Study On The Soft-Sensing Models Of Pour Point Of Light Diesel Oil In The Atmosphere Distillation Column

In this thesis, soft-sensing models of pour point of light diesel oil which are produced by the second column of atmosphere distillation unit in refinery are studied and established. The soft-sensing model is regulated through Distributed Control Systems (DCS) and the test result is effective.This job are made up of next main stages: analyze the mechanism of the atmosphere distillation column; collecting data forms sample set; analyze and dispose data with data statistic theory; establish the fitting model; operate and correct the model on line. The modeling way is based on mechanism and data. In the selection of secondary variables, the method combining Principal Component Analysis (PCA) with mechanism is adopted, and it is proved to be effective in the practice. On the analying of that there is strong linear relationship between pour point of light diesel oil and secondary variable, the soft-sensing model is established by applying Multivariable Linear Regression (MLR), Partial-Least-Squares Regression (PLSR), BP(Back Propagation)networks and RBF(Radial Basis Function)networks.At last, running environment of these soft-sensing models is ADVANT-500 of ABB Company. The models are implementing at control module with programming language, and are verified on line with constant variety. Data ratio with error less than 4℃gets to 95%, which satisfies the requirement of production.