改进的BP算法在中小型高炉炉温预测上的应用  

Improved Back Propagation Network Algorithms on the Middle and Small Blast Furnace of Temperature's Prediction

作　　者：高峰[1] 马翠红[1] 刘惠欣[1] 

机构地区：[1]河北理工大学计算机与自动控制学院,河北唐山063009

出　　处：《甘肃冶金》2008年第3期17-19,共3页Gansu Metallurgy

摘　　要：本研究应用人工神经网络方法对高炉炉温进行预测,确定了三层前向网络结构。针对基于梯度下降法的BP网络存在的学习效率低、收敛速度慢、易陷于局部极小状态、网络的泛化及适应能力较差等缺陷,采用了附加动量法及自适应学习率、合理选择激励函数等策略对算法中的缺陷给出了改进措施,取得了较好的应用效果。In the course of the blast fumace＇s temperature predicting, the artificial neural networks method is studied and three-layer former network structure is confirmed. In allusion to the defect of grads dissension of traditional back propagation network algorithms, some improving measures such as adaptive learning and additive momentum has been taken, and better application result is acquired.

关 键 词：高炉 神经网络 炉温预测 铁水硅含量 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]