冯昌森, 张瑜, 谢路耀, 文福拴, 张凯怡, 张有兵. 配电系统双时间尺度电压管理的深度强化学习方法[J]. 电力系统自动化, 2022, 46(12): 202-209.
引用本文: 冯昌森, 张瑜, 谢路耀, 文福拴, 张凯怡, 张有兵. 配电系统双时间尺度电压管理的深度强化学习方法[J]. 电力系统自动化, 2022, 46(12): 202-209.
FENG Changsen, ZHANG Yu, XIE Luyao, WEN Fushuan, ZHANG Kaiyi, ZHANG Youbing. Deep Reinforcement Learning Approach for Dual-timescale Voltage Management in Distribution System[J]. Automation of Electric Power Systems, 2022, 46(12): 202-209.
Citation: FENG Changsen, ZHANG Yu, XIE Luyao, WEN Fushuan, ZHANG Kaiyi, ZHANG Youbing. Deep Reinforcement Learning Approach for Dual-timescale Voltage Management in Distribution System[J]. Automation of Electric Power Systems, 2022, 46(12): 202-209.

配电系统双时间尺度电压管理的深度强化学习方法

Deep Reinforcement Learning Approach for Dual-timescale Voltage Management in Distribution System

  • 摘要: 随着可再生能源发电渗透率的不断增大,配电系统的电压越限问题愈发频繁,亟需高效的电压管理策略以保证配电系统的安全经济运行。首先,文中建立了双时间尺度的配电系统电压管理模型,实现不同时间响应特性的调压设备协调控制。然后,将2个时间尺度的电压管理模型建模为马尔可夫决策过程,在有效考虑两者的时间耦合关系和可控设备物理特性的基础上,分别利用多智能体深度确定性策略梯度算法和双深度Q网络算法求解模型,实现了双时间尺度的实时电压管理。最后,基于IEEE 33节点配电系统进行算例分析,验证了所提模型和方法的有效性。

     

    Abstract: With the increasing penetration rate of renewable energy generation, the problem of voltage violation in the distribution system becomes more frequent, and efficient voltage management strategies are urgently needed to ensure the secure and economic operation of the distribution system. First, this paper establishes a dual-timescale voltage management model for the distribution system to realize the coordinated control of voltage regulators with different time response characteristics. Then, the voltage management models of the two time scales are modeled as Markov decision process(MDP). Effectively considering the temporal coupling relationship between the two time scales and the physical characteristics of controllable devices, the dual-timescale realtime voltage management is realized by using the multi-agent deep deterministic policy gradient algorithm and the double deep Q network algorithm to solve the model, respectively. Finally, the effectiveness of the proposed model and method is demonstrated by case studies on the IEEE 33-bus standard distribution system.

     

/

返回文章
返回