【作者】 杨俊杰；

【导师】 周建中；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2007， 博士

【摘要】 流域梯级联合调度决策是在水文循环、发电控制、电网安全、电能需求、市场交易规则、以及用电行为等约束条件下的大型、动态、非凸、非线性的多目标不确定性决策问题,较传统水电能源优化调度复杂得多,国内外众多学者一直致力于研究能有效解决上述问题的各种方法。然而,流域梯级水电系统中复杂目标的相互冲突和约束条件的耦合作用使得问题的描述和模型的求解极为困难,至今几乎没有令人满意的解决方法,亟待进一步发展新的理论并探索其技术实现方法。因此,流域梯级复杂水电能源系统多目标调度决策的理论与方法的研究始终是学术前沿的热点问题。本文通过对流域梯级水电能源系统的复杂性分析,采用复杂系统理论和现代智能进化方法,对流域梯级水电联合优化调度决策的理论与方法进行了深入研究。针对所建立的流域梯级水电系统多目标优化调度模型,提出了自适应网格多目标粒子群优化算法和基于广义集对分析的多属性决策方法,证明了优化算法处理大规模、多目标、复杂约束条件优化问题的有效性,体现了集对分析决策方法处理不确定多属性决策问题的优势,发展了流域梯级水电系统优化调度决策的理论。研究成果成功应用于三峡梯级水电联合多目标调度工程实践,为流域梯级复杂水电站群的分散决策和现代化管理提供了科学的决策依据。主要研究工作和创新性成果如下:针对现有多目标进化算法在求解流域梯级联合调度等复杂大规模优化问题时存在计算复杂度高、非劣解多样性差和难以处理复杂约束条件等问题,提出了一种新的多目标进化算法——自适应网格多目标粒子群优化(AG-MOPSO)算法,其核心内容包括非劣解密度自适应网格估计算法及其基于非劣解密度信息的Pareto最优解搜索方法和非劣解多样性保留方法。通过对典型测试函数的计算表明,AG-MOPSO算法处理大规模复杂优化问题时在收敛性能、计算效率等方面,较有代表性的多目标进化算法有不同程度的改善。针对传统调度决策方法不能有效处理梯级联合调度决策问题中的不确定性因素、缺乏柔性和鲁棒性差等缺点,系统地研究并提出了集对分析联系数相似程度刻画、排序以及基于集对分析多属性决策的理论与方法,导出了联系数贴近度函数的若干典型计算公式,获得了一种基于相对可能势的联系数排序方法和基于联系数决策矩阵的多属性决策模型,研究成果不仅丰富了集对分析的基础理论,而且为不确定多属性决策问题的解决提供了有力的工具。提出了基于AG-MOPSO算法的梯级电站多目标调度模型及其求解方法,通过将梯级电站联合调度中各种约束条件映射到决策变量的可行域,从而使得复杂约束优化问题转换为无约束优化问题,提高了模型求解的收敛效率,使基于AG-MOPSO算法的梯级电站多目标调度优化的工程应用成为可能。以三峡梯级电站调度问题为应用背景,得到了相应调度问题的非劣调度方案集,为梯级电站的多属性决策提供数量依据。在提出的广义集对分析多属性决策理论与方法的基础上,根据梯级水电系统的运行环境、运行时期以及决策层次的要求,实现对梯级电站多目标调度方案集的排序优选,并通过不确定演化因子对决策结果的敏感性分析,找出影响调度决策的关键因素,进而获得决策者最终满意的调度方案,为梯级电站多目标调度决策问题的解决提供了一条有效的新途径。更多还原

【Abstract】 In uncertain electric power market environment, the issue of cascade optimal regulation and decision making is a large scale, dynamic, nonconvex and nonlinear discrete multi-objective decision-making problem, with the constraint conditions of electrical market trading rules, hydrological cycle, generation control, power system security and reliability, power demand and consumers’ reaction, which is more complicated than the traditional optimal regulation of cascade hydropower system. Although many researchers have been devoted to finding the effective solutions to the above problems, the mutual conflict of complex objectives and the coupling relations between restrictions in valley cascade hydropower system make the problem description and model solving very difficult, which hardly have satisfying solutions and the new improved theories and reliaztions are in urgent need. Consequently, the study of multi-objective decision-making theory and method in complex cascade hydropower system is always the hot issues in academic frontier. The thesis makes a thorough study on the optimal regulation and decision-making theory and method of valley cascade hydropower stations by analyzing the complicacy in valley cascade hydropower system and adopting the complex system theory and modern intelligence evolutionary method. Focusing on the established multi-objective optimal regulation models of valley cascade hydropower system, multi-objective particle swarm optimization algorithm based on adaptive grids and multi-attribution decision-making approach based on Set Pair Analysis are brought forward for the sake of proving the validity of the promoted optimal algorithms when dealing with large scale, multiple objective and complex multiple constraint conditions optimization problem, which also exhibit the predominance of the decision-making approach based on Set Pair Analysis in the uncertain multi-attribution decision-making problem and develop the theory of valley cascade hydropower system optimal regulation and decision making The investigation results are successfully applied in the multi-objective optimal regulation of Three Gorges cascade, which provide scientific basis for decentralized decision and modern management of valley cascade hydropower stations. The study work and innovations are listed as follows:Multi-objective Particle Swarm Optimization based on adaptive super grids (AG-MOPSO) is proposed for the purpose of overcoming the the defects in the existing multi-objective evolutionary algorithms, such as high computational complexity, bad solution diversities and difficulties in dealing with complicated constraints while solving cascade optimal regulation and decision-making problems,.Compared with some representative multi-objective evolutionary algorithms on a set of well-designed test functions, the presented algorithm has stable convergence, good properties of dealing with complex large scale optimization problem and preferable diversity of solutions by employing the methods of density information estimation in Pareto set, pruning Pareto set, and Pareto optimal solution searching based on adaptive super grids algorithm.Aiming at dispelling the limitations of the conventional decision-making methods when dealing with the flexibility, robustness and uncertainties, this thesis investigates into the theories and methods of the matching-degree depiction, ranking methods of connection numbers and the multi-attribution decision-making approach based on generalized Set Pair Analysis. Several typical calculation methods of the interrelated function, ranking methods based on relatively certainty probability power, and multi-attribution decision-making approaches based on connection number decision matrix are developed in the thesis. The contents mentioned above not only develop the basic theory of Set Pair Analysis, but also supply effective tool to solve uncertain multi-attribution decision-making problems.The thesis establishes and implements the model of the multi-objective optimal regulation of cascade hydropower stations based on AG-MOPSO. It employs water level as the decision variable in AG-MOPSO, and converts the problem of solving cascade regulation into unconstrained optimization problem by calculating the water level range of reservoirs in each period of time, which cuts down computing cost and improves convergence rate due to the guaranteeing process of evolution in feasible regions. The non-inferior solution set of the problem of Three Gorges cascade multi-objective regulation are obtained in the end to suply the data preparation for the final multi-attribution decision.Based on the proposed the multi-attribution decision-making approach based on Set Pair Analysis, the thesis realizes preferred ranked choice and order for the multi-objective regulation schemes of cascade hydropower stations, finds out the key factors influencing the decision making results by the stability analysis of the uncertain evolutive factor, and obtains the satisfactory regulation scheme in the end, which meets the requirements of operation environment and the different hierarchy of the decision making, and provides a new effective way to solve the problems of multi-objective regulation and decision making of cascade hydropower stations as well.更多还原