Evaluating the Use of Optical Gas Imaging Cameras for Above Ground Facilities

Natural gas is the largest primary energy source in the United States. Reliance on natural gas is only increasing as its role in electricity systems becomes more significant and that of coal diminishes. While this has air quality and health benefits over the use of coal, the global warming potential of methane - the primary component of natural gas - cannot be ignored. In order to mitigate methane leaks, periodic leak detection and repair programs are required in the United States. Various different technologies exist to detect and/or quantify methane leaks. Studying them and evaluating their performance is an important step in evaluating equivalence in emissions reductions between technologies.

In this study, we evaluate the performance of two optical gas imaging cameras. The first is the FLIR GF320, an infrared camera, which we coupled with the Providence Photonics QL320 to enable it to quantify methane leaks. The second is the Rebellion Photonics mini-GCI, a hyperspectral imaging camera, which provides automated alerts when a leak is detected. Experiments to test the two systems were carried out at the Methane Emissions Technology Evaluation Center in Fort Collins, Colorado over two weeks. We tested both technologies at a variety of leak size and imaging distance combinations. In order to better simulate real-world conditions, we also tested the performance of the two systems in the presence of different types of interference. For the first technology, we evaluate the quantification performance and for the second, the detection performance.

We report performance metrics at different distances, interference scenarios and leak sizes in the case of Rebellion. The two technologies differ in terms of automation, detection and quantification capabilities, imaging distance and minimum observable leak size. Quantification accuracy of the FLIR + Providence system was poor. Machine learning algorithms may be able to improve the quantification accuracy in this kind of system. The Rebellion Photonics mini-GCI had a high false positive rate under most conditions and might benefit from adjusting its sensitivity to allow some smaller leaks to go undetected while issuing fewer false positives. Different types of interference had either no effect or a negative effect on the performance of each technology. More data points are needed to gain tighter confidence intervals around the statistics presented.