基于先验分布的成品油管道停输数据有效性校验方法OA
A prior distribution-based validation method for shutdown data validity in refined oil pipelines
成品油长输管道在停输工况下的压力监测是识别泄漏、保障管道安全的关键手段.随着工业智能化发展,基于各种机器学习算法的压力预测技术被广泛应用,但现有模型普遍缺乏对初始输入数据的物理一致性校验,若输入数据本身包含早期泄漏特征或传感器漂移,极易导致后续预测结果出现误报或漏报.提出一种基于先验分布统计理论的初始压力数据校验方法,在停输过程的热—水力耦合机理的基础上,创新性引入"二十四节气"作为环境热力学特征的离散化表征,建立不同停输时长下的压力变化率先验分布.该模型能有效识别非正常的压力波动,在保证极低误报率(5.0%)的同时,对流通能力系数≤0.10 的微小泄漏保持了97.5%的高检出率,而对于流通能力系数≥0.15 的泄漏工况,检出率达到 100%.此校验方法可为后续机器学习预测模型提供高置信度的输入数据.
Pressure monitoring of long-distance refined oil pipelines under shutdown conditions is a critical approach for leak detection and pipeline safety assurance.With the advancement of industrial intelligence,pressure prediction technologies based on various machine learning algorithms have been widely applied.However,existing models generally lack physical consistency validation of initial input data.If the input data contains early leakage characteristics or sensor drift,it can easily lead to false alarms or missed detections in subsequent predictions.To address this issue,this paper proposes an initial pressure data validation method based on prior distribution statistical theory.The method is grounded in the thermo-hydraulic coupling mechanism during shutdown processes and innovatively introduces the"Twenty-four solar terms"as a discretized representation of environmental thermodynamic characteristics to establish prior distributions of pressure change rates under different shutdown durations.The model effectively identifies abnormal pressure fluctuations while maintaining an extremely low false alarm rate(5.0%),achieving a detection rate of 97.5%for minor leaks with flow coefficient≤0.10 and 100%detection rate for leak conditions with flow coefficient≥0.15,thereby providing high-confidence input data for subsequent machine learning prediction models.
杨文;吴玥蓉;黄晓茵;李晓平;李苗;宫敬
国家石油天然气管网集团有限公司华南分公司,广东 广州 510620中国石油大学(北京)机械与储运工程学院,北京 102249国家石油天然气管网集团有限公司华南分公司,广东 广州 510620中国石油大学(北京)机械与储运工程学院,北京 102249国家石油天然气管网集团有限公司华南分公司,广东 广州 510620中国石油大学(北京)机械与储运工程学院,北京 102249
成品油管道停输压力先验分布二十四节气
Refined oil pipelineShut-down pressurePrior distributionTwenty-four solar terms
《天然气与石油》 2026 (4)
10-20,11
国家石油天然气管网集团有限公司华南分公司项目"长输成品油管道停输期管内压力变化预测模型及系统开发"(HNGS-SSCC202401)
评论