Field operations system with filter

Patent metadata

Field	Value
Publication number	US-12078048-B2
Application number	US-202318308881-A
Country	US
Kind code	B2
Filing date	Apr 28, 2023
Priority date	Nov 15, 2017
Publication date	Sep 3, 2024
Grant date	Sep 3, 2024

Abstract

A system and method that can include training a deep neural network using time series data that represents functions of a non-linear Kalman filter that represents a dynamic system of equipment and environment and models a pre-defined operational procedure as a temporal sequence. The system and method can also include receiving operation data from the equipment responsive to operation in the environment and outputting an actual operation as an actual sequence of operational actions by the deep neural network. The system and method can additionally include performing an operation-level comparison to evaluate the temporal sequence against the actual sequence using a distance function in a latent space of the deep neural network and outputting a score function that quantifies the distance function in the latent space. The system and method can further include controlling an electronic component to execute an electronic operation based on the score function.

First claim

The invention claimed is:

1. A method comprising:
training a deep neural network using time series data that represents functions of a non-linear Kalman filter that represents a dynamic system of equipment and environment and models a pre-defined operational procedure as a temporal sequence of desired operational actions by the deep neural network;
receiving operation data from the equipment responsive to operation in the environment and outputting an actual operation as an actual sequence of actual operational actions by the deep neural network;
performing an operation-level comparison to evaluate the temporal sequence against the actual sequence using a comparison function in a latent space of the deep neural network, classifying one or more actual operation internal states by one or more activity labels from the pre-defined operational procedure, and outputting a score function that quantifies the comparison function in the latent space, wherein the comparison function in the latent space quantifies compliance of each of the actual operational actions with each of the desired operational actions; and
controlling a well construction operation based on the score function.

2. The method of claim 1 , wherein the pre-defined operational procedure comprises a series of proscribed actions that are defined as a sequence of internal states.

3. The method of claim 1 , further including representing the pre-defined operational procedure through a simulated control sequence, the one or more activity labels, and corresponding actual operation internal states by the deep neural network.

4. The method of claim 1 , wherein the deep neural network is trained using multi-channel data.

5. The method of claim 4 , wherein the multi-channel data is utilized as an input to the deep neural network with a deep Kalman filter to output the actual operation as the actual sequence of actual operational actions.

6. The method of claim 1 , wherein the deep neural network comprises a convolution neural network layer and a long short-term memory layer.

7. The method of claim 1 , wherein performing the operation-level comparison includes generating an internal state vector in the latent space using the operation data and the deep neural network and comparing at least the internal state vector to at least a base internal state vector in the latent space that corresponds to the pre-defined operational procedure.

8. The method of claim 1 , wherein controlling the well construction operation includes controlling a display to render a graphical representation of the actual operation and the pre-defined operational procedure.

9. The method of claim 1 , wherein controlling the well construction operation includes controlling at least one piece of the equipment at a rig site to perform at least one operation pertaining to well construction.

10. A system comprising:
a processor;
memory accessible by the processor;
processor-executable instructions stored in the memory and executable to instruct the system to:
train a deep neural network using time series data that represents functions of a non-linear Kalman filter that represents a dynamic system of equipment and environment and models a pre-defined operational procedure as a temporal sequence of desired operational actions by the deep neural network;
receive operation data from the equipment responsive to operation in the environment and output an actual operation as an actual sequence of actual operational actions by the deep neural network;
perform an operation-level comparison to evaluate the temporal sequence against the actual sequence using a comparison function in a latent space of the deep neural network, classify one or more actual operation internal states by one or more activity labels from the pre-defined operational procedure, and output a score function that quantifies the comparison function in the latent space, wherein the comparison function in the latent space quantifies compliance of each of the actual operational actions with each of the desired operational actions; and
control a well construction operation based on the score function.

11. The system of claim 10 , wherein the comparison function in the latent space is transformed via principle component analysis to output a plot in a processed space to display dynamics of the well construction operation.

12. The system of claim 10 , wherein the system is instructed to represent the pre-defined operational procedure through a simulated control sequence, the one or more activity labels, and corresponding actual operation internal states by the deep neural network.

13. The system of claim 10 , wherein the deep neural network is trained using multi-channel data.

14. The system of claim 13 , wherein the multi-channel data is utilized as an input to the deep neural network with a deep Kalman filter to output the actual operation as the actual sequence of actual operational actions.

15. The system of claim 10 , wherein the deep neural network comprises a convolution neural network layer and a long short-term memory layer.

16. The system of claim 10 , wherein the system is instructed to generate an internal state vector in the latent space using the operation data and the deep neural network and compare at least the internal state vector to at least a base internal state vector in the latent space that corresponds to the pre-defined operational procedure.

17. The system of claim 10 , wherein the system is instructed to control a display to render a graphical representation of the actual operation and the pre-defined operational procedure.

18. The system of claim 10 , wherein the system is instructed to control the well construction operation comprising control of at least one piece of the equipment at a rig site to perform at least one operation pertaining to well construction.

19. A non-transitory computer-readable storage medium storing instructions that when executed by a computer, which includes a processor performs a method, the method comprising:
training a deep neural network using time series data that represents functions of a non-linear Kalman filter that represents a dynamic system of equipment and environment and models a pre-defined operational procedure as a temporal sequence of desired operational actions by the deep neural network;
receiving operation data from the equipment responsive to operation in the environment and outputting an actual operation as an actual sequence of actual operational actions by the deep neural network;
performing an operation-level comparison to evaluate the temporal sequence against the actual sequence using a comparison function in a latent space of the deep neural network and outputting a score function that quantifies the comparison function in the latent space, wherein the comparison function in the latent space quantifies compliance of each of the actual operational actions with each of the desired operational actions; and
controlling a well construction operation based on the score function.

20. The non-transitory computer-readable storage medium of claim 19 , wherein performing the operation-level comparison includes generating an internal state vector in the latent space using the operation data and the deep neural network and comparing at least the internal state vector to at least a base internal state vector in the latent space that corresponds to the pre-defined operational procedure.

Assignees

• Schlumberger Technology Corp

Inventors

• Yu Yingwei
• Liu Qiuhua
• Meehan Richard John
• Chambon Sylvain
• Hamzah Mohammad Khairi

Classifications

• G06N3/08

  Learning methods · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/045

  Combinations of networks · CPC title

• E21B21/08

  Controlling or monitoring pressure or flow of drilling fluid, e.g. automatic filling of boreholes, automatic control of bottom pressure (valve arrangements therefor E21B21/10) · CPC title

• G06N3/09

  Supervised learning · CPC title