Evaluation of Crude Oil Production and Transportation Risk Indicator Based on Outlier Detection and Gaussian Process Regression

Many fire or explosion accidents have happened because of the high temperature which is the excellent indicator for risk evaluation caused by an oxidation exothermic reaction of sulfurized rust in the production and transportation of sulfur-containing oil. To efficiently evaluate the temperature and prevent such accidents, this paper proposes a method based on outlier detection technique and Gaussian process regression (GPR) method to predict the maximum temperature during the oxidation self-heating process of sulfurized rust. Firstly, the Box plot method and k-means algorithm are adopted to detect and eliminate the outliers in the raw data for more fitting of model. Then, the GPR model trained and validated is applied to predict the maximum temperature based on the remaining data. Finally, the prediction results obtained by the proposed method in this paper are compared with those by the Support Vector Machine (SVM) algorithm and the traditional GPR algorithm in which the outlier detection is not conducted in advance, the result shows the proposed method is more accurate and rational. It indicates that the method is more of significance to assess risk and prevent disaster during the production and transportation of the crude oil with high sulfur.