Study On Soft Measurement And Operation Optimization Of Atmospheric And Vacuum Distillation Unit

Atmospheric and vacuum distillation unit as the oil refining enterprises leading, realizing device product quality optimization operation is the premise of measurement feedback information to obtain the main products of plant quality. But a lot of important parameter of measuring the quality of the products can’t be directly measured, and the chromatograph online measuring product quality is very expensive, with measurement lag and inconvenient maintenance. Based on the analysis of the atmospheric distillation process, atmospheric and vacuum flash point、viscosity soft measurement and yield of the atmospheric and vacuum tower three lines are studied, to fully tap the potential of device optimization.Based on the analysis of the processing mechanism and operating characteristics of the atmospheric and vacuum distillation unit, related variables of flash point、viscosity、lateral line quality are studied, which are secondary variables of the soft sensor model. Having gotten the production data of atmospheric and vacuum distillation unit in a petrochemical factory, we do the data processing, such as errors processing、transform data and abnormal data preprocessing, to get the effective information of soft measurement modeling, and consider the measurement lag time, to effectively improve the accuracy of modeling. Select the RBF neural network technology as a solution method of soft sensor modeling of product quality, because of its excellent performance to approach any nonlinear function, learning the network by orthogonal least square method, correction the process model online by the recursive algorithm with forgetting factor. After training the network we can get the nonlinear system model with higher precision. The technique is applied to soft measurement modeling of atmospheric distillation three-line^vacuum tower second-line and three-line the oil viscosity and flash point, and we have achieved good results. After having recommended traditional complex optimization method, we did research a new method about swarm intelligence optimization—glowworm swarm optimization algorithm model (GSO), by updating the fluorescein value, searching the brightest individuals, updating individual position and sensing range to achieve the optimization process. For the steady-state optimization problem of atmospheric column light oil yield, we used both complex method and GSO method, and got the operating parameters of optimizing device, with product yield improvement, to increase the economic benefit of the plant.