基于深度确定性策略梯度的主动配电网有功-无功协调优化调度  

Coordinated Optimal Dispatch of Active and Reactive Power in Active Distribution Networks Using Deep Deterministic Strategy Gradient

作　　者：孙国强[1] 殷岩岩 卫志农[1] 臧海祥[1] 楚云飞[1] SUN Guoqiang;YIN Yanyan;WEI Zhinong;ZANG Haixiang;CHU Yunfei(College of Energy and Electrical Engineering,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学能源与电气学院,南京市211100

出　　处：《电力建设》2023年第11期33-42,共10页Electric Power Construction

基　　金：国家自然科学基金项目(U1966205)。

摘　　要：为了实现主动配电网(active distribution network,ADN)的有功-无功资源协调控制,提高配电系统供电可靠性及经济性,提出一种基于深度确定性策略梯度(deep deterministic policy gradient,DDPG)的ADN有功-无功协调优化调度策略。首先,在避免电压和潮流越限的情况下,以ADN日运行成本最小为目标,计及可投切电容器组、有载调压变压器、微型燃气轮机和能量储存系统构建ADN有功-无功协调调度模型。其次,将ADN实时调度问题转化成马尔科夫决策过程,并定义系统的状态空间、动作空间及奖励函数。然后,为提升深度确定性策略梯度的离线训练速度和奖励回报,在算法中加入优先经验回放(priority experience replay,PER)机制,并搭建了基于优先经验回放机制的深度确定性策略梯度(PER-DDPG)ADN在线调度框架。最后,在修改的IEEE-34节点配电系统上进行仿真,算例结果表明,PER-DDPG方法通过高效的经验学习,能够为ADN提供安全、经济的调度策略。To achieve coordinated control of active power and reactive power resources within the active distribution network(ADN)and enhance the reliability and cost-effectiveness of the distribution system's power supply,we propose an optimization strategy for ADN active power and reactive power coordination based on deep deterministic policy gradient(DDPG)scheduling.First and foremost,our approach focuses on minimizing the daily operating costs of the ADN while avoiding voltage and power flow exceeding their limits.It takes into account various factors,including switchable capacitor banks,on-load voltage regulating transformers,micro gas turbines,and energy storage systems.We establish a model for the coordinated dispatch of active and reactive power within the ADN.Next,we transform the real-time scheduling problem in the ADN into a Markov decision process,defining the state space,action space,and reward function for the system.To enhance the offline training speed and reward returns of the DDPG algorithm,we introduce a priority experience replay(PER)mechanism.This leads to the development of an online scheduling framework known as PER-DDPG for the ADN.Finally,we conduct simulations on the modified IEEE-34 node power distribution system.The results of these simulations demonstrate that the PER-DDPG method effectively provides a secure and cost-efficient dispatch strategy for the ADN,achieved through efficient empirical learning.

关 键 词：主动配电网 有功无功协调优化 深度确定性策略梯度(DDPG) 在线调度框架 优先经验回放机制 

分 类 号：TM732[电气工程—电力系统及自动化]