神经网络与遗传算法结合的球团竖炉燃烧优化

Optimization of combustion for pellet shaft furnace based on artificial neural network and genetic algorithm

对神经网络与遗传算法结合的球团竖炉燃烧优化方法进行了研究.首先构建了以矿料成分及含水率、相关操作参数以及燃烧室和炉膛温度等16个参数作为输入量,球团竖炉煤气吨耗和NOx污染物排放浓度作为输出量的人工神经网络模型.采用700组现场运行数据作为样本对神经网络进行训练,训练后的模型具有良好的泛化能力和预测精度,煤气吨耗预测误差低于3%且NOx排放浓度的相对误差在5%以内.此外,结合所建模型,采用实数编码的遗传算法,对球团竖炉燃烧进行优化计算,在寻优过程中对煤气吨耗及NOx排放这2个优化分量采用线性加权和的方法转化为单一数值的目标函数.通过选择不同的权重比例得出不同侧重条件下的优化目标函数,并给出该优化函数下寻优所得的操作参量优化控制方案.由所选优化方案数值解可以看出在煤气吨耗上升1.7%的情况下,NOx的排放浓度下降了20.37%.

Combined the neural network with genetic algorithms,a model for a shaft furnace which has tons of gas consumption and NO_x emission is built.There are sixteen input parameters in this model,containing mineral aggregate components,moisture content,furnace temperature and so on.Output parameters are the gas consumption and the concentration of NO_x emission.Based on the 700 groups of field data,the neural network has been trained.The results show that the prediction error of the gas consumption is less than 3% and the prediction error of NO_x emission is less than 5%.Base on this model,real-coded genetic algorithm is applied to linear weight low gas consumption and low NO_x emission and switch the model into a function with single variable parameter.Multiple objective functions and operating parameters focusing on different conditions can be discovered under different wight ratios.According to the optimization,the result shows that NO_x emission decreases by 20.37% while gas consumption increases by 1.7%.