RegPSO-ANN-Based Optimum Adaptive Load Shedding Technique

Utility Grid outages may affect the industrial power system load-generation balance and cause rapid frequency decays particularly due to low inertia and small spinning reserve. Frequency stability is an essential perspective for the invulnerable operation of islanded industrial power systems with in-plant generation. Multi-machine system frequency response (MMSFR) model with the dynamic of governors, overall inertia of the system (generators and loads), and load damping is considered to find the approximate value of load shedding (LS) amount. The frequency variation of the islanded system is evaluated for various operating scenarios with a calculated approximate LS amount to determine the optimum amount of LS. In this paper, an artificial neural network (ANN)-based optimum and adaptive LS technique have been presented. The total in-plant generation, spinning reserve, total power import, total demand, and frequency decay rate have been selected as the input neurons of the ANN. The regrouping particle swarm optimization (RegPSO) algorithm has been adopted to obtain more accurate and faster training of the ANN. The oil refinery power distribution system with in-plant generation is considered to evaluate the performance of the proposed RegPSO-ANN load shedding algorithm. The performance of the proposed scheme is evaluated in comparison with the under-frequency relay-based conventional load shedding scheme and Levenberg-Marquardt back-propagation ANN-based LS scheme.