【作者】 陈坚红；

【导师】 岑可法；

【作者基本信息】 浙江大学 ， 工程热物理， 2004， 博士

【摘要】 论文研究了大型多轴布置的燃气—蒸汽联合循环机组运行计划和负荷分配问题。涉及的内容有大型多轴布置的燃气—蒸汽联合循环机组关键部件(燃气轮机、余热锅炉和蒸汽轮机)的建模：水、水蒸汽和燃气的热力学性质通用计算模型；联合循环机组的变工况性能计算方法；根据自动发电控制(AGC)的实时调度负荷对联合循环机组的负荷进行在线最优分配以及多种复杂约束条件下联合循环机组运行计划和负荷分配问题的启发式遗传算法求解。论文最突出的贡献是综合应用了机理分析建模、基于小脑模型(CMAC)神经网络的建模、融合机理和CMAC神经网络的混合建模、数据挖掘技术在建模中的应用、用于数据挖掘建模的数据处理算法和自适应启发式遗传算法等多种方法，以满足问题求解工作的需要。作者的主要工作和创新点有： 1、研究了燃气—蒸汽联合循环机组关键部件建模这样一个逆向工程课题，建立了融合机理和CMAC神经网络方法的燃气轮机、余热锅炉和蒸汽轮机的数学模型。首先进行机理分析，建立机理模型：其次，在系统论述小脑模型神经网络原理的基础上，推导了小脑模型神经网络的概念映射算法、物理映射算法、输出算法和学习算法，并把小脑模型神经网络引入建模过程中，建立了基于小脑模型神经网络的燃气轮机、余热锅炉和蒸汽轮机的模型；最后，在启发式知识的引导下，将机理模型和基于CMAC神经网络的模型有机地融合，建立了混合模型。 2、论文系统地阐述了水、水蒸汽和燃气的热力学性质通用计算模型，并采用面向对象的程序设计方法，在计算机中编程实现，大大提高了计算程序模块的可靠性、可重用性、可扩充性以及方便性。 3、基于融合机理和CMAC神经网络方法的燃气轮机、余热锅炉和蒸汽轮机的数学模型，进行了联合循环机组的变工况性能计算，并得出了一些相关的结论。 4、研究了以联合循环机组的变工况性能计算为基础，根据大气环境温度、压力和自动发电控制(AGC)的实时调度负荷，对联合循环机组的负荷进行在线最优分配。 5、针对大型多轴布置的燃气—蒸汽联合循环机组运行计划和负荷分配问题，论文阐述了计及多种复杂约束条件的问题数学模型，提出解决问浙江大学博士学位论文摘要题的自适应、启发式遗传算法。目标就是在一个运行周期内使包含启动耗量、停机耗量和正常运行耗量等在内的花费最小化。基于启发性知识，构造若干新的有效的遗传操作算子，采用改进的算法控制策略，有效地克服局部极小值。最后，测试了应用自适应、启发式遗传算法实现的软件系统，结果表明，可以有效地求解机组负荷分配和运行计划这样一个NP完全问题。 6、首次系统提出和分析了用于数据处理过程中的数据融合理论和方法。把数据融合技术应用于关键部件建模，可以使所建的模型能更完善、更准确地反映实际情况。文中详细推导了数据融合算法，这种数据融合方法计算简便，可以反映传感器在空间或时间上的冗余或互补的信息，获得比有限个传感器的算术平均值更准确的测量结果，具有较高的可靠性，实际应用结果也证实了算法的准确性。 7、由于燃气轮机、余热锅炉和蒸汽轮机的非线性特性，联合循环机组关键部件的模型也必须是非线性的。论文在建模过程中采用了与以往完全不同的基于联合循环电厂实时数据库的数据挖掘的建模方法。这种新方法采用数据挖掘技术发现联合循环电厂实时数据库中参数之间的内在关系，基于这些内在关系和启发性机理知识，建立融合机理和CMAC神经网络的混合模型。 关键词:联合循环，运行计划，负荷分配，建模，机理模型，混合模型，小脑模型神经网络，水、水蒸汽热力学性质，燃气热力学性质，变工况性能计算，自动发电控制，在线，自适应，启发式遗传算法，数据融合，数据挖掘，数据处理更多还原

【Abstract】 This dissertation is mainly focused on the operation schedule and load dispatch of heavy-duty multi-shaft combined cycle generating unit. It involves key components ( including gas turbine, heat recovery steam generator and steam turbine ) modeling; general calculating models of thermodynamic properties for water, steam and gas; off-design performance calculating method for combined cycle generating unit; on-line optimum load dispatch for heavy-duty multi-shaft combined cycle generating unit according to the real-time dispatching load of automatic generation control; and heuristic genetic algorithm solving for operation schedule and load dispatch of combined cycle generating unit with multiple complex constraint condition. The most important contribution this dissertation presents is that multiple methods have been comprehensively applied so as to meet the needs of problem solving. The methods include modeling by mechanism analysis; modeling based on cerebellar model articulation controller ( CMAC ) neural networks; hybrid modeling combined mechanism analysis and CMAC neural networks; data mining technique and its applications in modeling; data processing algorithm used in modeling applied data mining technique; self-adaptive heuristic genetic algorithm; etc. The main work and innovations in this dissertation are as follows:1. The reverse engineering subject of combined cycle generating unit’s key components modeling is discussed. The mathematic models for gas turbine, heat recovery steam generator and steam turbine combined mechanism analysis and CMAC neural networks are established. Firstly, the mechanism is analyzed and then mechanism models are established; secondly, on the basis of discussing the principles of CMAC, the author deduces the conceptual mapping algorithm, the physical mapping algorithm, the output mapping algorithm, and the learning algorithm of CMAC, then, introduces CMAC in the procedure of modeling, establishes the mathematic models for gas turbine, heat recovery steam generator and steam turbine based on CMAC; lastly, with the guidance of heuristic knowledges, the mechanism models andthe mathematic models based on CMAC are combined some hybrid models.2. The general calculating models of thermodynamic properties for water, steam and gas are systematically discussed in the dissertation. Then, the software is implemented and developed by adopting object oriented program design method, and the reliability, expandability, convenience are greatly improved.3. Based on the mathematic models for gas turbine, heat recovery steam generator and steam turbine combined mechanism analysis and CMAC neural networks, off-design performance calculation of combined cycle generating unit is performed, and some correlative conclusions are educed from the off-design performance calculation.4. On the basis of off-design performance calculation of combined cycle generating unit, on-line optimum load dispatch for heavy-duty multi-shaft combined cycle generating unit according to the ambient temperature, ambient pressure and real-time dispatching load of automatic generation control is studied.5. For the operation schedule and load dispatch of heavy-duty multi-shaft combined cycle generating unit, this research attempts to formulate a mathematical model for the operation schedule and load dispatch problem and propose a self-adaptive heuristic genetic algorithm to solve the problem. Multiple complex constraint conditions are considered in the mathematical model of the problem. The objective in this dissertation is to minimize the total cost among an operation period that consists of the start-up cost, the shutdown costs, and the normal operation cost etc. Based on the heuristic knowledge, some new and effective genetic operating operators are constructed, a modified algorithm control strategy is adopted, in order to overcome local minimum problem. Finally, the implemented software system applied self-adaptive heuristic genetic algorithm was tes更多还原