回转干馏炉内油页岩停留时间模型建立及其预测

Mean residence time experiment and prediction of oil shale in rotary retorting

为了提高估算MRT精度,文中采用支持向量机算法对求解MRT问题进行了建模,并在一定约束条件下,利用Gridregression.py寻找回归最优参数方法对支持向量机模型的参数进行了优化,获得了最优的模型参数。支持向量机模型将操作参数和结构参数作为输入量,MRT作为输出量,用实验数据对模型进行了校验和参数的寻优,利用优化后的模型对MRT进行了预测,并将预测结果与实验结果进行了对比,结果表明,优化后的支持向量机模型实现了对MRT较精确的预测。通过实验值和运动模型构建了平均停留时间的经验公式,结果表明:该平均停留时间经验公式的拟合曲线相对均方差为0.05,拟合效果良好。

In order to improve the accuracy of predicting performance for the mean residence time,a support vector machine（SVM）model is employed,and parameters of the SVM model optimized by gridregression and best parameters were obtained.The compositions of the operation and structure parameters were employed as inputs,and the mean residence time was used as output of the SVM model.The model was verified with the experiment datum,result of prediction by the optimized SVM model was compared with the test datum,and the result show the SVM model has achieved good predicting performance for both the mean residence time.This article builds an empirical equation of mean residence time by experimental value and motion model,and the results indicate：the relative mean squared error of the empirical equation fitting curve is 0.05,with a good fitting effect.