Adaptive MPC-PSO Excitation Control with RNN Forecasting for Enhanced Stability in Renewable-rich Multi-machine Power Systems
This paper proposes an adaptive excitation control architecture integrating Model Predictive Control (MPC), Particle Swarm Optimization (PSO), and short-term renewable energy forecasting using Recurrent Neural Networks (RNN). Tested on the IEEE 9-bus electrical system with a hybrid renewable source (wind and solar), the controller leverages RNN forecasts to anticipate renewable variability, enabling faster voltage settling, improved damping, and enhanced fault recovery. Compared to conventional methods, the proposed approach reduces settling time by 46%, overshoot by 22% and voltage variance by 35%, offering a scalable solution for stability in renewable-rich multi-machine power systems. Major Findings: Compared to traditional techniques, the adaptive MPC-PSO-RNN controller produced 49% faster fault recovery, 22% less overshoot, 20% stronger damping, and 46% faster settling. It also reduced voltage variation by 35%, demonstrating its great resilience for power systems that rely heavily on renewable energy.