A novel LF-TLBO-based optimisation scheme for islanding detection in microgrids

In the framework of contemporary power systems, a distributed generation (DG) system has benefits, but it also presents several operational challenges. In a networked distributed generation system, islanding is one such problem. Because of the negative consequences, an islanding event should be distinguished from other events, like transients within the minimum time. A multi-level adaptive neuro-fuzzy inference system (ANFIS) is being developed in this work to effectively detect islanding. To train the ANFIS model, a novel hybrid scheme based on Lévy flights and teaching–learning-based optimiser is suggested. The performance of the developed algorithm is evaluated using the IEEE CEC-C06 and other traditional benchmark functions. The ANFIS model’s classification regime is significantly improved by the optimisation algorithm. In the test system, renewable energy sources are used to power a voltage source converter unit in a network-forming mode via an energy storage medium. Based on measured frequency, RMS voltage and current, active and reactive power, voltage, and current THD at the point of common coupling (PCC), the trained ANFIS controller deduces the islanding detection command to the circuit breaker. The plan is verified following the UL1741 standard for islanding prevention, yielding notable outcomes with an average detection time of 0.04s and an accuracy of 89.3%.