System and method for failure prediction for rod pump artificial lift systems
First Claim
Patent Images
1. A computer-implemented method for failure prediction for artificial lift well systems, the method comprising:
- obtaining well data representing a plurality of measured attributes from a plurality of wells over a time period;
generating features from the well data based on the measured attributes;
applying a random peek semi-supervised machine learning technique to the features derived from the plurality of wells to generate a failure prediction model, the applying of the random peek semi-supervised machine learning technique including;
(a) splitting the well data into at least two data clusters such that a first cluster of well data is larger than a second cluster of well data, wherein the first cluster of well data is assumed to correspond to wells that do not contain one or more well failures and the second cluster of well data is assumed to correspond to wells that contain one or more well failures, and (b) adding well data from the first cluster to a training data set used to generate the failure prediction model;
determining whether one or more of the plurality of wells is predicted to fail within a subsequent time period based on the failure prediction model; and
outputting an alert indicative of impending failure of one or more of the plurality of wells.
1 Assignment
0 Petitions
Accused Products
Abstract
A computer-implemented reservoir prediction system, method, and software are provided for failure prediction for artificial lift well systems, such as sucker rod pump systems. The method includes providing well data from a production well. Attributes are extracted from the well data. Data mining is applied to the attributes to determine whether the production well is predicted to fail within a given time period. An alert is output indicative of impending production well failures.
-
Citations
22 Claims
-
1. A computer-implemented method for failure prediction for artificial lift well systems, the method comprising:
-
obtaining well data representing a plurality of measured attributes from a plurality of wells over a time period; generating features from the well data based on the measured attributes; applying a random peek semi-supervised machine learning technique to the features derived from the plurality of wells to generate a failure prediction model, the applying of the random peek semi-supervised machine learning technique including;
(a) splitting the well data into at least two data clusters such that a first cluster of well data is larger than a second cluster of well data, wherein the first cluster of well data is assumed to correspond to wells that do not contain one or more well failures and the second cluster of well data is assumed to correspond to wells that contain one or more well failures, and (b) adding well data from the first cluster to a training data set used to generate the failure prediction model;determining whether one or more of the plurality of wells is predicted to fail within a subsequent time period based on the failure prediction model; and outputting an alert indicative of impending failure of one or more of the plurality of wells. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. A system for failure prediction for artificial lift well systems, the system comprising:
-
a database configured to store well data representing a plurality of measured attributes from a plurality of wells over a time period; a computer processor configured to receive the stored data from the database, and to execute software responsive to the stored data; and a software program executable on the computer processor to implement a method, the method comprising; extracting features from the well data from the plurality of wells based on the attributes; and generating a failure prediction model using a random peek semi-supervised machine learning technique, the random peek semi-supervised machine learning technique including;
(a) splitting the well data into at least two data clusters such that a first cluster of well data is larger than a second cluster of well data, wherein the first cluster of well data is assumed to correspond to wells that do not contain one or more well failures and the second cluster of well data is assumed to correspond to wells that contain one or more well failures, and (b) adding well data from the first cluster to a training data set used to generate the failure prediction model; anddetermining whether one or more of the plurality of wells is predicted to fail within a subsequent time period based on the extracted features. - View Dependent Claims (17, 18, 19, 20, 21)
-
-
22. A computer program product, comprising a non-transitory computer usable medium having a computer readable program code embodied therein, the computer readable program code configured to be executed on a computer system to implement a method for failure prediction for artificial lift well systems, the method comprising:
-
extracting features from well data representing a plurality of measured attributes from a plurality of wells over a time period; and applying a random peek semi-supervised machine learning technique to the features to create a failure prediction model based on the features from the plurality of wells, the random peek semi-supervised machine learning technique including;
(a) splitting the well data into at least two data clusters such that a first cluster of well data is larger than a second cluster of well data, wherein the first cluster of well data is assumed to correspond to wells that do not contain one or more well failures and the second cluster of well data is assumed to correspond to wells that contain one or more well failures, and (b) adding well data from the first cluster to a training data set used to generate the failure prediction model; anddetermining whether one or more of the plurality of wells is predicted to fail within a given time period based on the failure prediction model.
-
Specification