主管：中华人民共和国应急管理部
主办：应急管理部天津消防研究所
ISSN 1009-0029  CN 12-1311/TU
消防科学与技术

消防科学与技术 ›› 2021, Vol. 40 ›› Issue (10): 1479-1483.

• • 上一篇    下一篇

基于KPCA-FA-ELM 模型的管道剩余寿命预测

张 刚   

  1. 长庆油田分公司 页岩油产能建设项目组，甘肃 庆阳 745000
  • 出版日期:2021-10-15 发布日期:2021-10-15
  • 作者简介:张 刚（1986-），男，陕西西安人，长庆油田分公司页岩油产能建设项目组油气田开发工程师，学士，主要从事油气田开发、油田集输、油田地面工程建设等工作，甘肃省庆阳市西峰区陇东指挥中心，745000。

Prediction of remaining life of pipeline based on KPCA-FA-ELM model

ZHANG Gang   

  1. Shale Oil Production Capacity Construction Project Team, Changqing Oilfield Branch, Gansu Qingyang 745000, China)
  • Online:2021-10-15 Published:2021-10-15

摘要： 针对管道剩余寿命的预测问题，提出了一种基于KPCA-FA-ELM组合模型的预测方法，对腐蚀管道剩余寿命预测模型的构建方法以及预测模型的性能验证方法进行研究。以我国某油田的回注水管道和油气集输管道为例，对本次研究所提出的管道剩余寿命预测方法进行验证。研究表明：本次研究所提出的KPCA-FA-ELM模型在进行油田注水管道剩余寿命预测过程中，其最小相对误差为0.38%，最大相对误差为6.1%，平均相对误差为2.35%，均方根误差为0.207，希尔不等系数为0.011，在进行油气集输管道剩余寿命预测过程中，其评价指标均小于其他模型，因此，该种模型的性能优于其他常见预测模型。

关键词: 管道剩余寿命, 腐蚀管道, 核主成分分析, 萤火虫算法, 极限学习机

Abstract: Aiming at the prediction of the remaining life of pipelines, this study proposes a prediction method based on the KPCA-FA-ELM combined model. The construction method of the remaining life prediction model of the corroded pipeline and the performance verification method of the prediction model are studied. Take oilfield reinjection pipelines and oil and gas gathering pipelines as examples to verify the pipeline remaining life prediction method proposed in this research. The research shows that the minimum relative error of the KPCA-FA-ELM model proposed by this research is 0.38%, the maximum relative error is 6.1%, and the average relative error is 2.35% in the process of predicting the remaining life of oilfield water injection pipelines. The root square error is 0.207, and the Hill inequality coefficient is 0.011. In the process of predicting the remaining life of oil and gas gathering and transportation pipelines, the evaluation indicators are smaller than other models. Therefore, the performance of this model is better than other common prediction models.

Key words: remaining life of pipeline, corroded pipeline, nuclear principal component analysis, firefly algorithm, extreme learning machine