水库群长期优化调度时空组合降维算法  

Spatiotemporally Combined Dimensionality Reduction Algorithm for Optimizing Long-term Operation of Multi-reservoir Systems

作　　者：陈佳 章汉军 徐囡 CHEN Jia;ZHANG Hanjun;XU Nan(Zhuji Water Conservancy Bureau,Shaoxing 311800,China;Zheguang Middle School of Keqiao District of Shaoxing City,Shaoxing 312025,China)

机构地区：[1]诸暨市水利局,浙江绍兴311800 [2]绍兴市柯桥区浙光中学,浙江绍兴312025

出　　处：《人民珠江》2023年第8期58-67,92,共11页Pearl River

摘　　要：为了缓解水库群调度的“维数灾”效应,在保障解的质量的同时提高求解效率,对动态规划逐次逼近算法(DPSA)和逐步优化算法(POA)进行改进和综合,设计一种时、空组合降维算法。首先,采用基于“梯级水库链”的逐次逼近策略,将DPSA“单库轮流”优化模式拓展至“梯级水库链轮流”优化模式,弥补DPSA在处理梯级水库水力耦合关系上的不足;然后,采用动态变量解耦策略和扰动机制对POA进行改进,解决POA“盲目搜索”问题和“维数”问题;最后,对2种改进算法进行组合,在改进DPSA的框架下,采用改进POA求解“梯级水库链”优化问题。采用湖南沅水流域梯级水库发电调度问题和经典“十水库”问题进行测试,组合降维算法在解的质量和求解效率上优于7种现有算法。结果表明:组合降维算法能够有效地缓解水库群调度的“维数灾”效应,能够在不牺牲解的质量的情况下提高求解效率,具有应用于大规模复杂水库系统优化的潜力。To alleviate the“curse of dimensionality”and improve the solution efficiency while ensuring the quality of solutions in optimizing the operation of multi-reservoir systems,this paper proposes a spatiotemporally combined dimensionality reduction algorithm which integrates and improves the dynamic programming with successive approximation(DPSA)and the progressive optimality algorithm(POA).First,a chain-based successive approximation strategy is proposed to expand the DPSA's optimization mode from“single reservoir alternation”to“cascade reservoir chain alternation”,which makes up for the DPSA's shortcomings in dealing with the hydraulic coupling relationships among cascade reservoirs.Then,a dynamic variable decoupling strategy and perturbation mechanism are proposed to deal with the POA's blind search problem and dimensionality problem.Finally,the two improved algorithms are combined,in which the improved POA is applied to solving the optimization problems of cascade reservoir chains under the framework of the improved DPSA.The power generation operation problem of the cascade reservoirs in the Yuan River Basin of Hunan Province and the classical ten-reservoir problem are utilized to test the performance of the proposed algorithm.The proposed algorithm outperforms seven existing alternatives in terms of solution quality and efficiency.The results indicate that the proposed algorithm can effectively alleviate the“curse of dimensionality”in optimizing the operation of multi-reservoir systems,improve the efficiency while ensuring the quality of solutions and has potential to be applied to optimizing the operation of complex large-scale multi-reservoir systems.

关 键 词：水库调度 降维 逐次逼近 变量解耦 梯级水库链 偶极子优化 扰动机制 

分 类 号：TV213.4[水利工程—水文学及水资源] O229[理学—运筹学与控制论]