Adaptive proportional integral control using neural networks for secondary frequency regulation in microgrids

Abstract

Microgrids with high renewable energy integration face a challenge in maintaining frequency stability due to the reduced inertia of inverter-based generation and the intermittent nature of these sources. Although primary frequency regulation using virtual synchronous generator (VSG) strategies can provide fast support, it cannot fully bring the system frequency back to its nominal value. This limitation highlights the importance of secondary frequency regulation, which is implemented using proportional integral (PI) controllers. However, fixed parameter PI regulators often fail to adapt effectively to varying loads and fluctuating renewable generation. This paper proposes an adaptive secondary control strategy for microgrids that combines offline optimization with real time learning. Grey wolf optimization (GWO) is first applied offline to determine the optimal PI gains for multiple disturbance scenarios. These datasets are then used to train an artificial neural network (ANN), which updates the PI parameters in real time to achieve adaptive performance. The proposed control is implemented in a hybrid microgrid with a diesel generator, a permanent magnet synchronous generator (PMSG) wind turbine for primary support and a fuel cell for secondary regulation. Simulation results show that the adaptive PI controller improves frequency recovery and reduces steady-state error compared to conventional fixed gain PI.