基于机器学习的一次风量软测量技术研究

Study on Soft Sensing Technology for Primary Air Flow Based on Machine Learning

磨煤机风量的准确、可靠测量是保证火电机组控制系统稳定运行,进而提高锅炉燃烧效率的重要因素,然而受现场安装条件等方面的限制,仪表测量得到的磨煤机风量与实际值偏差很大。针对这一问题,应用软测量方法,结合火电厂的实际应用,对磨煤机风量软测量中辅助变量的选择、数据预处理、测量模型的建立及校正等问题进行了研究。采用基于支持向量机回归的方法建立了风量软测量模型,并对建模过程中核函数、惩罚因子的选择进行了分析和研究。电厂实际运行数据的验证表明:该软测量方法能够获得比现有硬件流量仪表更准确可靠的测量结果,且能适应机组工况的变化。

The accurate and reliable measurement of the coal mill air flow is an important factor for the stable operation of the thermal power unit control system and the improvement of the boiler combustion efficiency. However, limited by the site installation conditions, the values of coal mill air flow measured by the meters differ greatly from the actual values. Combined with the practical application in the thermal power plant, this paper adopts the soft sensing method to solve the problem. It studies the secondary variable selection, data preprocessing and measurement model establishment and correction etc. in the soft sensing of the coal mill. By applying the support vector machine （SVM） regression method, the air flow soft sensing model is estab- lished. The selection of kernel function and penalty factor during the establishment process is also analyzed and studied. The actual operation data in the plant verifies that the soft sensing method can obtain more accurate measurement results than the existing hardware flowmeters and can adapt to the changes in the unit conditions.