新型工业化
新型工業化
신형공업화
New Industrialization Straregy
2015年
6期
33-40
,共8页
吴雯美%陆江%谭敏%肖少华%韩传家%郭乐欣%余涛
吳雯美%陸江%譚敏%肖少華%韓傳傢%郭樂訢%餘濤
오문미%륙강%담민%초소화%한전가%곽악흔%여도
多区域无功优化%低碳电力%相关均衡%强化学习
多區域無功優化%低碳電力%相關均衡%彊化學習
다구역무공우화%저탄전력%상관균형%강화학습
multi-regional reactive power optimization%low-carbon electricity%correlated equilibrium%reinforcement learning
为了适应智能电网分布式发展趋势,提出一种基于相关均衡强化学习(CEQ)的分区多目标无功优化算法,以解决数据海量、通信瓶颈、协调互动等相关问题。同时为响应国家低碳环保战略,将碳排放引入到电力系统无功优化问题中,将其作为无功优化的目标之一。本文采用CEQ算法合理配置电力系统中控制变量,通过区域间的相关均衡博弈进行信息的沟通与共享实现分区多目标无功优化问题的寻优,有效解决了区域间信息共享机制受限和维数灾难问题。IEEE标准9节点电力系统仿真算例表明,通过预学习与在线学习的结合该算法能有效快速的进行多区域无功优化问题求解。
為瞭適應智能電網分佈式髮展趨勢,提齣一種基于相關均衡彊化學習(CEQ)的分區多目標無功優化算法,以解決數據海量、通信瓶頸、協調互動等相關問題。同時為響應國傢低碳環保戰略,將碳排放引入到電力繫統無功優化問題中,將其作為無功優化的目標之一。本文採用CEQ算法閤理配置電力繫統中控製變量,通過區域間的相關均衡博弈進行信息的溝通與共享實現分區多目標無功優化問題的尋優,有效解決瞭區域間信息共享機製受限和維數災難問題。IEEE標準9節點電力繫統倣真算例錶明,通過預學習與在線學習的結閤該算法能有效快速的進行多區域無功優化問題求解。
위료괄응지능전망분포식발전추세,제출일충기우상관균형강화학습(CEQ)적분구다목표무공우화산법,이해결수거해량、통신병경、협조호동등상관문제。동시위향응국가저탄배보전략,장탄배방인입도전력계통무공우화문제중,장기작위무공우화적목표지일。본문채용CEQ산법합리배치전력계통중공제변량,통과구역간적상관균형박혁진행신식적구통여공향실현분구다목표무공우화문제적심우,유효해결료구역간신식공향궤제수한화유수재난문제。IEEE표준9절점전력계통방진산례표명,통과예학습여재선학습적결합해산법능유효쾌속적진행다구역무공우화문제구해。
In order to meet the development trend of smart grid, the multi-regional reactive power optimization algorithm based on the correlated equilibrium Q-learning (CEQ) algorithm is proposed to solve the problems of tremendous data, communication bottleneck and interaction. Meanwhile, in response to the national strategy of low carbon environmental protection, CO2emission is considered as one of the control objectives in reactive power optimization. In this paper, CEQ algorithm is adopted to allocate the control variables rationally. Then the best multi-regional reactive power optimization method is obtained through the information communication and sharing which is realized by correlated equilibrium game among areas, thus offering a solution to the limited information-sharing mechanisms and curse of dimensionality problem effectively. The simulation of the IEEE 9-bus system indicates that CEQ algorithm solves the multi-regional collaborative reactive power optimization quickly and rationally with the combination of pre-learning and online learning.