制氢工艺水碳比神经网络模型研究

STUDY ON MODEL OF CARBON AND STEAM RATIO IN HYDROGEN PRODUCTION BASED ON ARTIFICIAL NEURAL NETWORK METHOD

摘要: 根据南京炼油厂制氢车间的生产数据，用人工神经网络(ANN)的反向传播(BP)算法对制氢装置转化生产中的水碳比进行预测。提出了适宜的人工神经网络拓扑结构，讨论了BP算法中学习速率、动量系数及过拟合现象对网络的影响，通过生产数据的检验表明，ANN方法能准确地关联和预报制氢装置转化生产中的水碳比，水碳比预测平均相对误差为2．83％。

Abstract: According to the hydrogen productiondate base, the Artificial Neural Network(ANN) was used for prediction carbon and steam ratio with Back- Propagation(BP) method. The appropriate topology of ANN was obtained. The learning rate, the momentum factor and ov