Neural network aided fractional-order sliding mode controller for frequency regulation of nonlinear power systems

Abstract

In this work, the performance of a fractional-order integral sliding mode controller (FOISMC) has been assessed for frequency regulation of interconnected power systems considering inherent nonlinearities, such as generation rate constraint and governor dead-band. The performance of the proposed controller has been measured on two- and three-area interconnected power systems integrated with a wind turbine generator. A Chebyshev neural network (NN)-based estimator is designed to estimate the lumped system uncertainties, including nonlinearities, unknown external disturbances, and parameter uncertainty. Afterwards, an improved FOISMC is developed, augmenting the estimated output of the NN-based estimator to cope with system disturbances effectively. To corroborate the potential benefits, the results obtained with NN-aided FOISMC (NN-FOISMC) are compared with the outputs of FOISMC and results reported in the literature. The simulation study confirms the superiority of NN-FOISMC over its other counterparts in terms of damping of power-frequency oscillations, weaker chattering, and a high degree of robustness.