引用本文:张延,唐昊,王珂,潘毅,李怡瑾.考虑源荷随机性的跨区互联电网直流联络线调度学习优化[J].控制理论与应用,2019,36(7):1047~1056.[点击复制]
ZHANG Yan,TANG Hao,WANG Ke,PAN Yi,Li Yi-jin.Learning-based optimization of direct current tie-line dispatch for inter-regional power grid considering the stochasticity of source-load[J].Control Theory and Technology,2019,36(7):1047~1056.[点击复制]
考虑源荷随机性的跨区互联电网直流联络线调度学习优化
Learning-based optimization of direct current tie-line dispatch for inter-regional power grid considering the stochasticity of source-load
摘要点击 2794  全文点击 1212  投稿时间：2018-05-08  修订日期：2018-12-11
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2018.80336
  2019,36(7):1047-1056
中文关键词  联络线调度  新能源消纳  柔性负荷  随机性  强化学习
英文关键词  tie-line dispatch  accommodation of new energy  flexible load  stochasticity  reinforcement learning
基金项目  国家重点研发计划项目(2017YFB0902600),国家电网公司科技项目(SGJS0000DKJS1700840)
作者单位E-mail
张延 合肥工业大学 zhangyan@mail.hfut.edu.cn 
唐昊* 合肥工业大学 htang@hfut.edu.cn 
王珂 中国电力科学研究院(南京）  
潘毅 中国电力科学研究院(北京）  
李怡瑾 合肥工业大学  
中文摘要
      在跨区互联电网中, 充分利用直流联络线调度能力可以有效地平衡电力资源的配置, 促进新能源的消纳. 本文针对源荷不确定性的跨区互联电网直流联络线调度问题, 首先用连续马尔科夫过程模型描述互联电网中风电 出力与负荷需求随机动态特性; 然后在功率平衡及联络线日交易电量约束等实际运行要求前提下, 将直流联络线 调度优化问题建立成离散马尔科夫决策过程模型. 在该模型下, 调度机构根据互联电网系统各时段源荷的功率情 况, 动态调整联络线输电计划和配套的柔性负荷调节方案, 以达到提升系统运行效益的优化目标; 最后引入强化学 习方法对调度策略进行优化求解. 通过学习优化, 系统平均日运行代价显著下降且最终收敛. 实验结果表明考虑源 荷随机性的直流联络线动态调整方法可有效地提高互联电网发输电系统的运行效益.
英文摘要
      In inter-regional power grid, the power resource can be allotted effectively by direct current tie-line to promote utilization ratio of renewable energy. The dispatch problem for direct current tie-line in inter-regional power grid with uncertain renewable sources and demands was researched in this paper. Firstly, the random dynamic characteristics of wind power output and load demand was described as continuous Markov process. Secondly, based on practical operation requirements including the power balance constraint and the limit of tie-line power, the optimal dispatch problem for direct current tie-line was described as a discrete Markov decision process. According to the power of renewable energy output and load demand, the optimized strategy for the plan of tie-line and flexible load in each period was established to promote the running benefit of the system in this model. Finally, a reinforcement learning method was adopted to obtain the optimal policy. The daily average cost of system operation decreases significantly and eventually converges by reinforcement learning. Simulation results show that the operational efficiency of inter-regional power grid is significantly enhanced by the proposed dynamic adjustment method for tie-line.