中文摘要：

针对CMAC神经网络的网络节点随输入维数的增大呈几何级数增加的问题,基于模糊聚类提出一种改进的超闭球CMAC神经网络算法,用于电站锅炉混煤燃烧污染物析出软测量模型的建立。以电站锅炉实际运行工况的煤质特性数据和炉内燃烧条件为输入参数,通过软测量实现大型电站锅炉混煤燃烧硫、氮污染物生成浓度的精确预估和在线测量,用于指导电厂运行人员进行锅炉燃烧调整,以控制污染物的超标排放。与超闭球CMAC算法比较,提出的改进算法可以大大降低高维神经网络节点数并提高神经网络软测量精度,实验结果表明该方法的有效性和可行性。

英文摘要：

To overcome the drawback that CMAC neural network node number increases with the increasing of input dimensions exponentially,an improved hyperball CMAC neural network algorithm based on clustering is proposed to establish the soft measuring model for pollutant release from blended coal combusting in the boiler in power plant.The characteristic data of coal quality in practical boiler operating situation and the combusting condition in the furnace are taken as the input parameters.The soft measuring model was applied to achieve precise prediction and on-line measurement of the concentration of pollutants such as sulfur and nitrogen released from blended coal combusting in the boiler in power plant.It was also used to guide the operators in power plant to optimize coal combustion and control pollutant release.Compared with the hyperball CMAC algorithm,the improved algorithm can effectively reduce neural network nodes and improve learning accuracy.Experiment results demonstrate the feasibility and superiority of the novel algorithm.