform was used to extract features from transient current and voltage signals, and then a Decision Tree classifier was employed to distinguish islanding events from other transients. Simulation based tests asserted that the proposed technique has a high reliability and fast response compared to most existing islanding detection methods. Also, the detection time of the proposed method was invariant with the power imbalance in the microgrid, and gave a zero non-detection-zone with any type of generator.

Opportunities for enhancing the security and reliability of power supply as well as the utilization of renewable and efficient energy sources have generated major interest in Microgrids. A microgrid typically consists of interconnected loads, distributed generators (DG) and energy storages, and should be able to operate in parallel with the utility grid or as a power-island. The main focus of this thesis is on the transition between parallel and islanded operation of a microgrid. A literature review on existing microgrids was carried out. Based on the survey, a microgrid test system was implemented on PSCAD/EMTDC simulation program. The microgrid controls essential for the study and a load shedding scheme were designed and implemented. When the microgrid changes from parallel to islanded operation, its controls need to be changed. It was found that delays in microgrid control mode transition can impact the amount of load need to be shed to preserve the frequency stability and the power quality of the islanded microgrid. The importance of fast detection of islanding was therefore highlighted. The IEEE standard 1547.4-2011 recommends application of the existing DG synchronization criteria for microgrid synchronization. The adequacy of these criteria for synchronization of a microgrid with highly unbalanced loading was investigated. It was found that the required criteria can be met with the support of switched capacitors for voltage balancing, and a circuit breaker supervised by a synchro-check relay is sufficient to successfully reconnect an islanded microgrid back to the utility. In order to meet the requirement for fast detection of islanding of microgrids, new islanding detection technique was proposed. In the proposed scheme, Discrete Wavelet Transform was used to extract features from transient current and voltage signals, and then a Decision Tree classifier was employed to distinguish islanding events from other transients. Simulation based tests asserted that the proposed technique has a high reliability and fast response compared to most existing islanding detection methods. Also, the detection time of the proposed method was invariant with the power imbalance in the microgrid, and gave a zero non-detection-zone with any type of generator.