常压塔塔顶腐蚀关键参量相关性分析与预测

Correlation Analysis and Prediction of Key Parameters of Corrosion in an Atmospheric Tower Overhead System

预测常压塔顶回流罐切水的关键离子浓度可以为常压塔顶系统工艺防腐提供技术指导。收集了某炼厂2014-2016年常压装置的生产实时工艺参数、原料及产品采样数据、水质分析化验参数,采用线性插值方法对不同时间尺度的数据进行补全。通过线性相关性分析获得了影响常压塔顶回流罐切水总铁离子浓度、pH的主要因素,并基于深度学习、支持向量机回归、粒子群优化方法建立了腐蚀关键参量预测模型。结果表明,pH值和总铁离子浓度的相关因素大部分重叠,与原料性质及产品馏出温度较强相关;建立的回归模型预测精度高,在训练集和预测集上,总铁离子浓度预测值与测量值最大偏差分别为4. 4%和9. 8%,pH值预测值与测量值最大偏差分别为0. 9%和1. 1%。

Prediction of the critical ion concentration of the cut water in the reflux tank at the top of the atmospheric tower can provide the technical guidance for the process anticorrosion of the atmospheric tower overhead system.The real-time production process parameters,raw material and product sampling data,and water quality analysis and test parameters of a refinery atmospheric unit from 2014 to 2016 were collected. Linear interpolation method was applied to complete the data of different time scales. Through linear correlation analysis,the main factors affecting the total iron ion concentration and pH of the cut water were obtained and the prediction model of key corrosion parameters was established based on deep learning,support vector machine regression and particle swam optimization. The results showed that the relevant factors of the pH value and total iron ion concentration mostly overlap and were strongly related to the properties of raw materials and the distillation temperature of products. The regression model had high prediction accuracy. For the training set and prediction set,the maximum deviations of total iron ion concentration prediction value and measured value were 4. 4% and 9.8% respectively. And the maximum deviations of p H prediction value and measured value were 0. 9% and 1. 1%.