燃煤机组烟气流量软测量技术研究

发布时间：2018-05-07 16:26

本文选题：烟气流量 + 软测量　；参考：《华北电力大学》2017年硕士论文

【摘要】：目前,节能、环保是各发电企业必须要面临的问题。烟气流量是节能环保控制优化过程中,非常重要的一个变量。但是烟气流量的测量却面临各种各样的问题,一方面,机组容量的扩大使得烟道截面较大,出现截面流场分布不均,传统硬件传感器测量出现较大误差,另一方面,受到测量传感器的影响,如目前传感器在高温、高尘、高腐蚀的影响,硬件传感器会常出现故障,运行维修困难等等,这些因素严重制约着发电企业节能环保的自动化运行程度,严重制约着发电企业的工作效率。本文以燃煤机组烟气流量软测量技术为研究对象,首先,分析影响烟气流量的各种因素,如影响烟气量产生与气体流量等因素,对这些因素进行理论分析与数据MATLAB分析,通过PLS变量投影重要性分析算法与前向搜索算法综合进行变量筛选,解决辅助变量之间存在相关性和辅助变量对主导变量影响重要性的问题。其次,在建模方面,根据机组运行情况选择典型工况数据作为静态建模的过程数据;在数据预处理方面,采用拉依达准则和归一化方法对奇异点、孤立点进行处理,消除对建模过程数据的影响;在建模方法上,静态建模采用改进最小二乘支持向量机算法建模,针对LSSVM丧失稀疏性问题,本文采用相似度函数法对建模数据进行冗余化处理,建模过程中采用相似度函数法和剪枝算法来解决稀疏性问题以增加模型的泛化能力。动态建模方面,由于静态建模过程中,无法选择全部的工况进行建模,在线修正是软测量技术研究必不可少的一步,本文采用自适应留一交叉预报误差的滑窗递推算法进行修正模型参数增加模型在线预测能力。本文在以上理论方法的研究基础上,采集数据,采用MATLAB编写程序,完成建模数据处理和辅助变量选择,进行烟气流量的静态和动态在线建模。仿真实验结果表明所建模型对各工况下的预测结果能够达到预想效果,为进一步完成燃煤机组节能环保优化提供依据。
[Abstract]:At present, energy conservation, environmental protection is the power generation enterprises must face the problem. Flue gas flow is a very important variable in the process of energy saving and environmental protection control optimization. However, the measurement of flue gas flow is faced with various problems. On the one hand, the expansion of unit capacity makes the flue section larger, the cross-section flow field uneven, the traditional hardware sensor measurement error, on the other hand, Affected by the measurement sensors, such as the high temperature, high dust, high corrosion of the sensor, the hardware sensor will often malfunction, operation and maintenance difficulties, etc. These factors seriously restrict the automation operation degree of energy saving and environmental protection of power generation enterprises, and seriously restrict the working efficiency of power generation enterprises. In this paper, the soft measurement technology of flue gas flow in coal-fired units is taken as the research object. Firstly, the factors influencing the flue gas flow, such as the generation of flue gas and the gas flow, are analyzed theoretically and MATLAB. Through the combination of PLS variable projection importance analysis algorithm and forward search algorithm, the problem of correlation between auxiliary variables and the influence of auxiliary variables on dominant variables is solved. Secondly, in the aspect of modeling, according to the operation condition of the unit, the data of typical working condition is selected as the process data of static modeling, and in the aspect of data preprocessing, the singularity and the isolated point are treated by using the Lagrangian criterion and the normalization method. In the modeling method, the improved least squares support vector machine (LS-SVM) algorithm is used to model the static modeling, and the similarity function method is used to deal with the redundancy of the modeling data, aiming at the problem of LSSVM losing sparsity. The similarity function method and pruning algorithm are used to solve the sparse problem in order to increase the generalization ability of the model. In dynamic modeling, due to the static modeling process, it is impossible to choose all the working conditions to model, so online correction is an indispensable step in the research of soft sensing technology. In this paper, a sliding window recursive algorithm with adaptive residual cross prediction error is used to modify the model parameters to increase the on-line prediction ability of the model. On the basis of the research of the above theories and methods, this paper collects the data, writes the program with MATLAB, completes the modeling data processing and the auxiliary variable selection, and carries on the static and dynamic on-line modeling of the flue gas flow. The simulation results show that the predicted results of the model can achieve the desired results, which provides the basis for further optimization of energy saving and environmental protection of coal-fired units.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X773;X831

【参考文献】