Precise power flow control of transmission corridors using an integrated HELM-SAC method

• A new HELM-SAC method combines Soft Actor-Critic algorithm and Holomorphic Embedding Load Flow method for precise power flow control. • The HELM indices can provide robust adjustment directions when power flow is divergent. • With well-trained SAC agent, the transmission corridor power can be quickly adjusted to the target value. • With the HELM indices and SAC method, the power flow can be adjusted from a non-convergence state to convergence with more stable operating state. With the growing uncertainties caused by the increased penetration level of renewable generation and EV charging loads, power flow control of key transmission corridors faces grand challenges. However, it usually requires operators’ manual intervention, and in certain circumstances, power flow calculation may not converge in real-time, affecting the control performance. This paper proposes a new HELM-SAC method, which combines the Soft Actor-Critic (SAC) deep reinforcement learning algorithm and the Holomorphic Embedding Load Flow method (HELM) for transmission power flow control. Firstly, a power grid model is established, and the target corridor is selected. The power sensitivity of the corridor with respect to generator outputs is calculated using the perturbation-based method, and the generators are ranked accordingly to participate in control. Then, the HELM is used to calculate the static voltage stability indices to formulate the reward mechanism of the SAC agent, which is easy to calculate and has its significance when it is not convergent so that the adjustment algorithm also has the adjustment direction when it is not convergent. Next, the environment and agent of SAC are trained, and the corridor power can be quickly adjusted to the target value by continuous optimization strategy. By introducing V S I m a x and V S I a v e indices for power flow adjustment, the power flow can be adjusted from a non-convergence state to convergence, and the power grid can be adjusted to a more stable operating state. The effectiveness of the proposed method is verified on the IEEE 14-bus and the IEEE 57-bus power grid model.