Neural Networks based Adaptive Predictive Control of Flywheel Energy Storage for Hybrid Power Systems

This paper describes a non-linear step-ahead neuro-adaptive predictive control (SNAPC) approach for a flywheel energy storage system (FESS) based on neural networks for frequency regulation of a hybrid multi-area power system. The focus is on constraining the low-rated FESS energy levels while ensuring frequency regulation in the power system. This involves using neural networks to model and predict the system’s behavior and adaptively adjust control actions based on the predicted performance. A control-oriented second-order structure is developed, which interprets the hardware constraints of the flywheel rotor speed and translates them into transformed power constraints. This ensures that the FES operates within the defined energy level boundaries. The time-domain simulations conducted on the three-area power system, which includes the incoming portions of wind energy, serve to verify the effectiveness of the developed control framework.