Research And Application Of Soft-sensing Technology For Oil-water Interface

In order to ensure product quality and production efficiency, we need to strictly controlthe process variables in industrial production. These variables are closely related with productquality. However, there are kinds of variables can’t be detected by sensors. We can get themonly through the method of offline analysis. Sometimes, these parameters must be detected inreal time, for the reason that they affect the production process and product quality directly orindirectly. As a new solution of these problems, soft-sensing is respected by people gradually.This paper researches the general situation of the soft-sensing technique which containsthe method of modeling the soft-sensing model, the selection of auxiliary variable and theway of data treatment, etc. According to various industry processes, there are many differentmodeling methods, all of them have their advantages and disadvantages. In this paper, wehave a brief description about the crude oil dehydration process. Using the soft-senor insteadof hardware to measure the oil-water interface highness can be meaningful.By studying a number of soft-sensor modeling methods, we selected a modeling methodwhich based on BP neural network. This kind of method has approximation ability, especiallyfor the uncertainty and nonlinear object. For the problem in the dehydration of crude oilproduction process, this article uses the neural network to establish the oil-water interfacehighness model. We simulated the experiment, verify the correctness of the model.According to the characteristics of the BP network and the study of improved algorithm,we use evolutionary algorithm to optimize neural network. Here chooses particle swarmoptimization algorithm to establish the soft sensor separately, then we simulate the experiment.Two algorithms are found can be combined into a hybrid algorithm. We use genetic algorithmto optimize the particle swarm optimization algorithm. Compared with the separate algorithm,this kind of algorithm can get better effect and fit in with the requirement of industry.