常压蒸馏产品质量软测量改进方法及应用

针对石油化工生产过程基于统计模型的产品质量软测量中普遍存在的软测量模型适用工作范围小、难以反映进料原料性质变化的问题，探讨和应用了机理分析和统计建模相结合的软测量方法。针对一个实际常减压蒸馏产品质量软测量和控制问题．通过机理分析、实际操作数据分析并结合操作人员经验，选择能够反映进料原料性质变化的过程变量作为统计模型的输入以克服进料原料性质变化的影响，将有些直接测量的输入变量按照机理关系进行计算得到的新的变量作为统计建模的输入变量，使其和产品质量之间具有更宽范围的近似线性关系，提高软测量模型的泛化能力：某厂常减压蒸馏装置的实际应用结果表明了该方法的有效性。

Improved Soft-sensor Method and its Applications to Crude Unit Product Quality

Aiming at the issues that the soft-sensor models have only narrow working scopes and are difficult to reflect material property changes in the petroleum industry product quality the soft-sensor method is discussed based on physical analysis and statistics modeling. For a soft-sensor and control problem of practical crude unit product quality, through physical and operation data analysis with experience of operators, this method chooses process variables reflecting the crude oil property changes as inputs to reduce their influence,and also chooses some new variables computed from process variables measured directly according to physical models as inputs to get approximate linearity of broader scope with product quality and improve the generalization capability of soft-sensor model. The application results in a crude unit show that this method is valid.