Back to results
US 12,399,296 B2Granted
System and method for automatic well log depth matching
AssigneeSchlumberger Technology Corporation
InventorsTimon Zimmermann, Lin Liang, Smaine Zeroug
Life of this patent
Estimating the term…
Abstract
A method and computing system device for receiving a plurality of well logs. A depth shift between at least one well log of the plurality of well logs and at least one other well log may be determined based upon, at least in part, processing the plurality of well logs with a neural network. The plurality of well logs may be matched with one another based upon, at least in part, the depth shift between the at least one well log and the at least one other well log.
Claims16
A computer-implemented method for well log depth matching, comprising:
- receiving, on a computing device, a plurality of well logs;
- training a neural network to generate a trained neural network configured to determine a depth shift between a first well log of the plurality of well logs corresponding to a first well and a second well log of the plurality of well logs corresponding to a second well, wherein training the neural network comprises:
- generating training data via one or more data augmentation transformations randomly applied on at least one well log of the plurality of well logs, wherein the one or more data augmentation transformations comprise applying additive noise to the at least one well log of the plurality of well logs;
- applying a first accuracy metric indicative of a proportion of output that is representative of a ground-truth to total output;
- and
- applying a second accuracy metric that increases the output that is representative of the ground-truth by including outputs representative of a convergence towards the ground-truth;
- and
- determining, using the trained neural network, the depth shift between the first well log and the second well log, wherein determining the depth shift comprises:
- automatically generating a plurality of points of a reference signal corresponding to at least a portion of the first well log;
- determining a point of interest of the plurality of points of the reference signal, wherein the point of interest corresponds to a first depth of the first well;
- creating a first window around the point of interest, wherein the first window comprises a first set of points;
- identifying a corresponding point in a desynchronized signal corresponding to at least a portion of the second well log using the trained neural network, wherein the corresponding point corresponds to the point of interest and a second depth of the second well;
- creating a second window around the corresponding point, wherein the second window comprises a second set of points;
- inputting the first set of points and the second set of points into the trained neural network to obtain an output from the trained neural network comprising a probability over possible depth shifts between the first set of points and the second set of points;
- and
- determining the depth shift between the first well log and the second well log based upon, at least in part, the probability over possible depth shifts between the first set of points and the second set of points;
- and
- matching the plurality of well logs with one another based upon, at least in part, the depth shift between the first well log and the second well log.
15 more elements
A computing system including one or more processors and one or more memories configured to perform operations comprising:
- receiving a plurality of well logs;
- training a neural network to generate a trained neural network configured to determine a depth shift between a first well log of the plurality of well logs corresponding to a first well and a second well log of the plurality of well logs corresponding to a second well, wherein training the neural network comprises:
- generating training data via one or more data augmentation transformations randomly applied on at least one well log of the plurality of well logs, wherein the one or more data augmentation transformations comprise applying additive noise to the at least one well log of the plurality of well logs;
- applying a first accuracy metric indicative of a proportion of output that is representative of a ground-truth to total output;
- and
- applying a second accuracy metric that increases the output that is representative of the ground-truth by including outputs representative of a convergence towards the ground-truth;
- determining, using the trained neural network, the depth shift between the first well log of the plurality of well logs and the second well log, wherein determining the depth shift comprises:
- automatically generating a plurality of points of a reference signal corresponding to at least a portion of the first well log;
- determining a point of interest of the plurality of points of the reference signal, wherein the point of interest corresponds to a first depth of the first well;
- creating a first window around the point of interest, wherein the first window comprises a first set of points;
- identifying a corresponding point in a desynchronized signal corresponding to at least a portion of the second well log using the trained neural network, wherein the corresponding point corresponds to the point of interest and a second depth of the second well;
- creating a second window around the corresponding point, wherein the second window comprises a second set of points;
- inputting the first set of points and the second set of points into the trained neural network to obtain an output from the trained neural network comprising a probability over possible depth shifts between the first set of points and the second set of points;
- and
- determining the depth shift between the first well log and the second well log based upon, at least in part, the probability over possible depth shifts between the first set of points and the second set of points;
- and
- matching the plurality of well logs with one another based upon, at least in part, the depth shift between the first well log and the second well log.
14 more elements
Dependencies are inferred from the claim text.