Interpreting seismic faults with machine learning techniques

1 . A method for interpreting seismic data, the method comprising:

receiving seismic data that represents a current condition of a subterranean volume in an oil and gas reservoir, the seismic data being received from at least one sensor disposed in the subterranean volume in the oil and gas reservoir;

training a first machine learning technique to:

receive one or more vertical fault lines in a seismic volume based on the seismic data; and

generate one or more inline probability values and one or more crossline probability values based on the received one or more vertical fault lines;

generating a merged data set by combining the one or more inline probability values and the one or more crossline probability values;

training a second machine learning technique based on a subset of labeled horizontal planes from the merged data set, the second machine learning technique being trained to:

receive one or more horizontal fault lines from the seismic volume from the merged data set;

remove false positive data points from the subset of labeled horizontal planes to improve continuity and consistency of the one or more inline probability values and the one or more crossline probability values generated by the first machine learning technique; and

identify high probability data values that are likely false positives for the identified one or more horizontal fault lines;

generating a representation of the seismic volume, the representation comprising a display of a plurality of three-dimensional continuous fault structures within the seismic volume including at least one of predicted horizontal fault lines or predicted vertical fault lines at locations where the seismic data does not include a fault line being represented by a coherent planar surface, the representation of the seismic volume identifying one or more of:

at least one region of the reservoir to be avoided in conventional drilling due to a presence of the at least one of predicted horizontal fault lines or predicted vertical fault lines; or

at least one region of the reservoir to be utilized in unconventional drilling comprising the at least one of predicted horizontal fault lines or predicted vertical fault lines enabling removal of at least one of oil or gas from the reservoir by allowing the at least one of oil or gas to travel through the at least one of predicted horizontal fault lines or predicted vertical fault lines and to become trapped in the reservoir; and

creating a three-dimensional subsurface model of the subterranean volume, the three-dimensional subsurface model comprising the representation of the seismic volume comprising the displayed plurality of three-dimensional fault structures including the at least one of the predicted horizontal fault lines or the predicted vertical fault lines, the generated one or more inline probability values and one or more crossline probability values indicating a likelihood of a presence of the at least one of the predicted horizontal fault lines or the predicted vertical fault lines at locations where the seismic data does not include a fault line;

selecting an oil and gas production plan based on the three-dimensional subsurface model, the oil and gas production plan comprising functional parameters for operation of a production tool, a weight on a bit, a pump rate, or a physical parameter of the production tool, the selected oil and gas production plan comprising one or more of:

avoiding the at least one region of the reservoir to be avoided due to the presence of the at least one of predicted horizontal fault lines or predicted vertical fault lines; or

utilizing the at least one region of the reservoir to be utilized comprising the at least one of predicted horizontal fault lines or predicted vertical fault lines enabling removal of at least one of oil or gas from the reservoir by allowing the at least one of oil or gas to travel through the at least one of predicted horizontal fault lines or predicted vertical fault lines and to become trapped in the reservoir;

transmitting the selected oil and gas production plan to equipment to adjust a retrieval of a resource from the oil and gas reservoir, the adjusting the retrieval comprising controlling at least one of: the operation of a production tool, the weight on the bit, the pump rate, or the physical parameter of the production tool to optimize drilling in accordance with the selected oil and gas production plan; and

retrieving the resource from the oil and gas reservoir, the retrieving comprising automatically controlling the at least one of: the operation of the production tool, the weight on the bit, the pump rate, or the physical parameter of the production tool to optimize the drilling in accordance with the selected oil and gas production plan and based on the current condition of the subterranean volume.

2 . The method of claim 1 , wherein the representation is a three-dimensional digital representation that is visualized on display.

3 . The method of claim 1 , wherein the generating the merged data set comprises applying a maximum value function to the one or more inline probability values and the one or more crossline probability values.

4 . The method of claim 1 , further comprising:

generating an inline prediction probability cube based on the one or more inline probability values; and

generating a crossline prediction probability cube based on the one or more crossline prediction values, the crossline prediction probability cube and the inline prediction probability cube being generated based on the seismic volume provided to the first machine learning technique.

5 . The method of claim 1 , wherein the first machine learning technique is a first neural network and the second machine learning technique is a second neural network.

6 . The method of claim 5 , wherein the second neural network is further trained with a second subset of labeled horizontal planes from a previously analyzed seismic volume, the previously analyzed seismic volume comprising a merged probability cube generated by the first neural network.

