Abstract:
To address the global energy demand, alternative energy sources are used more frequently daily. Grid connected inverter (GCI) penetration into power networks has accelerated and is one of the essential components of the future smart grid. If a portion of the system is islanded, GCI must efficiently maintain control of the islanded power system without compromising power quality. If not, GCI must be disconnected as soon as they are no longer negatively affecting the islanded grid. To disconnect the GCI or alter the GCI's control approach, the islanding should therefore be discovered as soon as possible.Islanding will damage the protection and reliability of the power system; it must be detected by the control system within a particular time. Islanding detection is a crucial function that needs to be handled in the distributed generation system. In the occurrence of unintended islanding in an inverter-based distributed generation system, an efficient and successful islanding detection mechanism is necessary to maintain the smooth operation of loads in the system.While active detection methods have catastrophic impacts on power quality, traditional islanding detection methods, such as the passive ones, have shortcomings in the form of the availability of sizable non-detection zones.An innovative islanding detection approach that is based on a machine learning algorithm has been developed here to address these shortcomings.A logistic regression algorithm is implemented for the islanding monitoring. Firstly, an inverter-based distributed power generation system model is defined in MATLAB/ SIMULINK. The model covers different possible island conditions, minimum observable parameters, and multiple inverter-based distributed power supplies and then applies. The data extracted from the above island conditions are trained in the proposed data-driven algorithm. Finally, the proposed trained model is applied to detect the islanding status. The outcomes validate the method's effectiveness.According to the simulation results, the method that has been provided may effectively and quickly detect islanding under all conditions for the selected system model.
