一种改进的调合辛烷值模型预测汽油研究法辛烷值

摘要

对中国石化安庆分公司（安庆石化）汽油调合组分油进行调合实验，建立了修正的调合辛烷值模型预测汽油的研究法辛烷值，利用基于 LM 改进的信赖域方法进行了模型非线型回归。选取安庆石化实际生产数据验证模型准确性，比较了线性模型、调合辛烷值模型、改进调合辛烷值模型的预测效果。通过对比3种模型的预测结果，发现3种模型中改进调合辛烷值模型具有较好的预测效果，平均绝对误差0．49，平均相对误差0．5％，残差均方和是0．74，模型的 R 2是1．08。%Based on the blending experiments of gasoline blending components(from Anqing branch) and the model parameters calculated by non-linear regression LM trust region method,an improved RON prediction model was established. The accuracies for RON prediction was tested and compared for linear model,blending model and improved blending model using real data of the production. The re-sults of the improved model are:mean absolute error of 0.49,average relative error of 0.5%,MSE of 0.74,and R 2 of 1.08,indicating the better prediction effect of the improved blending model than the other two models.