基于神经网络与遗传算法的常压塔顶油气系统腐蚀预测

Corrosion Prediction for Oil and Gas System of Atmospheric Tower Overhead by Neural Network and Genetic Algorithm

通过腐蚀检测系统对某公司常压蒸馏装置低温部位腐蚀情况进行了检测。将一段时间内检测到的塔顶污水的pH值、Cl-~浓度、Fe^(2+)浓度和硫化物含量作为输入数据,腐蚀速率作为输出数据,通过BP神经网络建立了腐蚀速率预测模型。预测数据与实际数据对比结果表明,预测误差较大。运用遗传算法对BP神经网络进行优化,优化后的模型能够较准确预测常压蒸馏装置低温部位腐蚀情况。

Corrosion situation for the low-temperature part of atmospheric distillation unit was tested by corrosion detection system,and then p H value,Cl-concentration,Fe2＋ concentration and sulfide content were used as input data,while corrosion rate was taken as output data. Predictive model for corrosion rate was established through BP neural network based on the detected data from overhead sewage,although error of the prediction results was large. BP neural network was optimized by genetic algorithm which could accurately predict corrosion situation in low-temperature part of the unit.