ASSESSING EMISSION DETECTION AND QUANTIFICATION SYSTEMS UNDER VARIED STRUCTURAL CONFIGURATIONS, METEOROLOGICAL CONDITIONS, AND CONTROLLED TESTING PROTOCOLS

Abstract

The potential use of continuous monitoring measurements for oil and gas inventory development has consistently motivated testing these solutions. Although some of these solutions are already used in voluntary emission reporting programs, their field performance remains poorly understood. Therefore, this study assesses the performance of continuous monitoring solutions using controlled testing protocols over four different testing campaigns.Phase 1 tested 13 continuous monitoring solutions over 12 weeks using a simplified controlled testing protocol. Release rates ranged from 0.08 to 6.75 kg CH4 hr-1 and lasted from 18 minutes to 8 hours. Six solutions demonstrated 90% method detection limits (DL90s), the emission rates at which the solutions have a 90% probability of detection across a range of test conditions, ranging from 0.5 [0.3, 0.6] kg CH4 hr-1 to 6.7 [5.8, NA] kg CH4 hr-1. Of these, five solutions had false-positive rates below 20%, ranging from 7.8% to 18.9%, and four had false-negative rates below 50%, ranging from 8.0% to 34.1%. All scanning/imaging solutions achieved localization precision and accuracy exceeding 40% at the equipment-unit level. Localization is the ability of a solution to pinpoint the source of an emission. Single emission estimates exhibited high relative errors. For small emissions (0.1 to 1 kg CH4 hr-1), errors ranged from 33 [0.9, 66] % (95% CI) to 1326 [1003, 1648] % (95% CI). For large emissions (> 1 kg CH4 hr-1), errors ranged from 3 [-20, 26] % (95% CI) to 3578 [-2832, 9988] % (95% CI). Although prior research has been conducted, the full potential of continuous monitoring solutions in real-world settings remains unclear. Therefore, the protocol was revised to better simulate field emissions and structural configurations. The revised protocol aims to test solutions against its operational objectives, using two testable interfaces: testable interface 1 and testable interface 2. Testable interface 1 assesses the accuracy of solution estimates against test center measurements, while testable interface 2 assesses a solution’s ability to alert an operator to the possibility of a failure condition at a facility. Testable interface 2 compares solution estimates and test center measurements against a solution-provided detection threshold. Phase 2 was a trial run of the revised protocol, which was completed over four weeks, with five solutions participating. Release rates ranged from 0.004 to 5.9 kg CH4 hr-1 and durations from 32 seconds to 24 hours. Results indicated that the alarm DL90, representing the emission rate above the threshold at which the solution detects emissions 90% of the time, exceeded the tested controlled-release rate range for two solutions, with values of 43.0 [25.4, 91.5] kg CH4 hr-1 and 63.0 [35.7, NA] kg CH4 hr-1. For the remaining three solutions, alarm DL90 could not be estimated. The relative error between controlled releases and reported masses ranged from -4% to +77%. The percentage of estimates within a factor of 3 of the controlled release rates ranged from 53% to 83%. The time to detection ranged from 0.3 to 1.8 hrs. Phase 3 was the first full implementation of the revised protocol, which tested three solutions over 16 weeks. Release rates ranged from 0.12 to 18 kg CH4 hr-1 and lasted from 10 seconds to 7 days. Two solutions were evaluated against their detection thresholds, while one was assessed against EPA’s threshold of 0.4 kg CH4 hr-1. The alarm DL90 for one solution was 0.8 [0.2, 2.0] kg CH4 hr-1; it could not be estimated for the other two solutions. The false-positive and false-negative rates ranged from 0.26 to 6.3% and from 1.7% to 18.4%, respectively. The relative error between controlled releases and reported masses ranged from -30% to +85%. The percentage of solution estimates within a factor of 3 of the test center measurements ranged from 60% to 86%. The mean time to detection ranged from 1.4 to 40.2 hours. Phase 4 was a second implementation of the revised protocol conducted at an active oil and gas site in Alberta, Canada, over 10 weeks. Four continuous monitoring solutions participated in single-blind testing. Release rates ranged from 0.12 to 8.52 kg CH4 hr-1 and lasted from 1 minute to 8 hours. The results show that the DL90 could not be estimated for all solutions. The false-positive and false-negative rates ranged from 1.3% to 52.0%, and from 6.0% to 15.0%, respectively. The percentage of reported single estimates that were within a factor of 3 of the controlled-release measurements ranged from 41% to 97% across all solutions. The mass-reported emissions were higher than the test center mass emissions by a factor of 1.1 to 14.8 across all solutions. The mean time to detection ranged from 0.06 to 0.24 hours. Overall, this work demonstrates that continuous monitoring solutions can enable rapid detection, with varying quantification capabilities. Solutions perform well across different metrics, suggesting the potential value of complementing solutions in actual field use rather than relying on a single solution for all functionalities. Finally, the work illustrates a wide variation between performance in field testing and controlled setups, such as at the METEC facility, suggesting that continuous, rigorous, and additional field testing is needed to better understand these solutions' performance and the extent to which the results transfer to real-world use.