发布时间：2015-01-24 来源：人大经济论坛

通信工程专业论文 目 录 第一章 绪论 1 1.1课题的来源与概述 1 1.2研究的意义 1 1.3国内外研究的现状 2 1.4论文的主要内容 3 第二章 焦炉立火道温度软测量原理分析 4 2.1 立火道温度与蓄热室温度的关系 4 2.2 影响焦炉立火道温度的因素 7 2.2.1 煤气换向操作 7 2.2.2 推焦串序 8 2.2.3 结焦时间 9 2.2.4 配合煤水分 10 2.2.5 煤气种类与煤气热值 11 2.2.6 大气温度和风向影响 11 2.3 影响因素与立火道温度的相关性分析 11 2.4 小结 13 第三章 基于线性回归的软测量模型 14 3.1 软测量方法 14 3.2 模型的学习机制 16 3.3 线性回归模型 17 3.3.1简单回归建模方法 17 3.3.2多元回归建模简介 19 3.3.3基于线性回归的软测量仿真模型 20 3.4 回归模型预测效果分析 23 3.5 小结 24 第四章 基于神经网络的软测量模型 25 4.1 神经网络的通用模型 25 4.2 神经网络建模方法 26 4.3 基于BP神经网络的软测量仿真模型 27 4.4 小结 30 第五章 结论与展望 31 5.1 结论 31 5.2 展望 31 致 谢 33 参考文献 34 附 录 36 摘 要 在焦炉加热控制系统中，立火道温度是其中重要的工艺参数。对立火道温度控制的优劣直接关系到焦炭的质量和产量以及焦炉的使用寿命。但是由于立火道温度很高，采用热电偶直接检测不仅热电偶寿命很短而且成本和维护费用很高。因此对立火道温度的检测一直是困扰人们的难题。 本文从软测量建模的角度入手，详细分析了影响焦炉立火道温度的各种因素，有如：蓄热室温度、煤气换向、推焦串序、入炉煤水分、煤气种类与热值、煤气流量、天气状况等。在原理分析的基础上，本文提出了立火道温度的软测量仿真模型，它以蓄热室顶部温度和配合煤水分为辅助变量，建立线性回归子模型和神经网络子模型分别对机焦两侧立火道温度进行预测。为了提高模型对立火道温度的预测精度，本文利用专家经验建立专家规则库，并对软测量子模型进行模糊组合。仿真实验的结果表明，这种组合的策略是可行的。 关键词： 焦炉；软测量；线性回归；神经网络；模糊组合 ABSTRACT As the quality of the coke and the life-span of the coke oven are connected closely with the coke oven flow temperature, the temperature of coke oven flow is one of significant technical parameters in the coke oven heating system. However, it is difficult to detect the coke oven flow temperature as the high temperature always makes the thermocouple out of use and consuming much in the thermocouple. And that has been a difficult problem which puzzles people for a long time. In the view of technical mechanism, this paper analyzed several factors associated with the coke oven flow temperature, including the temperature in the regenerator, the reverse operation of the coal gas, the operation of coke pushing, the percentage of water in the coal, the coal gas type and heat quality, the gas flowing flux and the weather and etc. Following a kind of soft sensor model was put forward, which can be divided into two sorts of sub-model, linear regress model and neural network model. The temperature at the top of the regenerator and the water percentage in coal were selected as the variables, and the flow temperature of double sides of the oven were predicted by the sub-models. In order to improve the precision of the model, the paper constructs the rules library based on the experts’ experiences. The method is valid and feasible which are proved by the simulation. KEY WORDS: coke oven; soft sensor; linear regress; neural network; fuzzy combination