粗糙集在控制燃煤电厂NO_x排放中的应用研究

发布时间：2018-03-23 22:33

本文选题：NO_x排放　切入点：粗糙集　出处：《华北电力大学》2015年硕士论文

【摘要】：随着环境污染的日益严重,对NO_x排放的控制越来越严格。控制NO_x排放的方法有很多,但是如果能够在燃料的燃烧阶段就减少NO_x的生成,会使NO_x的控制更加经济高效。本文首先对降低燃煤电站锅炉NO_x排放的几种研究方法进行了对比,指出目前对NO_x的控制主要包括燃烧优化和烟气吸收两种方法。相比于用神经网络等人工智能方法对燃煤锅炉进行燃烧优化,粗糙集无需建立复杂的数学模型,也无需现场的燃烧调整实验数据,只需要用简单的约简算法对运行历史数据进行挖掘,就能得到与NO_x排放值相关参数的优化运行区间,从燃烧阶段减少NO_x的生成。然后本文介绍了基于分辨矩阵的改进属性约简算法和基于核值的值约简算法。经过改进的属性约简算法不需要进行复杂的逻辑运算,便于用程序语言进行实现,且用于存储分辨函数的内存大大减少,提高了程序的运行效率。值约简算法能够对属性约简后的规则进行进一步的精简,使最终得到的规则数少而精。同时本文用可视化集成开发环境VC++6.0对该算法进行实现,使挖掘过程和得到的知识更加直观。用某机组的运行数据对该算法进行验证,最终得到的结果表明该算法能够获得正确且精炼的规则集。最后用该算法对某300MW机组200MW负荷下的NO_x排放数据进行挖掘。首先确认与NO_x排放相关的优化参数即待挖掘条件属性,得到待挖掘的决策表。在对数据预处理和离散化后,用上述属性约简算法和值约简算法进行挖掘。对约简后得到的排烟温度、燃烧器摆角、各层二次风风门开度、各引送分机风量等11个属性进行相关性分析,结果表明得到的条件属性与NO_x排放值之间存在较强的相关性。对得到的规则集进行分析,结果表明在200MW负荷下,较小的底层二次风配风,较大的上层二次风配风,偏下的燃烧器摆角,较低的过量空气系数以及较低的排烟温度能够有效的将NO_x的排放控制在低水平。实际中指导运行优化时,选取NO_x排放区间为较低的规则,可以在降低NO_x排放值的同时兼顾锅炉效率。
[Abstract]:With the increasingly serious environmental pollution, the control of NO_x emissions is becoming more and more strict. There are many ways to control NO_x emissions, but if the generation of NO_x can be reduced at the combustion stage of fuel, It will make the control of NO_x more economical and efficient. Firstly, several research methods of reducing NO_x emission from coal-fired power station boiler are compared in this paper. It is pointed out that the control of NO_x mainly includes two methods: combustion optimization and flue gas absorption. Compared with artificial intelligence methods such as neural network for combustion optimization of coal-fired boilers, rough set does not need to establish complex mathematical models. It is also necessary to mine the running history data by simple reduction algorithm, and to get the optimal operation interval of the parameters related to the emission value of NO_x, it is not necessary to adjust the experimental data of combustion on the spot, and only the simple reduction algorithm is used to mine the running history data. In this paper, the improved attribute reduction algorithm based on resolution matrix and the value reduction algorithm based on kernel value are introduced. The improved attribute reduction algorithm does not need complicated logic operation. It is easy to implement in program language, and the memory used to store resolution function is greatly reduced, which improves the efficiency of the program. The value reduction algorithm can further simplify the rules after attribute reduction. At the same time, the algorithm is implemented by visual integrated development environment (VC 6.0), which makes the mining process and the acquired knowledge more intuitionistic. The algorithm is verified with the running data of a certain unit. The final results show that the algorithm can obtain correct and refined rule set. Finally, the algorithm is used to mine the NO_x emission data under the 200MW load of a 300MW unit. Firstly, the optimization parameters related to NO_x emissions are confirmed, that is, the conditional attributes to be mined. After data preprocessing and discretization, the above attribute reduction algorithm and value reduction algorithm are used to mine the decision table to be mined. The correlation analysis of 11 attributes, such as the air volume of each extension, shows that there is a strong correlation between the conditional attribute and the NO_x emission value. The rule set is analyzed, and the result shows that under the 200MW load, there is a strong correlation between the obtained conditional attribute and the NO_x emission value. Smaller lower secondary air distribution, larger upper secondary air distribution, lower burner swing angle, lower excess air coefficient and lower exhaust temperature can effectively control NO_x emissions at low levels. When the NO_x emission interval is chosen as the lower rule, the boiler efficiency can be taken into account while reducing the NO_x emission value.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：X773

【相似文献】