【作者】 周婷；

【导师】 纪昌明；

【作者基本信息】 华北电力大学 ， 管理科学与工程， 2014， 博士

【摘要】 能源是人类生存和发展的重要物质基础,攸关国计民生和国家安全。水电作为目前开发规模庞大、开发技术最为成熟的可再生能源,以其良好的调节性能、低廉的运行成本和快速的负荷响应能力,在世界电力能源格局中发挥着重要作用。我国水力资源丰富,为经济社会发展提供了能源保障。加快开发水能资源是我国增加清洁能源供应、优化能源结构、应对世界气候变化、实现可持续发展的重要措施。“十二五”时期是我国全面建设小康社会的关键时期,从我国的能源特点和自然资源结构来看,加快水电发展也是实现2020年非化石能源目标的必经之路,也是有效降低单位GDP二氧化碳排放量的重要措施。水库调度是水库运行管理的重要环节,调度水平直接影响着水库水电站综合效益的发挥。合理优化的水库调度方式能够在不增加硬件投入的情况下,获得可观的社会效益和经济效益,也是优化能源结构、促进节能减排的有效措施。本论文在对水库水电站群隐随机优化理论回顾归纳的基础上,分别从确定性优化调度模型建立与求解、调度规则的制定与优化、基于调度规则的水库水电站群系统仿真、效益评价及隐随机优化调度因素影响分析等方面对径流不确定条件下的水电站群优化调度进行研究。主要研究工作包括：(1)水库水电站群隐随机优化调度理论研究及归纳。介绍了水库确定性优化调度和随机调度的概念和特征,从径流过程角度分析二者之间的区别和关系。在分析显随机优化调度和隐随机优化调度原理的基础上,重点综述隐随机优化理论方法的国内外研究进展及其在水电站水库调度规则制定中的应用,并总结各种调度规则制定方法的适用条件和优缺点。(2)基于网格搜索和交叉验证的改进支持向量机模型研究。基于支持向量机方法的原理分析其在回归预测领域的优势,针对支持向量机对参数敏感和小样本回归易受训练样本随机性影响的特点,建立基于网格搜索的参数寻优机制和基于交叉验证的样本随机性规避机制,对支持向量机性能进行改进。通过实例研究,验证改进机制对支持向量机在小样本训练拟合能力和预测能力方面的效果。(3)基于C++和MATLAB的水库水电站群混合编程仿真平台的建立。针对隐随机优化调度在实际运行中的实现难度,考虑隐随机优化调度模型复杂、计算机实现环境多样化的特点,以支持向量机理论为例,将基于MATLAB的调度决策生成算法预测编译为动态库文件,使其在基于C++的水库水电站群系统仿真程序中被调用,实现实时滚动模拟。通过案例应用,对仿真平台的结构及系统稳定性和可扩展性进行评价。(4)金沙江中下游12级梯级水电系统隐随机优化调度研究及其效益评价。以我国十三大水电梯级中规模最大的金沙江中下游梯级水电站系统为例,以系统发电量和保证出力为优化目标,建立并求解梯级中长期确定性优化调度模型,作为隐随机模型的训练样本。运用改进支持向量机方法对系统调度规则制定,并模拟系统1989-2000年运行过程。另基于多元逐步回归法制定调度规则并仿真,将同期确定性优化调度结果及两种仿真结果进行对比。对仿真结果的发电量、发电过程、保证出力等方面进行对比,分析仿真结果的效益和可靠性。(5)隐随机优化调度模型因素影响研究。定量研究梯级规模、径流预报误差、模型参数、输出决策等因素对梯级水电站群隐随机优化调度仿真结果的影响。基于金沙江下游——长江中游大型梯级水电系统,以其宗单库、其宗——向家坝12级和其宗——葛洲坝14级三种电站组合为研究对象,控制各影响因素变化范围,并分别进行仿真运行和效益评价。评价结果所揭示的各因素所带来的影响方式对于支持向量机理论的改进以及隐随机优化调度的下一步发展有着重要的参考价值。更多还原

【Abstract】 As one of the most important basis for human survival and development, energy is at stake of national economy and security. Hydropower, which is currently the most massive and technique mature renewable energy, has its unique advantage for its good regulating performance, low running costs and fast load response capability, thus plays an important role in the world of electrical energy pattern. Hydropower potential is abundant in China, which provides vital energy security for economic and social development. Accelerating hydropower development is important measure to increase our supply of clean energy, optimize energy structure, respond to climate change and achieve sustainable development. For the twelfth five year plan period is a critical stage of building our moderately prosperous society, from the characteristics of China’s natural resources and energy structure, accelerating hydropower development is not only the only way to achieve the2020non-fossil energy consumption goals, but also important measure to effectively reduce the carbon dioxide emissions per unit of GDP.Reservoir operation plays an important role in reservoir operation and management, which directly affects the overall efficiency of hydropower systems. Reasonable and optimal reservoir operation is able to gain considerable social and economic benefits without increasing any hardware investment, which is a key manner to optimize energy structure and energy conservation. This paper takes implicit stochastic optimization theory as basis, comprehensively researches on the deterministic reservoirs operation model and solution, derivation and optimization of operating rules, hydropower system operation based on operating rules, benefit evaluation and factors assessment. The major achievements are as following:(1) Implicit stochastic optimization theory analysis for hydropower operation. The concept and characteristics of deterministic optimal operation and stochastic optimal operation as well as their differences and relationships. Based on analysis of explicit stochastic optimization and implicit stochastic optimization, this paper reviewed research progress of implicit stochastic optimization theory and its application operating rules derivation. summarized applicable conditions, advantages and disadvantages of various ISO methods.(2) Support Vector Machine research based on grid search and cross validation. The advantages of Support Vector Machine are analyzed with its principle. Considering the parameter sensitivity feature and the model vulnerability caused by small-size training samples, grid search and cross validation mechanisms are established to improve the SVM performance. The effect of improved Support Vector Machine en regression and prediction precision is validated by case study.(3) Hybrid programming simulation platform for reservoirs and hydropower stations based on C++and MATLAB. Considering the difficulties of implicit stochastic optimization implementation in actual operation,and the simulation model complexity of implicit stochastic optimization and diversity in computer implementation languages, this paper takes SVM for example, compiled the operating rules derivation algorithm into dynamic library files, which are then called by reservoir simulation platform to implement real time rolling simulation. Based on case study, the stability and scalability of simulation system of system structure is evaluated.(4) Implicit stochastic optimization of Jinsha mid-downstream twelve cascaded hydropower system and benefit evaluation. Based on the largest hydropower system of China’s thirteen hydropower bases, deterministic optimization model with maximum electric production and maximum firm output is established and solved as training samples for ISO. System operating rules is established using improved SVM and simulated during the year1989-2000. Additionally, the system is simulated independently using multiple progressive linear regression and simulated. Three operation results are compared and discussed. According to comparisons on electric production, output process and firm output, the benefit and reliability is assessed comprehensively of simulation results.(5) Factors analysis of ISO model. The influence of cascade scale, inflow forecast error, mdel parameters and decision output of operating rules are analyzed. Based on Jinsha River-middle Yangtze hydropower system, three cases are formed Qi Zong single reservoir, cascade system with twelve reservoirs from Qi Zong to Xiang Jiaba and cascade system with fourteen reservoirs from Qi Zong to Ge Zhouba. By controlling the range of each factor, different scenarios are simulated and evaluated. The effect of each factor revealed by simulation results provides vital reference value for SVM theory improvement and future ISO research.更多还原