发电技术 ›› 2024, Vol. 45 ›› Issue (4): 734-743.DOI: 10.12096/j.2096-4528.pgt.23029

• 智能电网 • 上一篇    下一篇

基于深度强化学习的有源配电网电压分层控制策略

杜婉琳1, 王玲1, 罗威2, 朱远哲1, 吕鸿1, 马潇男3, 周霞3   

  1. 1.广东电网有限责任公司电能质量重点实验室(广东电网有限责任公司电力科学研究院),广东省 广州市 510080
    2.广东电网有限责任公司梅州供电局,广东省 梅州市 514021
    3.南京邮电大学自动化学院、人工智能学院,江苏省 南京市 210023
  • 收稿日期:2023-08-26 修回日期:2023-10-18 出版日期:2024-08-31 发布日期:2024-08-27
  • 通讯作者: 周霞
  • 作者简介:杜婉琳(1985),女,硕士,高级工程师,电力系统分析技术,2788829207@qq.com
    马潇男(1999),男,硕士研究生,主要研究方向为配电网运行技术、电压管理,344717084@qq.com
    周霞(1978),女,博士,副教授,从事电力通信、电力系统分析与控制研究,本文通信作者,zhouxia@njupt.edu.cn
  • 基金资助:
    国家自然科学基金项目(52207009);南方电网公司科技项目(GDKJXM20200331)

Voltage Hierarchical Control Strategy of Active Distribution Network Based on Deep Reinforcement Learning

Wanlin DU1, Ling WANG1, Wei LUO2, Yuanzhe ZHU1, Hong LÜ1, Xiaonan MA3, Xia ZHOU3   

  1. 1.Key Laboratory of Power Quality of Guangdong Power Grid Co. , Ltd. (Electric Power Research Institute of Guangdong Power Grid Co. , Ltd. ), Guangzhou 510080, Guangdong Province, China
    2.Meizhou Power Supply Bureau of Guangdong Power Grid Co. , Ltd. , Meizhou 514021, Guangdong Province, China
    3.College of Automation & College of Artificial Intelligence, Nanjing University of Posts and Telecommunications, Nanjing 210023, Jiangsu Province, China
  • Received:2023-08-26 Revised:2023-10-18 Published:2024-08-31 Online:2024-08-27
  • Contact: Xia ZHOU
  • Supported by:
    National Natural Science Foundation of China(52207009);Science and Technology Projects of China Southern Power Grid Corporation(GDKJXM20200331)

摘要:

目的 分布式电源发电的随机性和波动性,给有源配电网(active distribution network,ADN)的电压控制带来了严峻的挑战,在此背景下,亟需一种高效的电压控制策略来保证ADN的安全运行。 方法 基于深度强化学习方法,提出了一种双层区域配电网电压控制策略。首先,以调压设备的调节特性和可控元素复杂化的特点为前提,针对ADN辐射网架结构,设计了区域协调控制区域和本地自治控制区域,分别构建每个区域的电压控制模型;然后,通过深度Q网络(deep Q-network,DQN)算法和深度确定性策略梯度(deep deterministic policy gradient,DDPG)算法对该模型进行求解,以实现实时跟踪电压变化的目的,有效解决了ADN运行过程中电压控制问题;最后,通过IEEE 33节点仿真算例对该方法进行了验证。 结果 利用DQN算法和DDPG算法分别求解协调控制区域和本地自治区域的控制变量,实现了ADN系统电压调节的实时决策,解决了ADN潮流双向流动、电压复杂多变的问题。 结论 所提控制策略控制电压偏差效果明显,具有很强的准效性和实用性。

关键词: 有源配电网 (ADN), 区域协调控制, 本地自治控制, 深度强化学习, 电压控制策略

Abstract:

Objectives The randomness and volatility of distributed power generation poses significant challenges for the voltage control in active distribution network (AND). In this context, there is an urgent need for an efficient voltage control strategy to ensure the safe operation of ADN. Methods Based on the deep reinforcement learning method, a voltage control strategy for double-layer regional distribution networks was proposed. First, based on the adjustment characteristics of voltage regulating equipment and the complexity of controllable elements, a regional coordinated control area and a local autonomous control area were designed for the radiating grid structure of the ADN, and the voltage control model of each area was constructed. Then, the model was solved by deep Q-Network (DQN) algorithm and deep deterministic policy gradient (DDPG) algorithm to achieve the purpose of tracking voltage changes in real time, and effectively solve the voltage control problem during the operation of the ADN. Finally, the method was verified by IEEE 33-bus simulation examples. Results The DQN algorithm and the DDPG algorithm were used to solve the control variables in the coordinated control region and the local autonomous region respectively, realizing real-time decision-making of voltage regulation in the ADN system, and solving the problems of bidirectional flow of ADN power flow and complex and changeable voltage. Conclusions The proposed control strategy has obvious effect on controlling voltage deviation, and has strong accuracy and practicality.

Key words: active distribution network (ADN), regional coordination control, local autonomous control, deep reinforcement learning, voltage control strategy

中图分类号: