Coordinated Control of Thermal Power Unit–Hybrid Energy Storage System for AGC Based on Improved Markov Power Prediction

High penetration of renewable energy poses severe frequency stability challenges to modern power systems, while conventional thermal power units (TPUs) fail to satisfy high-precision frequency regulation demands due to slow dynamic response. A hybrid energy storage system (HESS) with a flywheel and battery has become a mainstream scheme for TPU-HESS coordinated regulation by realizing complementary advantages. However, existing strategies generally adopt rigid state-of-charge (SOC) constraints, lack battery lifespan optimization, and suffer from incomplete power prediction models with weak integration with real-time control. To fill these gaps, this paper proposes a prediction-assisted coordinated control framework. A three-matrix adaptive Markov prediction model corrected by K-means time-series morphological clustering is established, combined with a weighted random sampling scenario tree method. Furthermore, a prediction-driven dynamic SOC interval feedforward-feedback control strategy is designed for HESS power dispatch and SOC management. Simulations verify that the clustering correction reduces prediction RMSE by 22.25%. The proposed strategy increases the optimal SOC operation ratio of batteries by 34.25%, decreases equivalent cycles by 19.80%, and lifts the comprehensive AGC performance index by 160.93%. This study provides technical support for efficient and durable TPU-HESS frequency regulation systems.