Neural network training data selection using memory reduced cluster analysis for field model development
First Claim
1. A method for producing a training data set from a set of multidimensional geophysical input data samples for training a model to predict target data, comprising:
- dividing a set of geophysical input data samples into a plurality of first subsets of input data samples, dividing each of the first subsets into a plurality of first clusters, generating a first set of prototypes each representing one of the first clusters, and dividing the first set of prototypes into a plurality of second clusters.
1 Assignment
0 Petitions
Accused Products
Abstract
A system and method for selecting a training data set from a set of multidimensional geophysical input data samples for training a model to predict target data. The input data may be data sets produced by a pulsed neutron logging tool at multiple depth points in a cases well. Target data may be responses of an open hole logging tool. The input data is divided into clusters. Actual target data from the training well is linked to the clusters. The linked clusters are analyzed for variance, etc. and fuzzy inference is used to select a portion of each cluster to include in a training set. The reduced set is used to train a model, such as an artificial neural network. The trained model may then be used to produce synthetic open hole logs in response to inputs of cased hole log data.
-
Citations
59 Claims
-
1. A method for producing a training data set from a set of multidimensional geophysical input data samples for training a model to predict target data, comprising:
-
dividing a set of geophysical input data samples into a plurality of first subsets of input data samples, dividing each of the first subsets into a plurality of first clusters, generating a first set of prototypes each representing one of the first clusters, and dividing the first set of prototypes into a plurality of second clusters. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. A method for producing a training data set from a set of multidimensional geophysical input data samples for training a model to predict target data, comprising:
-
dividing multidimensional geophysical input data samples into a set of clusters, linking each multidimensional geophysical input data sample with corresponding target data, and performing an analysis of the input samples and target data in each cluster. - View Dependent Claims (17, 18, 19, 20, 21, 22, 23, 24, 25, 26)
-
-
27. A method for predicting open borehole logging measurements from actual cased borehole logging measurements, comprising:
-
collecting open hole logging measurements in a borehole, collecting cased borehole logging measurements in the borehole, dividing the cased borehole logging measurements into a set of clusters, linking each cased borehole logging measurement with corresponding open hole logging measurements, performing an analysis of the cased borehole logging measurements and corresponding open hole logging measurements for each cluster, selecting a percentage of the cased borehole logging measurements and corresponding open hole logging measurements from each cluster based on results of the analyses, training a predictive model with the selected measurements, and using the trained predictive model to predict open hole logging measurements in response to cased borehole logging measurements. - View Dependent Claims (28, 29, 30, 31, 32, 33, 34, 35)
-
-
36. A method for predicting open borehole geophysical measurements from actual cased borehole geophysical measurements, comprising:
-
collecting open hole geophysical measurements in a borehole, collecting cased borehole geophysical measurements in the borehole, selecting a percentage of the cased borehole measurements and corresponding open hole measurements as a training data set, training a predictive model with the selected measurements, and using the trained predictive model to predict open hole geophysical measurements in response to cased borehole geophysical measurements.
-
-
37. A method for predicting cased borehole geophysical measurements from actual open borehole geophysical measurements, comprising:
-
collecting open hole geophysical measurements in a borehole, collecting cased borehole geophysical measurements in the borehole, selecting a percentage of the open hole measurements and corresponding cased borehole measurements as a training data set, training a predictive model with the selected measurements, and using the trained predictive model to predict cased hole geophysical measurements in response to open borehole geophysical measurements.
-
-
38. A method for producing a synthetic log of at least one geophysical parameter for a well, comprising:
-
collecting a first log of a plurality of geophysical parameters, including the at least one geophysical parameter, in a first well, the log comprising a plurality of multidimensional data samples, dividing the data samples into a set of clusters based on the geophysical parameters other than the at least one geophysical parameter, selecting data from each cluster, training a predictive model with the selected data, collecting a second log of the plurality of geophysical parameters, excluding the at least one geophysical parameter, in a second well, and inputting the second log to the predictive model to produce a synthetic log of the at least one geophysical parameter for the second well. - View Dependent Claims (39, 40, 41, 42)
-
-
43. A method for producing a synthetic value of at least one geophysical parameter for a well, comprising:
-
collecting a first data sample set of a plurality of geophysical parameters, including the at least one geophysical parameter, relating to a first well, dividing the first data sample set into a set of clusters based on the geophysical parameters other than the at least one geophysical parameter, selecting data from each cluster, training a predictive model with the selected data, collecting a second data sample set of the plurality of geophysical parameters, excluding the at least one geophysical parameter, relating to a second well, and inputting the second data sample set to the predictive model to produce a synthetic value of the at least one geophysical parameter for the second well. - View Dependent Claims (44, 45, 46, 47, 48)
-
-
49. A method of operating a hydrocarbon bearing field, comprising:
-
drilling a plurality of wells in the hydrocarbon bearing field, performing open hole logging in a subset of the wells, performing cased hole logging in substantially all of the wells including the subset of wells, using open hole logging data and cased hole logging data from the subset of wells to train a predictive model to produce synthetic open hole data in response to inputs of cased hole data, and using the trained predictive model and cased hole data from the wells to produce synthetic open hole data. - View Dependent Claims (50, 51, 52)
-
-
53. Apparatus for producing synthetic values of at least one geophysical parameter for a well, comprising a predictive model trained by:
-
collecting a first data sample set of a plurality of geophysical parameters, including the at least one geophysical parameter, relating to a first well, dividing the first data sample set into a set of clusters based on the geophysical parameters other than the at least one geophysical parameter, selecting data from each cluster, and training the predictive model to produce a synthetic value of the at least one geophysical parameter in response to inputs of the plurality of geophysical parameters, excluding the at least one geophysical parameter. - View Dependent Claims (54, 55)
-
-
56. A method for producing synthetic geophysical measurements in a well having at least one depth interval in which one or more actual measurements cannot be accurately taken, comprising:
-
collecting at least one log of a plurality of geophysical measurements in a borehole, the at least one log having missing or defective measurements of at least one parameter in at least one depth interval, selecting a training data set comprising at least a portion of the plurality of geophysical measurements from depth intervals other than the at least one depth interval, training a predictive model with the training data set, and using the trained predictive model to produce synthetic values of the missing or defective measurements of the at least one parameter in response to inputs comprising at least a portion of the geophysical measurements taken in the at least one depth interval. - View Dependent Claims (57, 58, 59)
-
Specification