| 引用本文: | 谢黎龙,李勇汇,范培潇,万黎,黄梦旗,杨军.基于深度强化学习的孤立多微电网系统频率和电压综合控制[J].电力自动化设备,2024,44(6):118-126. |
| XIE Lilong,LI Yonghui,FAN Peixiao,WAN Li,HUANG Mengqi,YANG Jun.Deep reinforcement learning-based integrated frequency and voltage control for isolated multi-microgrid system[J].Electric Power Automation Equipment,2024,44(6):118-126. |
|
| 本文已被：浏览 3216次 下载 595次 |
码上扫一扫！ |
|
|
| 基于深度强化学习的孤立多微电网系统频率和电压综合控制 |
|
谢黎龙1,2, 李勇汇1,2, 范培潇1,2, 万黎3, 黄梦旗1,2, 杨军1,2
|
|
1.交直流智能配电网湖北省工程技术研究中心，湖北 武汉 430072;2.武汉大学 电气与自动化学院，湖北 武汉 430072;3.国网湖北省电力科学研究院，湖北 武汉 430077
|
|
| 摘要: |
| 分布式电源出力不确定性和负荷功率扰动给孤立多微电网系统稳定带来较大威胁。提出基于多智能体柔性动作评价(MA-SAC)算法的孤立多微电网负荷频率控制器(LFC)，同时采用柔性动作评价(SAC)算法对自动电压调节器(AVR)的比例积分(PI)控制参数进行优化调整。建立了多微电网LFC和AVR组合模型。对于电压和频率控制器的设计，分别根据SAC算法和多智能体深度强化学习(MA-DRL)框架建立各自的状态、动作空间与奖励函数。选择合适的神经网络与训练参数经过预学习生成深度强化学习控制器。最后通过仿真分析，基于SAC算法优化的PI控制器能更快跟踪电压参考值；多微电网系统遭遇功率扰动时，MA-SAC控制器可以快速维持频率稳定。 |
| 关键词: 多微电网系统 柔性负荷 负荷频率控制 自动电压调节 MA-SAC算法 |
| DOI：10.16081/j.epae.202311013 |
| 分类号: |
| 基金项目:国家自然科学基金资助项目(51977154)；国网湖北省电力有限公司科技项目(B31532226407) |
|
| Deep reinforcement learning-based integrated frequency and voltage control for isolated multi-microgrid system |
|
XIE Lilong1,2, LI Yonghui1,2, FAN Peixiao1,2, WAN Li3, HUANG Mengqi1,2, YANG Jun1,2
|
|
1.Hubei Engineering and Technology Research Center for AC/DC Intelligent Distribution Network, Wuhan 430072, China;2.School of Electrical Engineering and Automation, Wuhan University, Wuhan 430072, China;3.State Grid Hubei Electric Power Research Institute, Wuhan 430077, China
|
| Abstract: |
| The output uncertainty of distributed power source and the power disturbance caused by load pose a greater threat to the stability of isolated multi-microgrid systems. An isolated multi-microgrid load frequency controller(LFC) based on multi-agent soft actor critic(MA-SAC) algorithm is proposed, and the soft actor critic(SAC) algorithm is used to optimize and adjust the proportional integral(PI) control parameters of the automatic voltage regulator(AVR). A combined model of LFC and AVR for multi-microgrid is developed. For the design of voltage and frequency controllers, the corresponding states, action spaces and reward functions are established according to SAC algorithm and multi-agent deep reinforcement learning (MA-DRL) framework, respectively. The appropriate neural network and training parameters are selected to generate the deep reinforcement learning controller through pre-learning. Finally, through simulation analysis, the PI controller optimized based on SAC algorithm can track the voltage the reference value faster, and the MA-SAC controller can maintain the frequency stability quickly when the multi-microgrid system encounters power disturbance. |
| Key words: multi-microgrid system flexible load load frequency control automatic voltage regulation MA-SAC algorithm |
