Simultaneous quantification and changepoint detection of point source gas emissions using recursive Bayesian inference

TL;DR

This paper introduces a recursive Bayesian changepoint detection algorithm for real-time identification of abnormal gas emissions, demonstrating high success rates in controlled experiments and adaptability to stationary sensor data.

Contribution

The paper presents a novel recursive Bayesian framework for simultaneous emission quantification and fault detection, improving rapid identification of equipment faults in methane emission monitoring.

Findings

Successful detection (>90%) of emission rate changes in experiments

Measurement statistics predict algorithm performance

Method adaptable to stationary sensor concentration data

Abstract

Recent findings suggest that abnormal operating conditions of equipment in the oil and gas supply chain represent a large fraction of anthropogenic methane emissions. Thus, effective mitigation of emissions necessitates rapid identification and repair of sources caused by faulty equipment. In addition to advances in sensing technology that allow for more frequent surveillance, prompt and cost-effective identification of sources requires computational frameworks that provide automatic fault detection. Here, we present a changepoint detection algorithm based on a recursive Bayesian scheme that allows for simultaneous emission rate estimation and fault detection. The proposed algorithm is tested on a series of near-field controlled release mobile experiments, with promising results demonstrating successful detection (>90% success rate) of changes in the leak rate when the emission rate is tripled after an abrupt change. Moreover, we show that the statistics of the measurements, such as the coefficient of variation and range are good predictors of the performance of the algorithm. Finally, we describe how this methodology can be easily adapted to suit time-averaged concentration data measured by stationary sensors, thus showcasing its flexibility.