US 12,264,580 B2
Detecting gas leaks in oil wells using machine learning
Mohammed Y. Al Daif, Qatif (SA); and Sultan S. Al Sumat, Dammam (SA)
Assigned to Saudi Arabian Oil Company, Dhahran (SA)
Filed by Saudi Arabian Oil Company, Dhahran (SA)
Filed on Sep. 1, 2021, as Appl. No. 17/464,080.
Prior Publication US 2023/0063604 A1, Mar. 2, 2023
Int. Cl. E21B 47/11 (2012.01); E21B 47/002 (2012.01); E21B 47/103 (2012.01); E21B 47/113 (2012.01); E21B 47/117 (2012.01); G06F 18/214 (2023.01); G06N 3/08 (2023.01)
CPC E21B 47/117 (2020.05) [E21B 47/002 (2020.05); E21B 47/103 (2020.05); E21B 47/113 (2020.05); G06F 18/214 (2023.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
obtaining, using one or more processors, first data regarding a first oil well, wherein the first data comprises one or more first thermal images of the oil well generated by one or more first thermal cameras;
determining, using the one or more processors implementing a computerized neural network, a presence of a gas leak at one or more locations on the first oil well based on the first data, wherein the one or more locations comprise at least one of:
a first location along a pipeline configured to convey gas to a flare area of the first oil well, or
a second location at a rig floor of the first oil well; and
responsive to determining the presence of the gas leak at the one or more locations, generating, using the one or more processors, a notification indicating the presence of the gas leak at the one or more locations,
wherein the computerized neural network comprises a plurality of interconnected nodes, including:
a plurality of input nodes,
a plurality of output nodes, and
a plurality of weighted nodes interconnecting the plurality of input nodes and the plurality of output nodes,
wherein the computerized neural network is trained to determine one or more transfer functions, wherein the one or more transfer functions define a relationship between the plurality of input nodes and the plurality of output nodes according to the plurality of weighted nodes,
wherein at least some of the input nodes of the computerized neural network corresponds to the first data,
wherein at least one of the output nodes of the computerized neural network corresponds to a first likelihood that the gas leak is present at the one or more locations on the first oil well, and
wherein at least another one of the output nodes of the computerized neural network corresponds to a second likelihood that one or more conditions of an ambient environment of the first oil well have changed.