System and method for determining well characteristics

Abstract

Methods and systems for predicting total organic carbon (TOC) throughout a well. The method includes deploying a logging tool in the well and obtaining logging data from the well using the logging tool, where the logging data has one or more data values at each depth in a set of depths. The method further includes determining, with a first machine learning model, elemental data for the well at each depth in the set of depths based on the logging data and determining, with a second machine learning model, TOC at each depth in the set of depths based on the logging data and elemental data. The method further includes determining a hydrocarbon content of the well based on the determined TOC and executing a well operation plan based on the hydrocarbon content.

What is claimed is: 
     
         1 . A method for predicting total organic carbon (TOC) throughout a well, the method comprising:
 deploying a logging tool in the well;   obtaining logging data from the well using the logging tool, wherein the logging data comprises one or more data values at each depth in a set of depths;   determining, with a first machine learning model, elemental data for the well at each depth in the set of depths based on the logging data;   determining, with a second machine learning model, TOC at each depth in the set of depths based on the logging data and elemental data;   determining a hydrocarbon content of the well based on the determined TOC; and   executing a well operation plan based on the hydrocarbon content.   
     
     
         2 . The method according to  claim 1 , wherein the first machine learning model comprises a neural network. 
     
     
         3 . The method according to  claim 1 , wherein the logging tool is a wireline tool. 
     
     
         4 . The method according to  claim 1 , further comprising training the first and second machine learning models, wherein training the first and second machine learning models comprises:
 obtaining modelling data comprising modelling logging data, modelling elemental data, and modelling TOC data from at least one of a sampled interval of the well and one or more offset wells;   identifying contaminated samples in the modelling TOC data using an outlier detection or correlation method in view of the modelling elemental data;   removing identified contaminated samples from the modelling data;   forming a training dataset comprising depth aligned examples from the modelling data; and   jointly training the first and second machine learning models, wherein:
 the first machine learning model receives, as input, the modelling logging data of the training dataset and returns, as output, predicted elemental data, 
 the second machine learning model receives, as input, the modelling logging data and of the training dataset and the predicted elemental data and returns, as output, predicted TOC data, and 
 the joint training is guided by a first comparison of the modelling elemental data of the training dataset and the predicted elemental data and a second comparison of the modelling TOC data of the training dataset and the predicted TOC data. 
   
     
     
         5 . The method according to  claim 4 , wherein the modelling elemental data are obtained from one or more of inductively coupled plasma-mass spectrometry (ICP-MS) and x-ray fluorescence (XRF). 
     
     
         6 . The method according to  claim 4 , wherein the modelling TOC data are obtained using a pyrolysis technique on one or more core samples collected from one or more of the well and offset wells within the same geological setting. 
     
     
         7 . The method according to  claim 4 , wherein training the first and second machine learning models further comprises:
 forming a validation dataset comprising depth aligned examples from the modelling data; and   validating the first and second machine learning models using the validation dataset.   
     
     
         8 . A non-transitory computer-readable memory comprising computer-executable instructions stored thereon that, when executed on a processor, cause the processor to perform steps comprising:
 obtaining logging data from a well using a logging tool deployed in the well, wherein the logging data comprises one or more data values at each depth in a set of depths;   determining, with a first machine learning model, elemental data for the well at each depth in the set of depths based on the logging data;   determining, with a second machine learning model, total organic carbon (TOC) at each depth in the set of depths based on the logging data and elemental data; and   determining a hydrocarbon content of the well based on the determined TOC.   
     
     
         9 . The non-transitory computer-readable memory according to  claim 8 , wherein the first machine learning model comprises a neural network. 
     
     
         10 . The non-transitory computer-readable memory according to  claim 8 , the steps further comprising training the first and second machine learning models, wherein training the first and second machine learning models comprises:
 obtaining modelling data comprising modelling logging data, modelling elemental data, and modelling TOC data from at least one of a sampled interval of the well and one or more offset wells;   identifying contaminated samples in the modelling TOC data using an outlier detection or correlation method in view of the modelling elemental data;   removing identified contaminated samples from the modelling data;   forming a training dataset comprising depth aligned examples from the modelling data; and   jointly training the first and second machine learning models, wherein:
 the first machine learning model receives, as input, the modelling logging data of the training dataset and returns, as output, predicted elemental data, 
 the second machine learning model receives, as input, the modelling logging data and of the training dataset and the predicted elemental data and returns, as output, predicted TOC data, and 
 the joint training is guided by a first comparison of the modelling elemental data of the training dataset and the predicted elemental data and a second comparison of the modelling TOC data of the training dataset and the predicted TOC data. 
   
     
     
         11 . The non-transitory computer-readable memory according to  claim 10 , wherein the modelling elemental data are obtained from one or more of inductively coupled plasma-mass spectrometry (ICP-MS) and x-ray fluorescence (XRF). 
     
     
         12 . The non-transitory computer-readable memory according to  claim 10 , wherein the modelling TOC data are obtained using a pyrolysis technique on one or more core samples collected from one or more of the well and offset wells within the same geological setting. 
     
     
         13 . The non-transitory computer-readable memory according to  claim 10 , wherein training the first and second machine learning models further comprises:
 forming a validation dataset comprising depth aligned examples from the modelling data; and   validating the first and second machine learning models using the validation dataset.   
     
     
         14 . A system, comprising:
 a logging tool configured to obtain logging data from a well;   a computer processor; and   a non-transitory computer readable medium storing instructions that when executed by the computer processor cause the processor to perform operations comprising:
 obtaining logging data from the well using the logging tool, wherein the logging data comprises one or more data values at each depth in a set of depths; 
 determining, with a first machine learning model, elemental data for the well at each depth in the set of depths based on the logging data; 
 determining, with a second machine learning model, total organic carbon (TOC) at each depth in the set of depths based on the logging data and elemental data; and 
 determining a hydrocarbon content of the well based on the determined TOC. 
   
     
     
         15 . The system according to  claim 14 , wherein the first machine learning model comprises a neural network. 
     
     
         16 . The system according to  claim 14 , wherein the logging tool is a wireline tool. 
     
     
         17 . The system according to  claim 14 , wherein the operations further comprise training the first and second machine learning models, wherein training the first and second machine learning models comprises:
 obtaining modelling data comprising modelling logging data, modelling elemental data, and modelling TOC data from at least one of a sampled interval of the well and one or more offset wells;   identifying contaminated samples in the modelling TOC data using an outlier detection or correlation method in view of the modelling elemental data;   removing identified contaminated samples from the modelling data;   forming a training dataset comprising depth aligned examples from the modelling data; and   jointly training the first and second machine learning models, wherein:
 the first machine learning model receives, as input, the modelling logging data of the training dataset and returns, as output, predicted elemental data, 
 the second machine learning model receives, as input, the modelling logging data and of the training dataset and the predicted elemental data and returns, as output, predicted TOC data, and 
 the joint training is guided by a first comparison of the modelling elemental data of the training dataset and the predicted elemental data and a second comparison of the modelling TOC data of the training dataset and the predicted TOC data. 
   
     
     
         18 . The system according to  claim 17 , wherein the modelling elemental data are obtained from one or more of inductively coupled plasma-mass spectrometry (ICP-MS) and x-ray fluorescence (XRF). 
     
     
         19 . The system according to  claim 17 , wherein the modelling TOC data are obtained a pyrolysis technique on one or more core samples collected from one or more of the well and offset wells within the same geological setting. 
     
     
         20 . The system according to  claim 17 , wherein training the first and second machine learning models further comprises:
 forming a validation dataset comprising depth aligned examples from the modelling data; and   validating the first and second machine learning models using the validation dataset.