Improved PLS Algorithm And Its Application In Delayed Coking

With the light crude oil consumed gradually, the petrochemical enterprises are forced to use the heavy crude oil to produce light oil. In order to make full use of the limited oil resources and meet the rapidly growing demand of light oil, heavy oil processing technology, especially the delayed coking technology, is developed quickly and used widely. Owing to its advantages that low investment and the ability of processing various heavy crude oil which increases the production of gasoline and diesel, how to improve the technology of delayed coking is increasingly important in the future. In this paper the study focuses on the model of delayed coking and the main work of the dissertation can be summarized as follows:(1)In this paper, because of its advantages in extracting useful information, removing noises and, Partial Least Squares method is used with the combination of hybrid PI-SIGMA fuzzy neural network algorithm, which has a strong ability to handle nonlinear problems. The new algorithm is applied in the model of coking speed and the coking height can be predicted according to the coking speed.The result shows that the precision of coking height is in the required range,(2)In order to predict95%distillation point of diesel, save labor cost and improve the efficiency of adjusting the device, this paper proposes a method to predict95%distillation point of diesel which based on improved Kernel Partial Least Squares by researching the process of delayed coking unit and analyzing the influencing factors about95%distillation point of diesel. The method deals with the non-linear relationship among variables effectively and quickly. In addition, the result shows that the model based on KPLS is better than that based on Partial Least Squares and so on.(3)On account of delayed coking production equipment, a delayed coking analysis and measurement software is developed. The software provides the following services:connecting with real-time database, calculating the value of95%distillation point of diesel online, forecasting coking height, displaying results online, alarming, as well as generating the electronic production reports. Good effects have been achieved since the software was put into operation.