کنفرانس بین المللی ماشین و محرکه های الکتریکی

صفحه اصلی / The 3rd International Conference on Electrical Machines and Drives

Transformative Control Optimization in PMSG-Based Wind Energy Systems: A Deep Reinforcement Learning Approach

نویسندگان :

Bagher Khadem Hamedani¹ Masoud Izadi² Nahid Izadi³ Amin Azarshab⁴ Ali Najari⁵ Behnam Zaker⁶

1- Hamedan University of Technology 2- Hamedan Azad University 3- Hamedan Azad University 4- Petroleum University of Technology 5- Sapienza University of Rome 6- Politecnico di Torino

کلمات کلیدی :

(Permanent-Magnet Synchronous Generators (PMSGs

چکیده :

This paper addresses the optimization of control strategies in Permanent-Magnet Synchronous Generator (PMSG)-based wind energy systems, with a specific focus on the Proportional-Integral (PI) controller. Traditionally, PI controller coefficients are optimized using methods like the response surface method and artificial neural networks. However, this research introduces a transformative approach by applying deep reinforcement learning (DRL), inspired by pioneering work in continuous control. The paper compares DRL with established optimization methods such as Particle Swarm Optimization (PSO), Krill Herd, and Grey Wolf. DRL exhibits adaptability to dynamic environments, continuous learning, effective handling of complex nonlinearities, autonomous decision-making, and real-time adaptation. These qualities position DRL as a versatile and efficient choice for optimizing control in PMSG-based wind energy systems. The paper is organized into sections, including system modeling and control methodology, exploration of the Krill Herd algorithm, and the empirical comparison of DRL with the Grey Wolf optimization algorithm. The findings emphasize DRL's transformative contributions to wind energy research, marking a pioneering step towards enhancing efficiency and sustainability in PMSG-based wind energy systems.