Texaco Gasifier Soft Sensor Modeling

Coal gasification as the main raw material in a comprehensive chemical enterprise is a very important aspect. Texaco coal gasification process of follow-up system for the synthesis gas as a raw material, which is a follow-up and stable operation of the system to ensure reliable operation. But the gasifier temperature measurement has always been a problem because the temperature measurement component must be able to meet the strong current, strong corrosion, high temperature. This device is not easy to choose, and reliability is not high. To find a reliable temperature measurement technology to overcome the temperature measurement devices were damaged, the entire process of adverse effects, it has become an important research.Soft sensor technology is an application of computer technology, using software instead of the sensor's role in achieving the purpose of online measurement of process parameters. In recent years, soft sensor technology has been used as a valuable alternative and an important method to instead of traditional method of measuring, in the chemical industrial processes, there are many variables are applied to soft sensor technology. In this paper, combined with a 600,000 t/year acetic acid for gas technology transformation projects, the soft sensor technology is used to measure temperature of the Texaco gasifier, for this project to find a suitable Gasifier Soft-sensing method.This paper mainly studies the support vector machine (SVM) algorithm, using BP neural network respectively, the standard support vector machine (SVM) and least square support vector machines (LS-SVM) to establish Gasifier soft sensor models. The results show that the model prediction errors of the first two methods fall within the control area are less than 85%, while least squares support vector machine prediction accuracy higher, model predictive control of the error falls range can be increased to a ratio of 97.0%. Visible, least squares support vector machine based on soft-sensor model can be modified to meet the technological requirements of the gasifier project.The papar also compares with support vector machines and neural networks. Finally, from the soft sensor modeling approach, support vector machine kernel function parameter selection and support vector machine algorithm optimization aspects, discuss the improvement of the subject in further research.