| 引用本文: | 祁向龙,陈 健,赵浩然,等.多时间尺度协同的配电网分层深度强化学习电压控制策略[J].电力系统保护与控制,2024,52(18):53-64. |
| QI Xianglong,CHEN Jian,ZHAO Haoran,et al.Multi-time scale cooperative voltage control strategy of a distribution network based on hierarchical deep reinforcement learning[J].Power System Protection and Control,2024,52(18):53-64 |
|
| 摘要: |
| 光伏和电动汽车的大量接入给配电网电压控制带来了严峻挑战。为此，提出了一种多时间尺度协同的配电网分层深度强化学习电压控制策略。首先，考虑电动汽车用户需求响应特性，在日前协同调度分组投切电容器制定分时电价引导用户改变充电行为实现有序充电。其次，在日内构建具有双层结构的深度强化学习策略，上层以长时间尺度通过电价调整引导用户响应实时价格激励以调节电动汽车充电功率，下层以短时间尺度对光伏逆变器和无功补偿装置进行控制，通过双层策略实现对具有不同时间响应特性资源的实时协同调控以降低系统电压偏差。最后，通过对改进的IEEE 33节点系统进行算例分析验证了所提策略的有效性。 |
| 关键词: 电动汽车 多时间尺度 电压控制 多智能体 深度强化学习 |
| DOI：10.19783/j.cnki.pspc.240122 |
| 分类号: |
| 基金项目:国家重点研发计划项目资助(2018YFA0702200) |
|
| Multi-time scale cooperative voltage control strategy of a distribution network based on hierarchical deep reinforcement learning |
|
QI Xianglong, CHEN Jian, ZHAO Haoran, ZHANG Wen, ZHANG Keyu
|
|
Key Laboratory of Power System Intelligent Dispatch and Control of Ministry of Education,
Shandong University, Jinan 250061, China
|
| Abstract: |
| The access of a high proportion of photovoltaic energy and a large number of electric vehicles imposes severe challenges to the voltage control of a distribution network. This paper presents a multi-time scale cooperative voltage control strategy based on hierarchical deep reinforcement learning. First, considering the demand response characteristics of electric vehicle users, the day-ahead time-of-use price is formulated to guide users to change their charging behaviors to achieve orderly charging of coordinated scheduling of switching capacitors. Secondly, a deep reinforcement learning strategy with a two-layer structure is constructed within the day. The upper layer guides users to respond to real-time price incentives through price adjustment to adjust electric vehicle charging load over a long time-scale, and the lower layer controls photovoltaic inverters and reactive power compensation devices on a short time-scale. By the two-layer strategy, the real-time coordinated regulation of resources with different time response characteristics is realized to reduce system voltage deviation. Finally, the effectiveness of the proposed strategy is verified on the improved IEEE 33-bus system. |
| Key words: electric vehicle multi-time scale voltage control multi-agent deep reinforcement learning |
