一种改进的调合辛烷值模型预测汽油研究法辛烷值

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。

RON PREDICTION BY IMPROVED MODEL FOR BLENDED GASOLINE

Motor gasoline blending, a critical step in refinery, is a processing of enormous benefit. Due to octane number is the important quality index, it is so significant for refinery to possess a precise octane number model. By the blending experiment, we established a developed blending octane number model and obtained the model parameters through developed LM. Furthermore we made a comparison between linear model、blending octane number model and developed blending octane number model. These predicting date derived from three models indicate that the developed blending octane model is advantageous. In order to verify the correctness of the model, this article randomly selected 31 An Qing refinery production data to verify the accuracy of the actual model, the result is that mean absolute error is 0.49, the average relative error is 0.5%, MSE is 0.74,R2 is 1.08.