Reinforcement Learning Based Short Time Scale Operation of Wind-Solar-Thermal Coupling Generation System Considering Nonconvex Ramping Constraint | IEEE Conference Publication | IEEE Xplore

Reinforcement Learning Based Short Time Scale Operation of Wind-Solar-Thermal Coupling Generation System Considering Nonconvex Ramping Constraint


Abstract:

This paper proposes a reinforcement learning (RL) based short time scale economic operation approach for wind-solar-thermal coupling generation system (WSTCGS) considerin...Show More

Abstract:

This paper proposes a reinforcement learning (RL) based short time scale economic operation approach for wind-solar-thermal coupling generation system (WSTCGS) considering the non-convex ramping constraints of thermal power unit (TPU) with deep peak regulation (DPR) capability. The WSTCGS operation is modeled as a Markov decision process (MDP) with an agent controlling the power sources in the WSTCGS. The proposed approach utilizes Soft Actor-Critic (SAC) algorithm to train the RL agent for achieving economic power sources scheduling in WSTCGS. The established RL framework considers the non-convex ladder type ramping constraints of TPUs, short-term forecasts of RES and load to accurately formulate the WSTCGS stochastic economic operation in a short time scale. The simulation results based on a real-world WSTCGS test system demonstrate that the proposed RL framework can effectively schedules TPUs to provides basic load, tracks net load fluctuations, and achieves high RES utilization in short time scale.
Date of Conference: 16-19 October 2023
Date Added to IEEE Xplore: 16 November 2023
ISBN Information:

ISSN Information:

Conference Location: Singapore, Singapore

References

References is not available for this document.