电站锅炉低NO_x燃烧建模优化研究与应用

Optimization Study and Application of the Modeling of the Low NO_x Combustion of a Utility Boiler

为了降低电站锅炉NOx排放,对锅炉燃烧进行优化和控制,分别应用NOx排放特性的实炉试验数据和在线采集的大量实时运行数据进行支持向量机算法和BP神经网络算法的建模研究,对比了不同建模数据情况下两种算法建模的性能,结果表明,支持向量机模型在泛化能力方面优于BP神经网络模型。应用支持向量机算法与遗传算法相结合,针对某330 MW煤粉炉进行了实际的锅炉燃烧优化试验,将该炉的NOx排放由708.5 mg/m3降低为576.1 mg/m3。表明该方法是一种有效的降低NOx排放的锅炉燃烧优化方法。

To reduce the NOx emissions of utility boilers,optimized and controlled was the combustion of a boiler to utilize the test data and a large number of on-line acquired operation data of the NOx emission characteristics of an actual boiler respectively to study how to establish a model based on the supporting vector machine algorithm and BP neural network algorithm. On this basis,the performance of the modeling based on both algorithms under different data for modeling was compared. It has been found that the supporting vector machine model is superior to BP neural network model in terms of the generalization ability. The supporting vector machine algorithm was used in combination with the genetic algorithm to conduct an actual combustion optimization test of a 330 MW pulverized coal-fired boiler,lowering the NOx emissions of the boiler from 708. 5 mg/m 3 to 576. 1 mg/m 3 . This shows that the method in question is an effective combustion optimization method for boilers to reduce NOx emissions.