7 . The method of claim 6 , wherein the second neural network is configured to identify a horizontal fault line in a subsequent seismic volume without a subset of labeled horizontal planes from the subsequent seismic volume.

8 . The method of claim 1 , wherein the second machine learning technique is smaller than the first machine learning technique and includes at least one of fewer layers of neurons or fewer neurons per layer than the first machine learning technique.

9 . The method of claim 8 , wherein the neurons of the second machine learning technique are not fully connected.

10 . A computing system for interpreting seismic data, comprising:

one or more processors; and

a memory system comprising one or more non-transitory computer-readable media storing instructions that, when executed by at least one of the one or more processors, cause the computing system to perform operations, the operations comprising:

receiving seismic data that represents a current condition of a subterranean volume in an oil and gas reservoir, the seismic data being received from at least one sensor disposed in the subterranean volume in the oil and gas reservoir;

training a first machine learning technique to:

receive one or more vertical fault lines in a seismic volume based on the seismic data; and

generate one or more inline probability values and one or more crossline probability values;

generating a merged data set by combining the one or more inline probability values and the one or more crossline probability values;

training a second machine learning technique based on a subset of labeled horizontal planes from the merged data set, the second machine learning technique being trained to:

receive one or more horizontal fault lines from the seismic volume from the merged data set;

remove false positive data points from the subset of labeled horizontal planes to improve continuity and consistency of the one or more inline probability values and the one or more crossline probability values generated by the first machine learning technique; and

identify high probability data values that are likely false positives for the identified one or more horizontal fault lines;

generating a representation of the seismic volume, the representation comprising a display of a plurality of three-dimensional continuous fault structures within the seismic volume including at least one of predicted horizontal fault lines or predicted vertical fault lines at locations where the seismic data does not include a fault line being represented by a coherent planar surface, the representation of the seismic volume identifying one or more of:

at least one region of the reservoir to be avoided in conventional drilling due to a presence of the at least one of predicted horizontal fault lines or predicted vertical fault lines; or

at least one region of the reservoir to be utilized in unconventional drilling comprising the at least one of predicted horizontal fault lines or predicted vertical fault lines enabling removal of at least one of oil or gas from the reservoir by allowing the at least one of oil or gas to travel through the at least one of predicted horizontal fault lines or predicted vertical fault lines and to become trapped in the reservoir; and

creating a three-dimensional subsurface model of the subterranean volume, the three-dimensional subsurface model comprising the representation of the seismic volume comprising the displayed plurality of three-dimensional fault structures including the at least one of the predicted horizontal fault lines or the predicted vertical fault lines, the generated one or more inline probability values and one or more crossline probability values indicating a likelihood of a presence of the at least one of the predicted horizontal fault lines or the predicted vertical fault lines at locations where the seismic data does not include a fault line;

selecting an oil and gas production plan based on the three-dimensional subsurface model, the oil and gas production plan comprising functional parameters for operation of a production tool, a weight on a bit, a pump rate, or a physical parameter of the production tool, the selected oil and gas production plan comprising one or more of:

avoiding the at least one region of the reservoir to be avoided due to the presence of the at least one of predicted horizontal fault lines or predicted vertical fault lines; or

utilizing the at least one region of the reservoir to be utilized comprising the at least one of predicted horizontal fault lines or predicted vertical fault lines enabling removal of at least one of oil or gas from the reservoir by allowing the at least one of oil or gas to travel through the at least one of predicted horizontal fault lines or predicted vertical fault lines and to become trapped in the reservoir;

transmitting the selected oil and gas production plan to equipment to adjust a retrieval of a resource from the oil and gas reservoir, the adjusting the retrieval comprising controlling at least one of: the operation of a production tool, the weight on the bit, the pump rate, or the physical parameter of the production tool to optimize drilling in accordance with the selected oil and gas production plan; and

retrieving the resource from the oil and gas reservoir, the retrieving comprising automatically controlling the at least one of: the operation of the production tool, the weight on the bit, the pump rate, or the physical parameter of the production tool to optimize the drilling in accordance with the selected oil and gas production plan and based on the current condition of the subterranean volume.

11 . The computing system of claim 10 , wherein the generating the merged data set comprises applying a maximum value function to the one or more inline probability values and the one or more crossline probability values.

12 . The computing system of claim 10 , wherein the seismic volume comprises a set of data representing one or more subterranean rock formation characteristics.

