|Thesis Advisor||谭杰 ; 王学雷|
|Place of Conferral||北京|
|Keyword||炼焦烟气 脱硫脱硝 换向过程 运行优化 数据驱动|
（4）针对复杂运行条件下炼焦烟气脱硝过程的实时操作优化问题，本文提出一种数据驱动的脱硝过程操作参数实时优化方案，整个优化过程分为两步：首先基于案例推理（Case Based Reasoning, CBR）方法及反馈调节器得到优化指标的预设定值，考虑到传统案例重用方法的缺点，提出了基于主成分回归（Principal Component Regression, PCR）-多案例融合的案例检索与重用方法；然后提出一种基于时间有序性即时学习（Time-order Just-in-time Learning, T-JITL）算法的局部非参数建模方法，将此模型作为优化约束条件，利用优化算法进一步求解优化问题得到最优值。该方法不基于任何参数模型，同时充分考虑了炼焦生产工况的复杂性、时序性，解决了炼焦烟气脱硝复杂工况所导致的稳态模型难以获取的问题，完全基于数据得到优化设定值。
|Other Abstract||With the deepening of atmospheric environmental governance in China, the new pollutants emission standard of coking industry puts forward stricter restrictions on coke oven emissions of SO2 and NOx. As a result, it is imminent for nearly 1000 domestic coking enterprises to implement coking flue gas treatmemnt so as to meet the national standard. In view of this increased pressure, a domestic coking group has taken the lead in building and operating an coking flue gas desulfurization and denitration integrated device, however, not only the flue gas purification effect can not be satisfied, enormous waste of energy also exists due to the manual operation mode at the present stage. In this dissertation, we take the coking flue gas desulfurization and denitration integrated device as the research object, the modeling, operation optimization and control strategy of denitration process is studied in consideration of the particularity and complexity of coking production, aiming to form a efficient, reliable and economical matching control technology for coking flue gas denitration process, promoting the industrialization of wet desulfurization and denitration technology as well as the flue gas comprehensive governance of coking industry in China. More specifically, the main content of this paper has been concluded as bellow:|
(1) The coking flue gas desulfurization and denitration process is analysed. The modeling and operation optimization of the process are based on fully understanding of the process characteristics, however, there is no mechanism research of the coking flue gas wet desulfurization and denitration process. The main influencing factors of the denitration process are determined first using the existing research results of similar processes and some experiments. Then the difficulties, operation status and existing problems of the process are analysed. Finally, the overall structure of the optimization control of coking flue gas denitrification process is proposed, pointing out the main research contents of the following chapters.
(2) The current condition of inlet flue gas indices cannot be determined timely and precisely by experience owing to the large detection lag and complex upstream coking process. In order to solve this problem, we take the heating combustion process of the coke oven in the coking plant as the background, an state estimation method based on the flue gas soft sensor model is proposed. The mechanism models are established firstly according to the principle of material balance and reaction kinetics, an improved neural network combining optimal stopping principle and dual momentum adaptive learning rate is proposed and used to compensate the prediction error. Based on the integrated model, we propose a steady-state value calculating method of the coking flue gas indices by using time windows and the combustion condition change criterion, laying the foundation for the coke oven reversing process control and operation optimization of the denitration process in the later chapters.
(3) A large amount of energy is wasted in the coking flue gas denitration process during the periodic coke oven reversing operation. In order to solve this problem, causes to NOx concentration change during this process are thoroughly analyzed, the disturbance modeling method of flue gas NOx concentration and control strategy of the ozone output are proposed: The flue gas NOx disturbance models are divided into the inlet and outlet disturbance models, the response curve of the pulse signal and a first-order inertial transfer function, which are connected in series is taken as the general model structure, the NOx disturbance models can be established after parameters being identified. Then the feedforward control system is designed, and control strategy according to the relation between the lowest inlet flue gas NOx concentration in reversing process and the limited value of outlet NOx concentration is proposed, which can minimize the ozone generating unit power consumption cost.
(4) For the real-time operation optimization problem under complex operating condition of coking flue gas denitration process, we propose a data-based operating parameter real-time optimization strategy of the denitration process, which is a two-step process. First, the preset operational indices are obtained by case based reasoning (CBR) method and a feedback regulator, a case reuse method based on principal component regression (PCR) multicase fusion is proposed to solve the shortcomings of the traditional case reuse method. Then a local nonparametric modeling method based on time-order just-in-time learning (T-JITL) algorithm is proposed, the model will be used as the optimization constraints, and the optimal value can be obtained by solving the established optimization problem. This scheme is not based on any parameter model, moreover with the complexity and time-order of coking production being taken into account, which can solve the steady-state modeling difficulties caused by the complex coking flue gas denitration conditions and obtain the optimized set points based on data.
(5) Using the methods proposed before, a prototype system for coking flue gas denitration process optimization based on the existing hardware and software platform is designed and developed. The application verification results in industrial field show that the method proposed in this paper has strong practicability and can realize stable and economic operation of the denitration process.
|First Author Affilication||Institute of Automation, Chinese Academy of Sciences|
|李亚宁. 炼焦烟气脱硝过程建模与运行优化控制方法研究[D]. 北京. 中国科学院研究生院,2018.|
|Files in This Item:|
|Recommend this item|
|Export to Endnote|
|Similar articles in Google Scholar|
|Similar articles in Baidu academic|
|Similar articles in Bing Scholar|
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.