基于DE_BPANN的单料烟烟气指标预测方法

为了分析和挖掘单料烟的化学成分和烟气指标之间的关系,采用基于差分进化的人工神经网络用于预测单料烟的烟气指标。建立了单隐层的人工神经网络,并针对基于误差反向传播的人工神经网络的缺陷,将差分进化算法应用于神经网络的训练过程。该预测方法的主要思想是结合了人工神经网络的局部搜索能力和差分进化的全局搜索能力。通过采用某烟草公司的单料烟烟叶、烟气数据,将单料烟的7种常规化学成分作为预测模型的输入变量,将主流烟气中的焦油、烟气烟碱和CO作为预测模型的输出变量,建立了人工神经网络预测模型。实验结果表明:焦油、烟气烟碱和CO的预测均方差达到了较好水平,与传统神经网络相比,分别提高了27%,10%和26%,表明该方法的预测准确度更高。

FGD Capacity Prediction of Thermal Power Plants in China

Through analyzing the proportion of SO2 emission from thermal power plants in the nationwide SO2 emis- sion in USA, Japan etc. developed countries, and the developmental course of thermal power installed capacity and the FGD capacity in USA, the FGD capacity of thermal power plants in China is forecasted from two angles. One is to predict FGD capacity in accordance with the policy in force in China. The other is to predict FGD capacity based upon the emission right trading policy. As compared, it is held that FGD equipment should be mainly installed on the large size units burning high sulfur coal according to the emission right trading policy. Such a method of work not only can economize large amount of investments and operation costs, but also can realize the same environmental effect.