梯级水电站群中长期优化调度的正交离散微分动态规划方法  

Orthogonal Discrete Differential Dynamic Programming for Mid-long Term Optimal Operation of Cascade Hydropower System

作　　者：冯仲恺[1] 廖胜利[1] 牛文静[1] 程春田[1] 唐建兴[2] 苏华英[2] 

机构地区：[1]大连理工大学水电与水信息研究所,辽宁省大连市116024 [2]贵州电网公司电力调度控制中心,贵州省贵阳市550000

出　　处：《中国电机工程学报》2015年第18期4635-4644,共10页Proceedings of the CSEE

基　　金：国家自然科学基金重大国际合作项目(51210014);国家重点基础研究发展计划项目(973计划)(2013CB035906)~~

摘　　要：我国特大流域梯级水电站群正在形成,这些梯级水电站群普遍具有装机规模大、级数众多的特点。随着巨型梯级水电站的集中投产,其优化调度计算规模不断增加,维数灾问题日益凸显。在分析离散微分动态规划(discrete differential dynamic programming,DDDP)算法的基础上,提出了正交试验设计和DDDP相结合的正交离散微分动态规划方法(orthogonal discrete differential dynamic programming,ODDDP)。该方法以DDDP为基础,采用正交试验设计选取具有"均衡分散,整齐可比"性质的部分状态组合,以减少方法所需存储量与计算量,进而提高计算的规模和效率。乌江干流梯级水电站群仿真调度结果表明,ODDDP在大幅缩短计算耗时的同时可获得与DDDP相近的优化结果,系统求解效率和计算规模显著提高。With the rapid expansion of the huge cascade hydropower system in China, the computing scale of hydro scheduling optimization is undergoing an explosive growth and faces the increasingly severe curse of dimensionality. It is essential to find some novel and effective methods to enhance the computing efficiency. An orthogonal discretedifferential dynamic programming （ODDDP） algorithm for mid-long term optimal operation of cascade hydroplants was proposed. On the basis of classical discrete discrete differential dynamic programming （DDDP）, the proposed algorithm took advantage of orthogonal design to choose small but representive state combinations of different plants at each period, which led to a significant reduction in the computation time and memory requirement. The results of Wujiang River show that ODDDP is comparable to DDDP while making a significant reduction in computing time. The ODDDP is an effective algorithm for the mid-long term optimal operation of hydropower system.

关 键 词：离散微分动态规划 水电站群 优化调度 降维 维数灾 试验设计 

分 类 号：TM73[电气工程—电力系统及自动化]