摘要
以伊朗原油不大于350℃的宽馏分油的恩氏馏程为基础,建立了可预测轻质油闭口闪点的人工神经网络（ANN）模型,并利用现场采集的5种轻质油品的测定值检验了模型的预测能力。结果表明,对伊朗原油的宽馏分油,采用恩氏馏程数据进行学习训练,样本的记忆平均相对误差为0.65%,平均绝对误差为2.0℃。模型对轻质油品闪点的预测能力较好,预测的绝对误差为1.6~3.6℃,能满足工程计算要求。
An artificial neural network （ANN） model for predicting closed flash point of light oil was built based on Engler distillation range of wide distillate oil not more than 350 ℃ from Iran crude oil and the prediction ability of the model was examined hy the measured value of 5 light oils collected from different refineries. The results showed that Iran＇s awide distillate oil was trained with Engler distillation range data, the average relative error of sample＇s memory was O, 65% and average absolute error was 1.6 - 3, 6 ℃ , which could meet the requirement of enginevring calculation.
出处
《石化技术与应用》
CAS
2016年第1期24-28,共5页
Petrochemical Technology & Application
关键词
馏分油
闪点
人工神经网络模型
distillate oil
flash point
artificial network model
