预测循环流化床L阀颗粒质量流量的改进BP算法  

Using the Improved BP Algorithm of Neural Networks to Predict the Solid Mass Flow of L-Valve in Circulating Fluidized Bed

作　　者：张伟[1] 王恩禄[1] 马天星[1] 马健[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《上海交通大学学报》2006年第8期1316-1319,共4页Journal of Shanghai Jiaotong University

摘　　要：在采用相对误差函数、训练集重组、增加惯性项、参数动态变化、权矩阵随机变化量等措施对传统的BP神经网络算法改进的基础上,利用改进的BP神经网络算法建立了循环流化床L阀固体颗粒质量流量的预测模型,得出了循环流化床L阀的充气位置、充气方式和充气体积流量与固体颗粒质量流量之间的映射关系,获得了对循环流化床L阀固体颗粒质量流量较好的预测结果.In the basis of improving the traditional BP （back propagation） algorithm of neural networks by using relative error function, regrouping the training sets, adding inertial part, letting parameter change dynamically, and adding random value to the connection matrix, a model of the solid mass flow of L-valve in a circulating fluidized bed with different air charging manner, position, and volume flow was built up by using the improved BP algorithm. The prediction result of the solid mass flow of L-valve through using the improved BP algorithm has a good agreement with the experimental results acquired from the L-valve test facility.

关 键 词：循环流化床 L阀 BP算法 神经网络 

分 类 号：TK221[动力工程及工程热物理—动力机械及工程]