中文摘要：

为准确预测双循环流化床生物质气化的颗粒循环流率Gs,设计并搭建了双循环流化床冷态试验台,研究了提升管流化风速、二次风量、二次风送风方式、二次风口高度及二次风口数量对颗粒循环流率的影响,并建立了基于Levenberg-Marquardt优化算法的BP神经网络预测模型,通过对比找出了最优模型,对颗粒循环流率进行了预测.结果表明：Gs随着提升管流化风速和二次风量的增大而增加,二次风量超过一定值后,增加的趋势变缓;二次风径向引入比切向引入时的Gs大;Gs对二次风口高度的变化十分敏感;应用该BP神经网络模型得出的Gs预测值与试验值的平均偏差为0.07 kg/（m2.s）,平均相对误差仅为0.49%,模型准确地预测了提升管送风特性对颗粒循环流率的影响.

英文摘要：

To accurately predict the circulation rate Gs of particles from biomass gasification in a dual circulating fluidized bed（DCFB）,a cold-state test rig has been set up so as to study how following factors influence the Gs,such as the fluidized air velocity in riser tube,the secondary air flow rate and supply mode as well as the height and number of secondary air tuyeres,etc.BP neural network prediction models are established based on Levenberg-Marquardt（LM） algorithm,from which an optimal network model is derived for prediction of the Gs.Results show that Gs rises with increasing fluidized air velocity and secondary air flow rate,however,it tends to be stable when the secondary air flow rate gets to a certain value.The Gs obtained at the supply mode of secondary air in radial direction is higher than that in tangential direction.The Gs is very sensitive to the height of secondary air tuyere.The mean deviation of Gs between calculated data and actual measurements is 0.07 kg/（m2·s）,and the mean relative error is merely 0.49%,indicating a precise prediction on how the air supply characteristics in riser influence the particle circulation rate.