1.Key Lab of Ind Computer Ctrl Eng of Hebei Province, Yanshan University, Qinhuangdao, Hebei 066004, China
2. Hebei Iron & Steel Group, Energy Ctrl Center of Chengde Iron & Steel Company, Chengde, Hebei 067102, China
Abstract：In order to reduce NOx emissions from utility boilers, a new machine learning method——relevance vector machine is presented. This is to build the model of a 330MW pulverized coal boiler for NOx output and twenty-six inputs such as drum first and secondary air, oxygen and so on, then gravitational search algorithm is used to optimize the parameters of the model to obtain the optimal pattern.Through comparing the outcome of particle swarm optimization‘s and genetic algorithm‘s optimizing relevance vector machine and gravitational search algorithm's optimizing support vector machine. Finally, the boiler adjustable variable input parameter is selected as the optimization variables for the target of cutting down NOx emissions to achieve the appropriate input parameters of lower NOx emissions. The result shows that gravitational search algorithm’s optimizing relevance vector machine gets better accuracy than the others, after the model of low NOx optimization, the results from the initial NOx output value of 906.65mg/m3 becomes 550.600mg/m3, a decrease of approximately 38.9%, to achieve a significant reduction in NOx emissions.
［1］王春林,周昊,李国能,等. 大型电厂锅炉NOx排放特性的支持向量机模型［J］.浙江大学学报（工学版）,2006,40(10):1787-1791.
［2］牛培峰,肖兴军,李国强,等.基于万有引力搜索算法的NOx排放模型的参数优化［J］.动力工程学报,2013, 33(2): 101-106.
［3］Tipping M E. Sparse Bayesian learning and the relevance vector machine［J］. Journal of Machine Learning Research,2001, 1:211-244.
2016, Vol. 37 
