摘要
设计并搭建双循环流化床冷态试验台,通过试验分析提升管二次风风量、二次风送风方式、二次风口高度和二次风口数目对颗粒循环流率的影响。建立BP神经网络预测模型,采用3种算法对颗粒循环流率进行预测,通过对比找出最优预测模型——基于LM算法的改进型BP神经网络预测模型。该预测模型很好地预测了二次风特性对颗粒循环流率的影响,试验值与模型预测值的平均绝对误差为0.23kg/(m2·s),平均相对误差仅为1.37%;最大偏差为1.23kg/(m2·s),最大相对误差5.75%。
The effect of flow rate, supply mode, inlet height and inlet number of secondary air on solids circulation rate were experimentally analyzed on the self-designed DFCB cold-state test bench. A BP neural network prediction model was established, and improved BP neural network prediction model based on Levenberg- Marquardt algorithm was determined to be the optimal model to predict the solid circulation rate via comparison of three algorithms. This model made a good prediction on the relationship between characteristics of secondary air and solids circulation rate. It shows that the mean absolute error between forecasted value and experimental value is 0.23 kg/（m2 -s）, the mean relative error is merely 1.37%, the maximum diversion is 1.23kg/（m2 ＂s）, and the maximum relative error is no more than 5.75%.
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2013年第6期1109-1114,共6页
Acta Energiae Solaris Sinica
基金
国家自然科学基金(50876030)
高等学校博士学科点专项科研基金(20090036110008)
关键词
双循环流化床
二次风
颗粒循环流率
改进型BP神经网络
预测模型
dual circulating fluidized bed
secondary air
solids circulation rate
improved BP neural network
prediction model
