Research on Time Series Prediction Model for the Trend of Corrosion Rate

摘要

In order to realize the prediction and early warning of corrosion status and reduce the risk of corrosion, the research on the prediction of corrosion rate trend for on-line monitoring of oil refining units is carried out. In this paper, the time series corrosion rate data of on-line monitoring probe is used to study the prediction model based on Autoregressive Integrated Moving Average (ARIMA). Firstly, the long-term monitoring data of corrosion rate is preprocessed and the data stability is judged. Then, the Akaike Information Criterion and Bayesian Information Criterion are used to select the parameters of ARIMA model and judge the applicability of the model. Finally, ARIMA(2,1,1) and ARIMA(1,1,1) parameters were used to realize the rapid prediction of corrosion rate trend, with the minimum average error of 10.08%; meanwhile, the accuracy of corrosion rate prediction was effectively improved by changing the modeling interval.