High Precision Pipeline Leak Detection and Localization Using Negative Pressure Wave Technique: An Application in a Real Field Case Study

One of the most important aspects of oil and gas production is the safe and efficient fluid transportation using pipelines. Pipelines transporting various fluids are the most efficient but are susceptible to failure and leaks. These leaks can come about through natural disaster, as well as from general wear from the pipes that could result in major environmental and economic problems. The ability to detect leaks with speed and accuracy, as well as locating these leaks within a narrow range, will aid with the maintenance response. Hasty responses will minimize the revenue loss and reduce potential environmental impact but bring about various computational challenges. Among all the leak detection techniques used in the industry the Negative Pressure Wave (NPW) is the most popular and cost-effective technique. Pressure analysis of several transducers makes it possible to both identify and locate the leak. However, there are several challenges to analyzing such pressure transducer data. It is extremely noisy (low quality data), there is a high noise to data ratio, requiring computationally expensive processes to denoise and make legible. Secondly, the initial pressure drop caused by the leak will dissipate quickly and the negative pressure wave decays as the system reaches a new equilibrium condition. The pressure data is also convoluted with both known and spontaneous events (i.e., multiple pumps and possible leak events). Finally, the robustness of the system needs to be verified to avoid complications and extra cost associated with false leak events detected. To remedy this issue, a new workflow is designed and applied in both complex real field flow networks in Texas and further assessed in a complex system with multiple and random leak and pump events. The new workflow incorporates i) data preprocessing including data cleansing, normalization and denoising; ii) developing dynamic pressure control limit lines for detecting and deconvolution of the pump events from actual leak events; iii) Performing multiple transducer analysis techniques to reduce and eliminate the possibility of the false events; iv) developing flow simulation software built on open-source Python package called WNTR to generate synthetic leak scenarios v) Finally, constructing a dashboard using the Python programming language and the Plotly open source graphing libraries for near real time visualization of different transducers response, quality check and verification of leak events and finally locating the leak event on the flow network map. Three months of data collected from a flow network is analyzed and one leak event is identified and confirmed with the operator. The leak occurred in the close vicinity of in-line pressure transducer #19 and the exact location was identified. The workflow is tested on a real network with synthetic leaks and high precision 10 and 1 millisecond recording and leak events are detected with 10-meter accuracy. The workflow showed great capability to be integrated with the SCADA system and being able to be used for near real time leak detection