论文部分内容阅读
原油炼制中减压塔侧线产品是精制润滑油的原料,其粘度测量对生产有重要意义。通过统计分析得出了影响侧线产品粘度的主要因素,并采用支持向量机回归方法建立粘度软测量模型。针对支持向量机训练参数确定问题,提出了采用差分进化算法的搜索策略,使模型训练参数的调整过程按预定目标自动快速优化。所构建的粘度软测量模型预报精度较高,趋势跟踪性能良好。
Crude oil refinery in the vacuum tower sideline product is refined lubricating oil raw materials, the viscosity measurement of great significance to the production. The main factors affecting the viscosity of the sideline product are obtained through statistical analysis. And the soft sensor model of viscosity is established by the support vector machine regression method. In order to solve the problem of determining training parameters of SVM, a search strategy based on differential evolution algorithm is proposed, which makes the adjustment process of the model training parameters be automatically and quickly optimized according to a predetermined target. The constructed viscosity soft-sensing model has high forecasting accuracy and good trend tracking performance.