基于SVR的燃煤机组NOX含量的软测量模型

Soft Sensor Model Based on SVR for the Measurementof NOX Concentration of Coal-fired Unit

摘要:

作为火电厂燃煤脱硝工作的基础,选择性催化还原(SCR)脱硝反应器入口氮氧化物(NO_X)含量的测量至关重要。针对难以准确现场实时测量NO_X含量的问题,提出了基于回归型支持向量机(SVR)的软测量模型。首先对SCR反应器生成NO_X的过程进行机理分析,并结合相关性分析、主成分分析等数据处理方法选取辅助变量,然后基于回归型支持向量机算法建立模型,最后运用BP神经网络对模型进行检验,解决了SCR反应器入口NO_X的含量的难以准确预测问题。为SCR反应器入口NO_X含量的实时、准确测量打下基础。

Abstract:

As a basis for denitration of coal-fired power plant, the measurement of nitrogen oxide (NOX) content at the inlet of the Selective Catalytic Reduction (SCR) reactor is critical. In order to solve the problem that the NOX content cannot be accurately measured in real time, a soft sensing model based on regression support vector machine (SVR) is proposed. Firstly, the process of generating NOX at the inlet of SCR reactor is analyzed. Then auxiliary variables are selected by correlation analysis and principal component analysis, and the mathematics model based on support vector machine for regression algorithm is built. Finally, the model is tested by the method of BP artificial neural network. The proposed model lays the foundation for the real-time and accurate measurement of the NOX concentration at the inlet of the SCR reactor.