基于七个运行参数建立煤粉锅炉NO_x排放模型（英文）

Development of a NO_x emission model with seven optimized input parameters for a coal-fired boiler

目的:采用最小二乘支持向量机建立煤粉锅炉NO_x排放模型,即建立输入参数与NO_x之间的关系。合理选择输入参数不仅会降低模型的复杂度,而且会提高模型的精度。为此,本文探讨各输入参数对模型的影响,并最终保留合适数量的输入参数建立NO_x排放模型。创新点:1.采用最小二乘支持向量机建立NO_x排放模型;2.通过敏感性分析确定模型的最终输入参数。方法:1.根据专家知识及运行经验确定NO_x排放模型的初始输入参数(图2);2.根据锅炉的运行历史数据,采用最小二乘支持向量机建立NO_x排放模型;3.采用敏感性分析方法确定NO_x排放模型的最终输入参数(图11),并用其进行建模以验证模型的有效性。结论:1.采用最小二乘支持向量机建立的1000 MW超超临界前后墙对冲锅炉NO_x排放模型,可靠性和精度较高;2.经过敏感性分析,NO_x排放模型的输入参数由初始的33个降为7个,模型的复杂度降低且精度提高。

Optimizing the operation of coal-fired power plants to reduce nitrogen oxide（NOx） emissions requires accurate modeling of the NOx emission process. The careful selection of input parameters not only forms the basis of accurate modeling, but can also be used to reduce the complexity of the model. The present study employs the least squares support vector machine-supervised learning method to model NOx emissions based on historical real time data obtained from a 1000-MW once-through boiler. The initial input parameters are determined by expert knowledge and operational experience, while the final input parameters are obtained by sensitivity analysis, where the variation in model accuracy for a given set of data is analyzed as one or several input parameters are successively omitted from the calculations, while retaining all other parameters. Here, model accuracy is evaluated according to the mean relative error（MRE）. This process reduces the parameters required for NOx emission modeling from an initial number of 33 to 7, while the corresponding MRE is reduced from 3.09% to 2.23%. Moreover, a correlation of 0.9566 between predicted and measured values was obtained by applying the model with just these seven input parameters to a validation dataset. As such, the proposed method for selecting input parameters serves as a reference for related studies.