An Adaptive Neural Network-Driven PID Control Framework for Load Frequency Regulation in Renewable-Integrated Power System

A load frequency controller is essential in maintaining frequency stability in power systems amid sudden load changes. This study presents an adaptive control strategy using an artificial neural network-tuned proportional–integral–derivative (ANN–PID) controller for a single-area hybrid power system integrating photovoltaic, wind, and thermal units. This setup reflects a realistic environment with inherent uncertainties. Initially, PID parameters are optimized via particle swarm optimization (PSO) for its fast convergence and effective performance. These optimized values serve as training targets for an ANN that dynamically tunes PID gains in real time to adapt to system variations. Regression analysis confirms the ANN model’s accuracy, achieving an R-value of 0.997, indicating high fidelity in capturing system behavior and reproducing optimal control actions. Simulations under six key scenarios, fixed and variable load disturbances, parameter uncertainties, nonlinearities, renewable generation variability, and stability assessment demonstrate the ANN–PID controller’s superiority over fixed-gain (PSO–PID) counterparts. The adaptive controller significantly improves frequency regulation, especially under variable load conditions, achieving faster settling times and reduced deviations. Quantitatively, it achieves up to 99.3% reduction in undershoot, 18.8% in overshoot, and 83% improvement in settling time. These consistent results across scenarios highlight the method’s robustness and adaptability. The findings underscore the practical applicability of the proposed adaptive controller in renewable-integrated power systems, especially under high uncertainty and stringent stability demands.