Predicting gas lift equipment failure with deep learning techniques

Abstract

A method and a system for predicting gas lift equipment failure are disclosed. The method includes obtaining data from a plurality of sources, the plurality of sources including sensor readings, maintenance records, operational parameters, and production targets and determining a plurality of initial equipment failure probabilities using a plurality of machine learning models. Further, the method includes determining an ensemble prediction using an ensemble model trained on the plurality of initial equipment failure probabilities and performing, in response to the ensemble prediction, a maintenance operation of the gas lift equipment.

What is claimed is: 
     
         1 . A method for predicting gas lift equipment failure, comprising:
 obtaining, using a computer processor, data from a plurality of sources, the plurality of sources including sensor readings, maintenance records, operational parameters, and production targets;   determining, using the computer processor, a plurality of initial equipment failure probabilities using a plurality of machine learning models;   determining, using the computer processor, an ensemble prediction using an ensemble model trained on the plurality of initial equipment failure probabilities; and   performing, in response to the ensemble prediction, a maintenance operation of the gas lift equipment.   
     
     
         2 . The method of  claim 1 , wherein the plurality of machine learning models include Support Vector Machine, Random Forest, and Deep Learning. 
     
     
         3 . The method of  claim 2 , wherein determining an initial equipment failure probability using SVM comprises:
 transforming, using the computer processor, input data to a feature space, the feature space being of higher dimension than an original input space;   parameterizing, using the computer processor, a hyperplane in the feature space, the hyperplane being the initial equipment failure probability of the SVM for a given input.   
     
     
         4 . The method of  claim 3 , wherein a decision boundary separating healthy and damaged lift gas equipment is determined using the SVM. 
     
     
         5 . The method of  claim 3 , wherein the transformation of the input data is based on a kernel function. 
     
     
         6 . The method of  claim 3 , wherein the SVM includes slack terms and regularization terms. 
     
     
         7 . The method of  claim 1 , wherein determining an initial equipment failure probability using Random Forest model comprises:
 determining hyperparameters of the Random Forest model including a maximum depth of decision trees and a minimum number of samples required to split a node;   determining, using the computer processor, the initial equipment failure probability by averaging predictions from all the decision trees.   
     
     
         8 . The method of  claim 1 , wherein determining an initial equipment failure probability using Deep Learning model comprises:
 generating, using the computer processor, relationships between the data and a target initial equipment failure probability by adjusting weights and biases of neurons in a Deep Learning network;   determining, using the computer processor, the initial equipment failure probability based on the generated relationships.   
     
     
         9 . The method of  claim 1 , wherein the maintenance operation comprises replacing gas lift equipment components. 
     
     
         10 . The method of  claim 1 , wherein the maintenance operation comprises adjusting operating parameters of the gas lift equipment to prevent failure. 
     
     
         11 . A non-transitory computer readable medium storing instructions executable by a computer processor, the instructions comprising functionality for:
 obtaining data from a plurality of sources, the plurality of sources including sensor readings, maintenance records, operational parameters, and production targets;   determining a plurality of initial equipment failure probabilities using a plurality of machine learning models;   determining an ensemble prediction using an ensemble model trained on the plurality of initial equipment failure probabilities; and   performing, in response to the ensemble prediction, a maintenance operation of a gas lift equipment.   
     
     
         12 . The non-transitory computer readable medium of  claim 11 , wherein the plurality of machine learning models include Support Vector Machine, Random Forest, and Deep Learning. 
     
     
         13 . The non-transitory computer readable medium of  claim 12 , wherein determining an initial equipment failure probability using SVM comprises:
 transforming input data to a feature space, the feature space being of higher dimension than an original input space;   parameterizing a hyperplane in the feature space, the hyperplane being the initial equipment failure probability of the SVM for a given input.   
     
     
         14 . The non-transitory computer readable medium of  claim 12 , wherein determining an initial equipment failure probability using Random Forest model comprises:
 determining hyperparameters of the Random Forest model including a maximum depth of decision trees and a minimum number of samples required to split a node;   determining the initial equipment failure probability by averaging predictions from all the decision trees.   
     
     
         15 . The non-transitory computer readable medium of  claim 12 , wherein determining an initial equipment failure probability using Deep Learning model comprises:
 generating relationships between the data and a target initial equipment failure probability by adjusting weights and biases of neurons in a Deep Learning network;   determining the initial equipment failure probability based on the generated relationships.   
     
     
         16 . The non-transitory computer readable medium of  claim 11 , wherein the maintenance operation comprises replacing gas lift equipment components. 
     
     
         17 . The non-transitory computer readable medium of  claim 11 , wherein the maintenance operation comprises refurbishing gas lift equipment components. 
     
     
         18 . A system comprising:
 a well logging system; and   a equipment failure simulator comprising a computer processor, wherein the equipment failure simulator is coupled to the well logging, the equipment failure simulator comprising functionality for:
 obtaining data from a plurality of sources, the plurality of sources including sensor readings, maintenance records, operational parameters, and production targets; 
 determining a plurality of initial equipment failure probabilities using a plurality of machine learning models; 
 determining an ensemble prediction using an ensemble model trained on the plurality of initial equipment failure probabilities; and 
 performing, in response to the ensemble prediction, a maintenance operation of a gas lift equipment. 
   
     
     
         19 . The system of  claim 18 , wherein the maintenance operation comprises replacing gas lift equipment components. 
     
     
         20 . The system of  claim 18 , wherein the maintenance operation comprises refurbishing gas lift equipment components.