双循环流化床颗粒循环流率实验研究

Experimental study on particle circulating rate of double-circulating fluidized bed

为实现颗粒循环流率的合理控制,在自行搭建的双循环流化床系统上对鼓泡床流化风速、快速床总风速和配风比、鼓泡床静床层高度、颗粒平均粒径等控制参数对颗粒循环流率的影响进行了研究,基于附加动量算法、Levenberg-Maraquardt算法和遗传算法3种不同的权值优化算法,建立了BP神经网络优化模型,并比较了模型预测值与实际值间的误差。研究结果表明:颗粒循环流率受鼓泡床流化风速变化影响较小;颗粒循环流率随快速床一次风比、总风速和鼓泡床静床层高度的增加而增加,随颗粒平均粒径增大而减小;基于遗传算法优化的BP神经网络在对测试样本测试时平均误差为0.436 5%,标准方差为0.064 1,预测值与实验值比较吻合,为较优的BP神经网络模型。

In order to realize reasonable control of particle circulating flow rate, the effects of control parameters on circulating flow rate were analyzed on a self-built double-circulating fluidized bed cold system, such as the wind speed of the bubbling bed, total wind speed and air distribution ratio of the fast bed, static bed height of the bubling bed, and the average particle size. Moreover, on the basis of three different weight optimization algorithms, like the additional momentum method, Levenberg-Marquardt（LM） algorithm and genetic algorithm（GA）, the BP neural network model was established, and the errors between the model predicted values and experimental values were compared. The results show that, the particle circulating flow rate was less influenced by the wind speed in bubbling bed, it increased with the primary air ratio and total wind speed of the fast bed and the static height of the bubling bed, but decreased with the increasing average particle size. The average error between the test sample and the one predicted by the GA-optimized BP neural network was 0.436 5%, the standard deviation was 0.064 1. The predicted values agreed well with the experimental values, indicating the BP neural network model was suitable.