US 12,229,689 B2
Hybrid modeling process for forecasting physical system parameters
Prasanna Amur Varadarajan, Clamart (FR); and Maurice Ringer, London (GB)
Assigned to SCHLUMBERGER TECHNOLOGY CORPORATION, Sugar Land, TX (US)
Filed by Schlumberger Technology Corporation, Sugar Land, TX (US)
Filed on Feb. 10, 2021, as Appl. No. 17/248,836.
Claims priority of provisional application 62/972,157, filed on Feb. 10, 2020.
Prior Publication US 2021/0248500 A1, Aug. 12, 2021
Int. Cl. G06N 5/04 (2023.01); G01V 20/00 (2024.01); G06F 30/27 (2020.01); G06N 20/00 (2019.01)
CPC G06N 5/04 (2013.01) [G01V 20/00 (2024.01); G06F 30/27 (2020.01); G06N 20/00 (2019.01)] 16 Claims
OG exemplary drawing
 
1. A method, comprising:
training a machine learning model to adjust modeled values generated by a physics model of a well system, wherein the training the machine learning model comprises:
receiving first input values for a first parameter of the well system, the well system including one or more surface components and one or more downhole components, wherein the first input values indicate at least one of: one or more parameters of the one or more surface components, one or more parameters of the one or more downhole components, or one or more parameters of a fluid in the well system;
calculating first modeled values for a second parameter using the physics model that represents the well system, wherein the calculating is based on the first input values, wherein the second parameter indicates a predicted current state or a predicted future state of the well system;
receiving, from one or more physical sensors in the well system, measured values for the second parameter; and
inputting the measured values for the second parameter and the first modeled values for the second parameter to the machine learning model to train the machine learning model to predict adjusted modeled values based on a difference between the first modeled values and the measured values;
receiving, by one or more processors of a computing system, second input values for the first parameter;
calculating, by the one or more processors of the computing system, second modeled values for the second parameter using the physics model;
generating, by the one or more processors of the computing system, adjusted values for the second parameter by adjusting the second modeled values using the trained machine learning model;
visualizing, by the one or more processors of the computing system, the adjusted values for the second parameter as representing operation of the well system;
predicting, by the one or more processors of the computing system, a wellbore-integrity event based on the adjusted values; and
controlling operation of the well system based on the predicted wellbore-integrity event, wherein controlling the operation includes adjusting one or more operating parameters of the well system to mitigate a risk of the wellbore-integrity event.