复杂工况下焦炉加热燃烧过程在线优化控制方法

With the characters of coke oven combustion process considered, such as complex mechanism and broad range of operating conditions, this project makes the research on the online optimization control methods for coke oven combustion process. Starting with process mechanism analysis of coke oven combustion process, the operative conditions are classified reasonably. Coke oven includes multiple coking process subsystems with respect to coking chambers, which are time correlated, and directly affect operative conditions of combustion process. In this project, the real time operative conditions identification method based on multiple coking process subsystems is studied in order to get the actual operative conditions of coke oven. Based on local modeling idea and just-in-time learning method, a supervised modeling neighborhood is build according to global characteristic, which are reflected by status parameters related with process outputs. Thus, with multiple operative conditions, a local dynamic nonlinear modeling method based on the supervised modeling neighborhood is studied to improve the modeling precision and efficiency. To handle different operative conditions, optimization control models with different opertive conditions are setup. An optimization control method for coke oven combustion process is studied, in which both controller parameters optimization method based on guidance approaches using knowledge model, and online performance assessment technologies are investigated. Finally, the simulation of the methods and the online optimization control system with complex operative conditions built in industrial application are carried out to realize the optimization operation. This project focuses on an effective optimization control method for coke oven combustion process with complex operative conditions, presents a feasible solution scheme for coke oven temperature control problem，and provides new theories, methods and system application technology.

本项目针对焦炉加热燃烧过程机理复杂、工况范围大的特点，研究复杂工况下焦炉加热燃烧过程在线优化控制方法。从焦炉加热燃烧过程工艺机理入手，对工况进行合理划分，研究基于多个工作时序关联子系统的焦炉加热燃烧过程在线工况识别方法，实现复杂工况的在线识别；基于即时学习局部建模思想，通过与过程输出相关的状态参数反映过程的全局特性，建立监督式的建模邻域，从建模精度和效率出发，研究焦炉加热燃烧过程多工况局部动态非线性建模方法；建立多工况优化控制模型，研究基于知识引导的控制器参数智能优化方法，结合控制系统性能在线评估技术，实现炼焦生产过程的性能维护和运行优化；通过研究复杂工况下焦炉加热燃烧过程在线优化控制系统结构，进行仿真与工业应用，实现焦炉加热燃烧过程的优化运行。通过研究，提出一种解决焦炉温度控制问题可行的解决方案，为焦炉加热燃烧过程优化控制提供新的理论与方法以及系统实现和应用技术。

焦炭是高炉炼铁中主要的热量来源以及冶炼过程的还原剂，同时起到炉料骨架的作用，是高炉炼铁中不可缺少的炉料。稳步提高焦炭质量以适应高炉大型化的需要，对提高我国炼铁工业生产技术具有重要的意义。焦炉温度是焦炉加热燃烧过程中直接影响焦炭质量的参数，其控制对于获得高质量的焦炭具有重要意义。.本项目针对焦炉加热燃烧过程机理复杂、工况范围大的特点，研究复杂工况下焦炉加热燃烧过程在线优化控制方法，提出一种基于多子系统信息熵的焦炉加热燃烧过程工况识别方法。从焦炉热平衡的角度对工况进行划分，结合焦炉结构和传热机理将焦炉分解为多个近似独立的加热子系统，分析各个加热子系统的耗热水平；采用信息熵确定各类子系统对焦炉不同工况影响权重，实时判断焦炉加热燃烧过程工况。针对焦炉加热燃烧过程中多工况的特点，本项目提出一种基于主动半监督仿射传播聚类算法的焦炉加热燃烧过程多模型建模方法，结合主动学习策略寻找样本信息间的成对约束关系，并将其应用到仿射聚类算法中，调整相似度矩阵并改善聚类性能，然后与最小二乘支持向量机方法相结合，形成多模型建模方法，获得了更好的精度和适应性以及训练过程的快速收敛性。针对焦炉加热燃烧过程工况变化范围大，利用即时学习模型能够快速反应过程变化的特性，提出基于动态模型的优化控制方法。基于局部建模的思想，通过分析焦炉加热燃烧过程的机理特性，综合考虑样本间的距离与样本变化趋势夹角的信息，从历史数据中选择数据建立动态样本库，采用即时学习算法建立焦炉加热燃烧过程动态模型；采用差分进化算法对模糊控制器论域参数进行优化，以保证控制器的性能能够满足焦炉加热燃烧过程的动态变化。.通过研究，本项目提出一种解决焦炉温度控制问题可行的解决方案，为焦炉加热燃烧过程优化控制提供新的理论与方法以及系统实现和应用技术。