Hierarchical Multi-Agent Deep Reinforcement Learning for Coordinated Voltage Regulation in Active Distribution Networks with Hybrid Devices

• Main Authors: Zhang, Tingjun, Yu, Liang, Yue, Dong, Zhang, Dingfan
• Format: Conference Proceeding
• Language: English
• Subjects: Active distribution networks; Couplings; Deep learning; Games; hierarchical multi-agent deep reinforcement learning; Markov processes; Reinforcement learning; Renewable energy sources; Simulation; two-timescale voltage regulation
• ISSN: 2161-2927
• DOI: 10.23919/CCC58697.2023.10240379

Voltage regulation is indispensable to the secure operation of active distribution networks (ADNs) with high penetrated renewables. In this paper, we investigate an optimal two-timescale voltage regulation problem of ADNs with hybrid devices. Specifically, we intend to minimize the total power loss of the whole ADNs while maintaining all bus voltages within a safe range. Due to the existence of uncertain parameters, temporal couplings, multiple timescales, mixed decision variables, and unknown system models, it is challenging to solve the above optimization problem. To this end, we propose a coordinated voltage regulation algorithm based on hierarchical multi-agent attention-based deep reinforcement learning. The proposed algorithm can support flexible collaboration among hybrid devices. Simulation results based on real-world traces show the effectiveness of the proposed algorithm.