The Performance of Optical Gas Imaging for Leak Detection in Pseudo-Realistic Controlled Conditions

Optical gas imaging (OGI) uses video cameras filtered to mid-IR wavelengths to image gas emissions at natural gas and industrial sites. While the cameras have been assessed in many experimental and theoretical studies, the emission detection performance of professional camera operators, using OGI cameras in realistic controlled conditions, has not been previously assessed.

In a CSU-led study 39 profession leak detection operators from 24 organizations - oil and gas operators, contractors and regulatory compliance teams - completed 29 combined days of testing at the Methane Emissions Technology Evaluation Center (METEC) at Colorado State University. METEC reconstructs a working gas production environment, including wells, separators, tanks and auxiliary equipment. Gas is metered and released from typical leak locations, at realistic rates. The single-blind trials were used, with methane leak rates of 19-375 g/h. Testing was spread from February to November to test performance in a wide range of weather conditions, including conditions under which OGI surveys would typically not be conducted.

As previously reported, detection rates drop with increasing wind and decreasing leak rates for all leak locations. In contrast to previous studies, we find the lower detection rates overall. For nearly all tests, a 100% detection rate is not achieved for any flow rate on any equipment type, even with standard operating wind speeds. However, further analysis indicates that self-reported operator experience is a strong indicator of detection performance. Operators who have screened more than 400 facilities have a detection rate over a range of conditions that is nearly double that of operators who had screened less than 100 facilities prior to the study. Qualitative analysis of captured OGI video indicates that performance differences may be due to experienced operators' intuitive feel for viewing angles that maximize the contrast between components and background scenes.

Results of this study indicates that a different training may regime for OGI operators may be advisable, and that survey protocols may need emphasize the need to view inspected component against several, contrasting, backgrounds during surveys.