【作者】 刘涵；

【导师】 黄强；

【作者基本信息】 西安理工大学 ， 水文学及水资源， 2006， 博士

【摘要】 随着我国经济建设的迅速发展,大规模的水电站水库群逐渐形成,对其进行综合调度与运行管理变的越来越复杂,其地位和作用也越来越突出。本论文系统地介绍了水库调度的研究现状与发展趋势,将宏观系统方法论引入到水库调度问题中,探讨了水库综合利用的多种目标之间如何形成统一的、一体化的调度模式,研究了确定型水库优化调度以及随机型水库优化调度的新方法,并采用数据挖掘算法提取了水库调度规则。论文主要取得了以下成果:1.引入物理-事理-人理(WSR)系统方法论,系统分析了水库调度的WSR过程,并分别以发电、防洪为主要目标探讨了WSR的具体应用过程,重点体现了人的参与过程和协调作用。首次将物理-事理-人理(WSR)宏观系统方法论引入到水库调度问题中,是一种观念上的创新,有利于解决水库调度过程中可能出现的矛盾冲突现象,协调各种利益关系,以最终形成统一的、一体化的水库调度模式。2.群体智能算法具有并行化、鲁棒性强、正反馈等优点,本论文将改进的蚁群算法和粒子群算法两种随机启发式搜索算法引入到确定型水库优化调度问题中,应用局部搜索和全局搜索相结合的机制优化路径,提高了算法的搜索效率,容易找到全局最优解,避免了求解确定性问题时出现的“维数灾”问题。3.考虑到水库调度在很大程度上依赖于未来的径流情势,但目前对未来径流尚无法准确预知,因此,确定型水库优化调度不可避免地具有一定的局限性。本论文以序列运算理论为基础,从数学的角度推导出水库随机调度的序列化过程,建立了水库随机调度的序列运算理论。提出了水库随机调度过程中各变量的序列化方法,通过使用各变量对应的概率性序列进行卷和、卷差、交积等运算,完成系统内各水库资源-需求的加载过程,即水库水量平衡过程,指出了算法中各种运算结果的期望值所代表的明确物理意义,该理论将成为水库随机优化调度的基础性工具。4.以水库随机调度的序列运算理论为基础,提出应用遗传算法进行随机型水库优化调度的求解,为水库随机优化调度提供了一种全新的思路,有助于缓解水库随机调度中存在的“维数灾”问题。5.以关中灌区某供水水库为例,研究了应用数据挖掘方法提取水库调度规则的新算法。通过将供水水库调度规则概化为几种模式,从而将传统推求水库调度规则的问题转化为一个模式关系的数据挖掘问题。主要选取调度时段、入流、水库蓄水量、水文年型以及需水5个特征属性作为下一步数据挖掘的基本数据集,采用决策树算法进行了数据挖掘计算,并与传统的线性回归获得的调度函数结果进行了比较分析。结果表明数据挖掘方法用于水库调度规则的提取是合理有效的,其准确率更高。更多还原

【Abstract】 With rapidly economy development, large-scale reservoir group system is come into being. Its integrative operation become from complex to complex, its station and function become from prominent to prominent. Study status and development trend of reservoir operation are introduced, macroscopical system methodology is introduced into reservoir operation problem, how to forming integrative operation mode among reservoir operation multi-objects is discussed, new methods of deterministic and stochastic reservoir optimal operation are studied, and data mining algorithm is adopted to extract reservoir operation rules. The main results of this paper are as follows:1. Wuli-Shili-Renli system methodology (WSR) is introduced to analyze reservoir operation process. Generation electricity and flood control are made as main objects to discuss detailed application process, participation course and harmony function of human is embodied. Firstly WSR is introduced into reservoir operation study, this is a concept innovation. It is advantaged to solve conflict problem of reservoir operation, to harmonize all sorts of benefit relations, and to form integrative reservoir operation mode ultimately.2. Swarm intelligence algorithm has being strong points of parallel, robustness, positive feedback and so on. Two heuristic search algorithms of improved ant colony optimization (ACO) and improved particle swarm optimization (PSO) are adopted to study deterministic reservoir optimal operation, combined local with global search to increase algorithm efficiency, easy to find global optimal solution, avoided the curse of dimensionality of deterministic reservoir operation.3. Considered reservoir operation depending on the future runoff state to a great extent, but without methods exactly forecasting the future runoff now. So there is some localization in problem of deterministic reservoir optimal operation. Based on sequence operation theory, from the point of mathematics sequence process of stochastic reservoir operation is deduced, sequence operation theory of stochastic reservoir operation is established. Sequence method of every variable in the process of stochastic reservoir optimal operation is put forward, probability sequence of every variable is adopted to make Addition-Type-Convolution (ATC), Subtraction-Type-Convolution (STC), AND-Type-Production (ATP) and so on, in order to complete resource-demand process of reservoir, which is reservoir water balance equation. Essential meaning of expected value of every operation results is provided. This theory will become basic tool of stochastic reservoir optimal operation.4. Based on sequence operation theory of stochastic reservoir operation, GA is adopted into solving the problem of stochastic reservoir optimal operation. It will provide a new thought for stochastic reservoir optimal operation and will be advantaged to solve the curse of dimensionality of stochastic reservoir operation.5. Case as water supply reservoir in Guanzhong irrigation area, data mining is adopted into reservoir operation rules extraction. Water supply reservoir operation rules are determined as several modes, so translating traditional reservoir operation rules extraction problem into data mining problem. Five characteristic attributes of operation period, inflow, reservoir storage, hydrologic year type and demand are chosen to make basic dataset of data mining. Decision tree is adopted to calculation, and compared to operation function results by linear regression. It shows that it is valid to adopt data mining extracting reservoir operation rules and it has higher exact efficiency.更多还原