燃烧过程复合建模与优化控制研究

发布时间：2018-05-11 11:39

本文选题：电站锅炉 + 最小二乘支持向量机　；参考：《华北电力大学》2014年硕士论文

【摘要】：目前火力发电在我国占主导地位,电站对煤的需求量持续增加,燃煤排放的污染物也在不断增加,因此电站锅炉燃烧优化成为一项重要的研究课题。电站锅炉燃烧优化主要分为两部分：电站锅炉燃烧系统模型是锅炉燃烧优化的基础,优化控制算法是锅炉燃烧优化的核心。锅炉燃烧系统机理复杂,利用燃烧机理建立的数值模拟模型运算速度十分缓慢,无法直接应用到锅炉燃烧优化中。机器学习建模运算速度较快,但是无法反应锅炉内部机理,而且模型受训练样本数据影响较大。结合机理建模与机器学习建模的优缺点,提出了一种以数值机理模型提供数据,并结合机组试验数据以及实时运行数据建立最小二乘支持向量机(LS-SVM)模型的复合建模方法。针对LS-SVM依赖建模样本数据的特性,数值机理模型提供的数据能够丰富LS-SVM模型的初始样本的多样性,提高模型精度。LS-SVM的快速性能够很好地满足燃烧优化控制的需求。结合特征选择提出了SFS-LS-SVM方法,能够精简样本数据,去除数据中冗余成分,提高建模速度。通过具体算例验证了变量选择的效果。根据LS-SVM算法提出了基于LS-SVM的模型更新算法。在复合建模中,模型更新算法用以更新机组的历史数据及实时运行数据,并且通过算例验证了建模效果。本文结合这两种优化方案的优缺点提出了一种以锅炉效率最大为目标的优化方案,并验证了优化效果。结合论文对建模与优化方案的研究成果,开发了燃烧优化控制系统。
[Abstract]:At present, thermal power generation plays a leading role in our country. The demand for coal in power stations is increasing continuously, and the pollutants emitted from coal combustion are also increasing. Therefore, the combustion optimization of power plant boilers has become an important research topic. The combustion optimization of utility boiler is divided into two parts: the model of boiler combustion system is the basis of boiler combustion optimization, and the optimization control algorithm is the core of boiler combustion optimization. The mechanism of boiler combustion system is complex, and the calculation speed of numerical simulation model established by combustion mechanism is very slow, which can not be directly applied to boiler combustion optimization. Machine learning modeling is fast, but it can not reflect the internal mechanism of boiler, and the model is greatly affected by the training sample data. Combining the advantages and disadvantages of mechanism modeling and machine learning modeling, this paper presents a compound modeling method of least square support vector machine (LS-SVM) model based on numerical mechanism model, unit test data and real-time operation data. According to the characteristics of LS-SVM dependent modeling sample data, the data provided by numerical mechanism model can enrich the diversity of initial samples of LS-SVM model, and improve the rapidity of model precision. LS-SVM can meet the needs of combustion optimization control. Based on feature selection, a SFS-LS-SVM method is proposed, which can simplify the sample data, remove redundant components from the data, and improve the modeling speed. The effect of variable selection is verified by an example. According to LS-SVM algorithm, a model updating algorithm based on LS-SVM is proposed. In the composite modeling, the model updating algorithm is used to update the historical data and real-time operation data of the unit, and the modeling effect is verified by an example. Combining the advantages and disadvantages of these two optimization schemes, this paper presents an optimization scheme aiming at the maximum boiler efficiency, and verifies the optimization effect. Based on the research results of modeling and optimization scheme, the combustion optimization control system is developed.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM621.2;TP273

【参考文献】