Safe voltage regulation strategy for distribution networks with multiple DERs based on the integrated DDPG-MPC method

With increasing penetration of distributed energy resources (DERs), the operation of modern distribution networks has been challenged by frequent overvoltage and undervoltage issues. Therefore, this paper proposes a novel safe voltage control (SVC) strategy for distribution networks with multiple DERs by fusing deep deterministic policy gradient (DDPG) and model predictive control (MPC) methods. In this context, different timescale voltage control devices, such as on-load tap changers (OLTC), capacitor banks (CBs), and distributed generations, are optimally coordinated to reduce the voltage violations and keep bus voltages within the feasible range. First, MPC-based slow-timescale control with enhanced interpretability is employed to optimize the tap operations of OLTC and CBs, which aims to regulate the long-term sizable voltage fluctuations in distribution networks. Subsequently, the reactive power outputs of DERs are further coordinated to reduce the voltage deviations on a faster timescale. The fast Volt-Var optimization problem is transformed into a constrained Markov decision process, which is then trained and solved using a data-driven DDPG algorithm. Furthermore, a safety verification mechanism is designed based on voltage sensitivity to ensure the voltage stability of distribution networks throughout the training process. Once trained, the fast-timescale control relies solely on local information to guarantee real-time voltage control performance, while the slow-timescale control utilizes global information to ensure the operational reliability of OLTC and CBs. A modified 20 kV Italian distribution network with various regulation devices was utilized to illustrate the voltage control performance of the proposed SVC strategy.