摘要
目前对燃煤电站烟气中汞形态浓度的预测模型尚不完善。将BP神经网络和GA遗传算法相结合组成GA-BP神经网络算法,用于燃煤烟气中汞形态浓度分布的预测。使用遗传算法对BP网络的初始权值进行优化,可以在解空间中定位出较好的搜索空间,然后采用BP算法在这个小的解空间中搜索出最优解。对75组燃煤电厂烟气中的汞浓度实测数据进行神经网络算法的训练和预测,结果表明GA-BP神经网络模型不仅可以预测燃煤烟气中汞形态浓度的分布,而且具有较高的预测精度。
There has not been an accurate model to predict the mercury speciation and its concentration in coal-fired flue gas. A GA-BP model comprised of the genetic algorithm （GA） and BP neural network was used to forecast the distribution of mercury speciation based on many available data extracted from literatures published on coal-fired flue gases. Through optimizing the original weights and biases of neural network by use of the genetic algorithms, a better searching space in solution space could be obtained, then the optimal solution could be achieved using BP neural network. It verified that the GA-BP neural networks that have been trained with a large number of data could predict the distribution of mercury speciation in coal-fired flue gas fast and accurately.
出处
《电站系统工程》
北大核心
2007年第6期15-18,共4页
Power System Engineering
基金
国家重点基础研究发展计划(973计划)资助项目(2002CB211604
2006CB200301)
关键词
燃煤烟气
汞形态分布
BP神经网络
遗传算法
coal-fired flue gas
distribution of mercury speciation
BP neural network
genetic algorithm