13 . The computing system of claim 10 , wherein:

the operations further comprise:

generating an inline prediction probability cube based on the one or more inline probability values; and

generating a crossline prediction probability cube based on the one or more crossline prediction values; and

the crossline prediction probability cube and the inline prediction probability cube are generated based on the seismic volume provided to the first machine learning technique.

14 . The computing system of claim 10 , wherein the first machine learning technique is a first neural network and the second machine learning technique is a second neural network.

15 . The computing system of claim 14 , wherein the second neural network is further trained with a second subset of labeled horizontal planes from a previously analyzed seismic volume, the previously analyzed seismic volume comprising a merged probability cube generated by the first neural network.

16 . The computing system of claim 15 , wherein the second neural network is configured to identify a horizontal fault line in a subsequent seismic volume without a subset of labeled horizontal planes from the subsequent seismic volume.

17 . At least one non-transitory computer-readable medium for interpreting seismic data, the at least one computer-readable medium comprising a plurality of computer-executable instructions that, in response to execution by a processor, cause the processor to:

receive seismic data that represents a current condition of a subterranean volume in an oil and gas reservoir, the seismic data being received from at least one sensor disposed in the subterranean volume in the oil and gas reservoir;

train a first machine learning technique to:

receive one or more vertical fault lines in a seismic volume based on the seismic data; and

generate one or more inline probability values and one or more crossline probability values based on the received one or more vertical fault lines;

generate a merged data set by combining the one or more inline probability values and the one or more crossline probability values;

train a second machine learning technique based on a subset of labeled horizontal planes from the merged data set, the second machine learning technique being trained to:

receive one or more horizontal fault lines from the seismic volume from the merged data set;

remove false positive data points from the subset of labeled horizontal planes to improve continuity and consistency of the one or more inline probability values and the one or more crossline probability values generated by the first machine learning technique; and

identify high probability data values that are likely false positives for the identified one or more horizontal fault lines;

generate a representation of the seismic volume, the representation comprising a display of a plurality of three-dimensional continuous fault structures within the seismic volume including at least one of predicted horizontal fault lines or predicted vertical fault lines at locations where the seismic data does not include a fault line being represented by a coherent planar surface, the representation of the seismic volume identifying one or more of:

at least one region of the reservoir to be avoided in conventional drilling due to a presence of the at least one of predicted horizontal fault lines or predicted vertical fault lines; or

at least one region of the reservoir to be utilized in unconventional drilling comprising the at least one of predicted horizontal fault lines or predicted vertical fault lines enabling removal of at least one of oil or gas from the reservoir by allowing the at least one of oil or gas to travel through the at least one of predicted horizontal fault lines or predicted vertical fault lines and to become trapped in the reservoir; and

create a three-dimensional subsurface model of the subterranean volume, the three-dimensional subsurface model comprising the representation of the seismic volume comprising the displayed plurality of three-dimensional fault structures including the at least one of the predicted horizontal fault lines or the predicted vertical fault lines, the generated one or more inline probability values and one or more crossline probability values indicating a likelihood of a presence of the at least one of the predicted horizontal fault lines or the predicted vertical fault lines at locations where the seismic data does not include a fault line;

select an oil and gas production plan based on the three-dimensional subsurface model, the oil and gas production plan comprising functional parameters for operation of a production tool, a weight on a bit, a pump rate, or a physical parameter of the production tool, the selected oil and gas production plan comprising one or more of:

avoiding the at least one region of the reservoir to be avoided due to the presence of the at least one of predicted horizontal fault lines or predicted vertical fault lines; or

utilizing the at least one region of the reservoir to be utilized comprising the at least one of predicted horizontal fault lines or predicted vertical fault lines enabling removal of at least one of oil or gas from the reservoir by allowing the at least one of oil or gas to travel through the at least one of predicted horizontal fault lines or predicted vertical fault lines and to become trapped in the reservoir;

transmit the selected oil and gas production plan to equipment to adjust a retrieval of a resource from the oil and gas reservoir, the adjusting the retrieval comprising controlling at least one of: the operation of a production tool, the weight on the bit, the pump rate, or the physical parameter of the production tool to optimize drilling in accordance with the selected oil and gas production plan; and

retrieve the resource from the oil and gas reservoir, the retrieving comprising automatically controlling the at least one of: the operation of the production tool, the weight on the bit, the pump rate, or the physical parameter of the production tool to optimize the drilling in accordance with the selected oil and gas production plan and based on the current condition of the subterranean volume.