Picarro, Inc.

Sean Patrick MacMullin

Chris W Rella

Sze Meng Tan

David Steele

Yonggang He

Aaron Van Pelt

Tim Clark

Alex Balkanski

Anders Nottrott

Benjamin Cohen-Stead

Chris W Relia

In some embodiments, data from multiple vehicle-based natural gas leak detection survey runs are used by computer-implemented machine learning systems to generate a list of natural gas leaks ranked by hazard level. A risk model embodies training data having known hazard levels, and is used to classify newly-discovered leaks. Hazard levels may be expressed by continuous variables, and/or probabilities that a given leak fits within a predefined category of hazard (e.g. Grades 1-3). Each leak is represented by a cluster of leak indications (peaks) originating from a common leak source. Hazard-predictive features may include maximum, minimum, mean, and/or median CH4/amplitude of aggregated leak indications; estimated leak flow rate, determined from an average of leak indications in a cluster; likelihood of leak being natural gas based on other indicator data (e.g. ethane concentration); probability the leak was detected on a given pass; and estimated distance to leak source.

Leak detection event aggregation and ranking systems and methods

In some embodiments, data from multiple vehicle-based natural gas leak detection survey runs are used by computer-implemented machine learning systems to generate a list of natural gas leaks ranked by hazard level. A risk model embodies training data having known hazard levels, and is used to classify newly-discovered leaks. Hazard levels may be expressed by continuous variables, and/or probabilities that a given leak fits within a predefined category of hazard (e.g. Grades 1-3). Each leak is represented by a cluster of leak indications (peaks) originating from a common leak sources. Hazard-predictive features may include maximum, minimum, mean, and/or median CH4/amplitude of aggregated leak indications; estimated leak flow rate, determined from an average of leak indications in a cluster; likelihood of leak being natural gas based on other indicator data (e.g. ethane concentration); probability the leak was detected on a given pass; and estimated distance to leak source.

In some embodiments, data from a vehicle-borne gas leak detection survey are used to generate an aggregate leak indication search area (LISA) indicator for a plurality of leak indications (measurement peaks) characterizing a single leak or localized set of leaks. A clustering algorithm (e.g. Markov, DBScan) may be used to group a set of indications into a cluster characterizing the leak. Leak indications may be pre-filtered for quality control before assignment to a cluster according to a number of parameters including background gas level, inter-peak distance, peak shape, wind speed, wind direction and/or variability, vehicle speed and/or acceleration, and/or a lower detection threshold for leak flow rate.

Aggregate leak indicator display systems and methods

Flux estimates for gas plumes from gas leaks are obtained from a 1-D horizontal line scan of gas concentration measurements, combined with an estimate of the vertical extent of the gas plume. In this manner, flux estimates for gas plumes can be obtained without having to gather a 2-D image of gas concentration data.

Emission quantification using a line scan of gas concentration data

In some embodiments, computer-implemented systems/methods detect and/or quantify changes in emission rates of gas emission sources (e.g. natural gas leaks originating from underground distribution pipelines) using data from multiple vehicle-based measurement runs. Exemplary described methods aim to address the observation that large (e.g. 10&#xd7;) changes in gas concentrations away from a source may be observed even in the absence of significant changes in source emission rate, due to changes in wind or other atmospheric conditions and local spatial variations in gas concentrations. Described methods are useful for identifying large increases in the emission rate(s) of known sources, for example due to frost heave or other dislocations. Multiple runs are performed along the same survey path in closely-related conditions (e.g. same time of day, same lanes), and a statistical test (e.g. a Kolmogorov-Smirnov test) is used to identify changes in concentration reflecting changes in emission rates.

Growing community of inventors

Livermore, CA, United States of America

Sean Patrick MacMullin