Deep reinforcement learning-based control of wind energy conversion system integrated with superconducting magnetic energy storage

Abstract

Traditional controllers, such as model predictive control, struggle to handle the highly dynamic and nonlinear nature of wind energy conversion systems effectively. They lack the flexibility and adaptability required to respond to fast-changing conditions in real time. Furthermore, when energy storage systems such as Superconducting Magnetic Energy Storage (SMES) are incorporated to smooth out power fluctuations, conventional control methods may not fully exploit the potential of such advanced storage technologies. This creates a pressing need for innovative, robust, and adaptive control strategies that can address the complexities of renewable energy systems while maintaining system stability and optimizing performance under variable conditions, as proposed in this research work. This research presents an advanced deep reinforcement learning (DRL)-based control framework for managing WECS integrated SMES to enhance power system stability under varying wind conditions. The study evaluates the performance of the proposed Twin Delayed Deep Deterministic Policy Gradient (TD3 DRL)-based controller against traditional controllers. Extensive simulations under scenarios such as wind gusts and step changes in wind speed reveal that the TD3 DRL controller significantly outperforms predictive control in minimizing errors in critical parameters such as grid power, point-of-common-coupling voltage, and DC-link voltage). Moreover, the incorporation of SMES further improves dynamic response and energy quality by effectively mitigating fluctuations. These results highlight the superior adaptability, precision, and robustness of the TD3 DRL-based control strategy, making it a viable solution for enhancing the reliability of renewable energy systems in modern power grids.