Adaptive and coordinated damping control of multiple converters for renewable energy transmission system via VSC-HVDC

This paper proposes a novel oscillation suppression method for renewable energy transmission systems via VSC-HVDC using a multi-agent deep reinforcement learning (DRL) approach. First, a neural network based technique is provided for identifying converter impedances. Then, the system frequency domain impedance network is constructed to calculate the system eigenvalues, which serves as the basis for DRL training. Subsequently, the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm is employed to design the adaptive supplementary damping control, ensuring an effective oscillation suppression under varying operating conditions. The reward function combines model-driven and data-driven insights, significantly improving training effects. Finally, a multi-agent DRL collaboration mechanism is designed to realize a coordinated oscillation mitigation among multiple converters. Simulation results in MATLAB/Simulink confirm the proposed method’s superiority over traditional approaches in suppressing oscillations and maintaining system stability across different scenarios.