【作者】 卢勇；

【导师】 徐向东；

【作者基本信息】 清华大学 ， 动力机械及工程， 2003， 博士

【摘要】 要改善我国目前能源利用效率低下、浪费严重的不利局面，其中一个重要的措施就是通过提高现有能耗设备的操作运行水平，在不增加更新设备投资的前提下，以实现最大限度的节能增效。本文课题研究的着眼点也在于此。本文第一部分为热工过程控制与数据信息采掘，研究的重点是通过先进控制策略的应用以及运行数据中发掘的有用信息，提高现有典型热工过程控制系统的性能，内容包括：首次提出了借鉴数据信息采掘技术的思路和方法，用于完善现有的热力系统优化控制方案；在此基础上结合生产实际，提出了运用多模型自适应策略实施锅炉负荷控制，并利用关联规则的数据信息采掘方法，通过对燃料发热量信息的在线诊断，优化控制器结构，同时实现锅炉燃烧配风的自适应调整，提高负荷、燃烧控制系统性能。提出了基于工况辨识的多线性模型多变量预测控制策略，其中包括热力系统的工况区域划分、辨识方法（基于运行数据信息采掘），以及在此基础之上的全新的多模型融合、控制器重构策略。为进一步提高系统性能，研究了多非线性模型预测控制，提出了一些非线性预测控制方法应用于热工过程控制的改进策略，并通过实时仿真研究，对方法可行性和有效性进行了验证。文章第二部分的研究内容主要集中在热工过程运行优化以及数据信息采掘技术在这一领域内的一些实际应用。首先，针对目前供热系统热源负荷控制中的难点问题，提出了通过综合分析系统运行数据，间接提炼出供热系统热负荷需求模式，并通过动态预测补偿的方法以实现真正意义上的按需供热。为解决锅炉变工况运行条件下的燃烧优化控制问题，利用数据信息采掘技术对锅炉变工况运行数据进行分析，提出了一种行之有效的基于工况分类、辨识和RBF静态神经网络模型的最佳氧量设定模块，并将其应用于生产实际，获得了良好的效果。更多还原

【Abstract】 Aiming at ameliorate the inefficient energy consumption method, improvement of the operation level is one of the most effective way in which energy can be saved to the maximum extent and additional investment for the substitution of equipment is unnecessary. This is the main focus of the research work of this dissertation. . The first part of this dissertation is about the automation control of the thermodynamic systems and the data mining technology applied in this field. The research objective is to improve the control system performance through the combination of the advanced control strategy with the meaningful information mined from the history operation database. Firstly, a new solution for the control performance enhancement is presented by the reference of the KDD technology. Aiming at practical application, a novel boiler load control method, which combines the multiple model adaptive control strategy with the data mining technology, is put forward in this dissertation. Through utilizing the association rules to identify the fuel heating characteristics, the auto-tuning of the air/fuel ratio in the boiler combustion control can be realized.Secondly, a thermodynamic system predictive control strategy based on the working regime identification online and multiple local linear models is presented, including a new working regime decomposition and identification method and the novel online multi-model combination and controller reconfiguration method. In order to upgrade the controller performance further, we extend our research work to the multiple nonlinear models predictive control and its application in the control of thermodynamic systems. Some amelioration strategies are presented to facilitate the practical application of the ordinary nonlinear predictive controller. Through the simulation test, the feasibility and superiority of these methods are proved.The second part of this dissertation mainly focuses on the thermal engineering process operation optimization and some applications of the data mining technology in this field. Firstly, aiming at solving the difficult problem in scheduling control of the heating load in a practical district heating system, a strategy is proposed which can realize the real balance between the quantity of heat needed for the resident and the quantity of heat provided by heating plant. This method is based on the heating load<WP=7>predictive model which is extracted from the practical operation data recorded in production field by the data mining tools, and is realized through the online model-based dynamic optimization.Finally, in order to improve the combustion efficiency of the boiler which is operated in a wide working condition range, a novel optimum oxygen content setting module is proposed by the statistic analysis of the operation data record, which is based on working condition classification, identification and a RBF static mapping function model. This setting module has been implemented in a practical combustion control system and is proved to be effective in fuel-saving.更多还原