کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Resilient Reinforcement Learning–Based Load Frequency Control in Fractional Microgrids With Stochastic Time Delays

نویسندگان :

Mostafa Maleki¹ Farhad Bayat² Abolfazl Jalilvand³ Reza Noroozian⁴

1- University of Zanjan 2- University of Zanjan 3- University of Zanjan 4- University of Zanjan

کلمات کلیدی :

Resilient reinforcement learning،load frequency control،fractional microgrids،stochastic time delays

چکیده :

The high penetration of renewable resources in islanded microgrids reduces system inertia, substantially increasing the challenge of load-frequency control (LFC). Furthermore, the widespread use of power electronic interfaces introduces nonlinear and memory-dependent dynamics into the overall system. These challenges are compounded by stochastic communication delays in the control channels, which further threaten closed-loop stability. This paper proposes a model-free TD3-based secondary controller for a fractional-order microgrid that explicitly accounts for bidirectional, time-varying delays. A fuzzy–PID controller is adopted as a realistic benchmark to ensure a fair evaluation. Simulation results demonstrate that the proposed controller achieves faster settling, smaller frequency deviations, and more stable control actions even under low-inertia operating conditions and stochastic delays. Hence, the proposed approach offers a robust, model-independent solution to enhance dynamic frequency stability in future renewable-dominated microgrids.