燃煤电站锅炉经济运行与污染排放多目标优化研究

Economic operation and pollution emissions control of coal-fired boiler in power plant is the two main themes of the thermal power enterprises, economic and efficient operation of coal-fired boiler is the requirements of enterprises, and pollution emission control is social requirements, both have the uniform side and contradictions side, how to achieve both the economic operation and lower pollution emission of coal-fired boiler in power plant overall optimization is a major problem of thermal power companies facing currently..The goal of this research project is to get a set of Pareto solutions on the problem of multi-objective optimization of coal-fired boiler combustions in power plant. Modeling and optimizing are two important steps, the first stage of the study was to model the functional relations between outputs (NOx emissions & efficient of the boiler) and inputs(operational parameters of the coal-fired boiler ),which is called the combustion characteristics model of the coal-fired boiler. Based on the field test data of coal-fired bolier under the state of cold and hot conditions, combined with the historical operation database of coal-fired boiler in power plant,taking into account the boiler structure and coal-fired boiler combustion mechanism, this research project intends to establish a gray box model predicting the functional relationship between outputs (NOx emissions & efficient of the boiler) and inputs(operational parameters of the coal-fired boiler ) of a coal-fired boiler by artificial neural network approach..In the second stage of this research project, with the combustion characteristics model of the coal-fired boiler established in the first stage, advanced multi-objective optimization algorithm such as NSGA-II and DEA was introduced to optimize the operational parameters of the coal-fired boiler.According to the existing problems of NSGA-II and DEA, some improvement research will be performed taking into account the characteristics of the multi-objective optimization problem of the coal-fired boiler combustion, a set of Pareto solutions supplied by this research project can be a good tool to increase the efficiency of the boiler and reduce the NOx emissions simultaneously. .This research can provided a reference for the economy of coal-fired boiler operation and pollution emissions integrated decision-making to achieve in both energy conservation and emission reduction for thermal power enterprises. Studies on the artificial neural network modeling, evolutionary multi-objective optimization algorithm, especially their application in the coal-fired boiler will be performed in-depth in this research.

燃煤电站锅炉经济运行及污染排放控制是火电企业的两大主题，燃煤电站锅炉经济高效运行是企业效益的要求，而污染排放控制是社会效益的要求，两者有统一的一面又有矛盾的一面，如何达到燃煤电站锅炉经济运行与排放污染的综合优化是火电企业目前面临的一大难题。.本研究项目拟在燃煤电站锅炉现场冷热态试验数据与历史运行数据库数据的基础上，结合及锅炉自身结构及燃烧机理，引入人工智能神经网络方法建立电站燃煤锅炉的热效率及污染物排放的灰箱模型，进而建立燃煤电站经济运行与污染排放的综合多目标优化模型，利用进化多目标算法对燃煤电站锅炉进行多目标整体优化，并对优化算法应用于锅炉燃烧多目标优化中存在的问题进行针对性的改进，为燃煤电站锅炉经济运行与污染排放综合决策提供参考依据，达到火电企业既节能又减排的综合优化目的。本项目将在神经网络建模、进化多目标优化算法等人工智能方法应用于燃煤电站锅炉进行系统的研究。

本项目实施期间，项目组按照国家基金管理规定及时提交项目进展报告，汇报项目研究进展情况。.在燃煤电站锅炉燃烧特性模型开发、多目标优化算法、数学建模与模拟、理论研究等方面取得预期成果，基本完成了项目计划任务书中的内容:本项目在燃煤锅炉燃烧效率及NOx排放特性预测模型的建立中，尝试比较了基于分别采用BP神经网络方法、径向基（RBF）神经网络方法、Elman神经网络方法和支持向量机（SVM）的预测模型，其中BP神经网络模型及支持向量机（SVM）的预测模型效果较为稳定，锅炉热效率的训练的最大相对误差小于0.176%，NOx排放量的训练的最大相对误差小于3.312%。在前面的锅炉燃烧特性预测模型基础上，结合不同的多目标优化算法上，建立了基于权重分配的GA多目标优化模型，基于改进的NSGA-II算法的锅炉燃烧多目标优化模型，基于BP-VEGA模型的燃煤电站锅炉燃烧多目标优化模型，基于分解的多目标进化算法（MOEA/D）燃煤电站锅炉多目标优化模型。针对NSGA-II在燃煤锅炉燃烧多目标优化问题应用中Pareto解集分布不理想、易早熟收敛的问题，在拥挤算子及交叉算子上进行了相应改进，优化结果表明，改进NSGA-II方法与BP神经网络模型结合可以对锅炉燃烧实现有效的多目标寻优、得到理想的Pareto解集，是对锅炉燃烧进行多目标优化的有效工具，同改进前的NSGA-II优化结果比较，其Pareto优化结果集分布更好、解的质量更优。通过计算机在线指导锅炉配风、配煤等燃烧运行调整，达到提高锅炉运行效率的同时减小NOx排放的目的。