【作者】 张慧峰；

【导师】 张勇传； 周建中；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2013， 博士

【摘要】 为了满足不同的经济、社会和生态需求，长江上游流域梯级水库群联合调度问题逐渐往高维、强耦合和多目标的方向发展，传统的调度模型已无法满足水库群防洪、发电、供水、生态及航运综合效益最优等需求，同时受到流域水文过程、调度方式及电力系统等多方面因素的影响，存在一系列亟待解决的关键科学难题。经典的优化理论在解决实际工程问题中往往出现模型适应性差、解的维数灾和计算精度不高等问题，迫切需要提高变化环境下的水文预报预测精度，综合考虑发电、防洪、生态等多个目标合理安排梯级水库群水量调配过程，最大化水量利用综合效益。因此，本文紧密结合梯级水库群的工程应用实际，建立一种以信息科学和系统控制理论为基础的优化理论与方法，解决长江上游径流非线性综合预测校正、水库群防洪、发电以及生态调度多目标联合调度与决策等关键科学技术难题。本文主要研究内容与创新点包括：（1）围绕流域水库群调度的实际工程需求，以流域径流高精度径流预报为目标，揭示了流域径流过程的趋势变化特征，提出了基于经验模式分解的卡尔曼滤波校正方法，结合不同水文预报模型的地域适应性特点，建立了适应于流域径流高精度的多预报模型间相匹配的串并联耦合校正模型，从而充分考虑了各类径流预报模型的预测优点，有效提高径流预报精度。（2）以帕累托优化为理论基础，研究适合于网络分布式计算的调度模型空间解构方式，并探究了该方法参数对调度模型求解精度和效率影响的变化规律，研究了可增强算法对多目标优化适用性的参数控制方法，采用混沌自适应控制理论对差分进化的局部进化过程进行改进，一定程度上避免了“早熟”现象的发生，以文化理念为框架，提出了基于文化理念框架的自适应多目标优化算法方法，充分运用外部档案集的精英保留策略及Pareto前沿的分布性控制策略，在保证群体进化收敛性的同时也保持了Pareto前沿分布的均匀性，得到了各目标之间的Pareto曲线或者曲面关系。（3）根据流域梯级水库群联合多目标优化运行模式，从防洪安全、综合兴利效益、生境修复条件等多个决策空间给出了多目标优化调度建模的量化依据。探究了不同调度时期、不同运行工况下梯级水电站群各调度目标间的均衡方式，建立了能有效保障梯级水库群防洪、发电、生态及水火电调度系统多目标优化调度模型，运用控制理论和智能优化理论相结合的方法，提出了能同时优化多个目标的高效求解算法，并分析了折衷调度方案的水位、流量及出力过程。（4）考虑水情、雨情、工情和梯级电站运行目标侧重点，辨识影响梯级水库群调度方案优选的关键因子，通过推导熵权计算公式和联系数贴近度函数若干典型计算公式，确定方案决策优选的主、客观指标权重；结合调度人员的知识经验、偏好、决策层次的差异，建立面向梯级电站水库群防洪、发电和生态调度的多属性决策模型，提出基于人工智能技术和多属性决策理论的梯级电站调度方案优劣排序方法，对不同运行时期、运行工况下优化控制方案的综合应用效果进行决策评价和优选，确定梯级水库群联合调度的最佳方案。更多还原

【Abstract】 In order to satisfy different economic requirements, Yangtze River basin cascadereservoirs scheduling problems gradually develop towards higher-dimensional, strongcoupling and multi-target direction, the traditional scheduling models have been unable tosatisfy reservoir flood control, power generation, water supply, ecology andcomprehensive benefits summa shipping requirements. Meanwhile, for being watershedby hydrological processes, scheduling and power systems, and many other factors, there isa series of key scientific problems to be solved. Classical optimization theory for solvingpractical engineering often arises model poor adaptability solutions curse ofdimensionality and low accuracy problems, it is needed to improve the changingenvironment hydrological forecasting prediction accuracy, considering the powergeneration, flood control, ecological and other reasonable arrangements for the targetdeployment process water cascade reservoirs to maximize the overall efficiency of wateruse. Therefore, this article closely relates cascade reservoirs practical engineeringapplications, takes the establishment of a kind of information science and systems basedon the theory of optimal control theory and method to solve nonlinear integratedforecasting Yangtze River runoff correction reservoir flood control, power generation andeco-scheduling and more target joint scheduling and decision-making, and other keyscientific and technological problems. This paper studies the content and innovationsinclude:(1) Around the of Hydropower Reservoirs scheduling the actual project requirements,the key scientific issues facing high-precision breakthrough basin runoff forecasting andengineering problems aim to reveal the evolution and trends of watershed runoff cyclevariation characteristics, this paper proposes empirical model decomposition basedkallman filter revision method for water forcasting. With analysis and evaluation ofdifferent hydrological forecasting model geographical adaptive characteristics, itestablishes model libraries to adapt to the high-precision forecast of runoff in differentprediction model that matches the series-parallel coupling correction technology, makefull use of the advantages of the various types of runoff forecasting model, and effectively improve the runoff forecast accuracy.(2) To Pareto optimization as the theoretical basis to study the way for networkdistributed computing the scheduling model space deconstruction, and explores thevariation of the method parameters affect the scheduling model for solving accuracy andefficiency enhancement algorithm for multi-objective optimization applicability ofparameter control methods, the use of the chaotic adaptive control theory of partialdifferential evolution evolutionary process to improve, to avoid the occurrence of thephenomenon of "premature" to a certain extent, to the concept of culture as a framework,adaptive framework based on cultural ideas multi-objective optimization algorithmmethod, make full use of an external file sets the elitist strategy and distribution of thePareto front control strategy to maintain the uniformity of the distribution of Pareto frontgroups of evolutionary convergence between each target Pareto curve or surface.(3) Basin cascade reservoirs joint multi-objective optimization mode of operation,given the multi-objective optimization scheduling modeling quantitative basis from theflood safety, Hennessy-effective, habitat restoration conditions and other decision-makingspace. Explores the different scheduling periods, different operating conditions cascadehydropower stations each scheduling a balanced manner between the target and theeffective protection of cascade reservoirs group of flood control, power generation,ecological and hydro-thermal scheduling system multi-objective optimization schedulingmodel, the use of control theory and intelligent optimization theory a combination ofmethods, to optimize multiple targets at the same time efficient solving algorithm, andanalyzes the water level, flow, and contribute to the process of compromise schedulingscheme.(4) Consider water, rainfall, engineering conditions and cascade hydropower stationsrun target focus, identification affect the cascade reservoirs scheduling program preferredkey factor calculated by derivation of the entropy weight and the number of contact closeto the function of certain typical formula to determine Optimization Decision subjectiveand objective index weight; differences in knowledge and experience combined withdispatchers, preferences, and decision-making level, the establishment of multi-attributedecision model is proposed based on artificial intelligence technology and multi-attribute decision theory rung for cascade Hydropower reservoirs in flood control, powergeneration and ecological operation power plant scheduling scheme pros and cons ofsorting methods, different run time, the application effect of operating conditionsoptimized control scheme for the evaluation and optimization of decision-making, todetermine the cascade reservoirs scheduling the best solution.更多还原