Autonomous data analytics for pipeline leakage detection

用于管道泄漏检测的自主数据分析

• 批准号: 538630-2019
• 负责人: Zhao, Qing
• 金额: $ 1.82万
• 依托单位: University of Alberta
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2019
• 资助国家: 加拿大
• 起止时间: 2019-01-01 至 2020-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=680316
• 关键词: Autonomous; data; analytics; pipeline; leakage

Leakage detection in crude oil pipelines is extremely crucial because it can help prevent pollution and reduce damage to environment. Furthermore it can reduce high cost related to repairs and loss of oil. Leakage in crude oil pipelines can often be detected by measuring pressure signals. Working with TOKU Industry Inc. a company based in Alberta specializing in design of monitoring systems for Oil&Gas pipelines, we propose to develop an autonomous leakage detection scheme based on more robust and accurate leak pattern identification. The scheme consists of an onboard leak detection algorithm which is capable of generating real-time alarms, and an off-line pressure data model used to extract leakage patterns and trend their change over time. Main considerations of the design include improved detection accuracy by reducing the number of false alarms and missed alarms, as well as low computational complexity for the onboard detection algorithm to be implemented on the portable smart pressure sensors. In this work, we focus on developing cost-effective software solutions that can be easily deployed. Instead of constructing transient models based on hydrodynamics and experiments, we propose to construct a data model for leak signatures by exploring a great number of historical datasets.

原油管道泄漏检测对于防止污染、减少对环境的破坏至关重要。此外，它还可以降低与维修和石油损失相关的高昂成本。原油管道泄漏通常可以通过测量压力信号来检测。与总部位于艾伯塔省的Toku Industry Inc.合作，我们建议开发一种基于更稳健和准确的泄漏模式识别的自主泄漏检测方案。Toku Industry Inc.是一家位于艾伯塔省的专门设计油气管道监控系统的公司。该方案由能够产生实时报警的车载泄漏检测算法和用于提取泄漏模式并跟踪其随时间变化的离线压力数据模型组成。设计的主要考虑因素包括通过减少虚警和漏警的数量来提高检测精度，以及将车载检测算法实现在便携式智能压力传感器上的低计算复杂度。在这项工作中，我们专注于开发成本效益高、易于部署的软件解决方案。我们不是基于流体力学和实验来构建暂态模型，而是通过探索大量的历史数据集来构建泄漏特征的数据模型。