基于PCA-SVR 模型的油气管道剩余强度预测

1. 西安石油大学石油工程学院;2. 中国石油管道局工程有限公司

PCA-SVR 模型;单一缺陷;油气管道;剩余强度;预测

Prediction on the residual strength of oil and gas pipelines based onPCA-SVR model
MA Gang1, LI Junfei2, BAI Rui1, DAI Zheng1

1. College of Petroleum Engineering, Xi'an Shiyou University;2. China Petroleum Pipeline Engineering Co. Ltd.

PCA-SVR model, single defect, oil and gas pipelines, residual strength, prediction

备注

针对单一缺陷油气管道的剩余强度预测问题,深入分析影响管道剩余强度的相关因素,对主成分分析(PCA)、支持向量回归法(SVR)及PCA+SVR 的基本原理和组合过程进行总结,利用从文献中获取的单一缺陷管道相关数据,使用PCA 对影响因素进行降维处理,最后使用SVR 模型进行剩余强度预测,并将预测结果与其他常见模型和计算方法进行对比,以此验证该模型的可行性。研究结果表明:在所有影响因素中,管道钢级对油气管道剩余强度的影响最大;PCA+SVR 预测模型的预测平均误差为1.91%,均方根误差为0.34,证明此方法具有较高的准确率,但所有预测结果均小于实际剩余强度,证明该方法的保守性相对较强,会导致管道工作效率下降。(图3表2,参[21]

In order to predict the residual strength of single-defect oil and gas pipelines correctly, the relevant factors affecting the residual strength of pipelines were analyzed thoroughly. The basic principles and combination processes of principal component analysis (PCA), support vector regression(SVR)and PCA+SVR were introduced. Then, according to the data related to the single-defect pipelines acquired from the literature, the dimension of influential factors was reduced by PCA. Finally, the residual strength was predicted in SVR model, and the prediction results were compared with the results of other common models and calculation methods to verify the feasibility of PCA-SVR model. It is indicated that among all influential factors, the pipeline steel grade has the greatest influence on the residual strength of oil and gas pipelines. In addition, the average prediction error of PCA+SVR prediction model is 1.91% and the root mean square error is 0.34, indicating higher accuracy of this method. However, all of the prediction results are lower than the actual residual strength, which proves that this method is relatively conservative and can decrease the operating efficiency of pipelines. (3 Figures, 2 Tabels, 21 References)