Machine learning workflow for predicting hydraulic fracture initiation
Systems and methods include a computer-implemented method for predicting hydraulic fracture initiation. A fracking operations dataset is prepared using historical field information for fracking wells. A set of hyper-parameters is tuned for use in a machine learning algorithm configured to predict fracture initiation for new fracturing wells. The dataset is divided into training and test datasets. A regression algorithm is applied to train the training dataset and to validate with the test dataset. A target variable of a breakdown pressure for a new hydraulic fracturing treatment is determined. A prediction dataset is updated using at least the target variable. The training dataset is trained using a classifier of the machine learning algorithm. A prediction is made using the prediction dataset whether the new hydraulic fracturing treatment can be initiated or not. The breakdown pressure is incrementally adjusted, and the method is repeated until successful hydraulic fracture initiation is predicted.
1 . A method, comprising:
preparing a fracking operations dataset based on historical information for a group of fracking wells from fields, the fracking operations dataset comprising a plurality of depth-specific fracturing treatments, each depth-specific fracturing treatment comprising a stress regime, a rock type, a porosity, a number of perforations, a perforation interval, a perforation diameter, a breakdown pressure, and a fracture initiation classification, wherein the stress regime represents one of a normal fault or a strike-slip, and the rock type represents one of shale, sandstone, or carbonate;
tuning a set of hyper-parameters of a machine learning algorithm configured to predict fracture initiation for new fracturing wells;
dividing the fracking operations dataset into a training dataset and a test dataset, wherein the training dataset is configured to train a fracture initiation prediction model;
applying a regression algorithm to train with the training dataset and to validate with the test dataset;
for each new depth-specific hydraulic fracturing treatment of a plurality of new depth-specific hydraulic fracturing treatments:
determining, using the fracture initiation prediction model and the training dataset, a breakdown pressure for the new depth-specific hydraulic fracturing treatment;
updating a prediction dataset based on at least the breakdown pressure;
training the training dataset using a classifier of the machine learning algorithm;
predicting, using the prediction dataset, whether the new depth-specific hydraulic fracturing treatment can be initiated or not; and
incrementally adjusting the breakdown pressure and repeating the updating, training, and predicting until successful hydraulic fracture initiation at the new depth-specific hydraulic fracturing treatment is predicted;
selecting a hydraulic fracturing design by selecting a casing, a treatment tubing, a wellhead, and a pumping schedule based on the breakdown pressures for the plurality of new depth-specific hydraulic fracturing treatments; and
injecting a fracking fluid containing proppants into a wellbore based on the pumping schedule such that the fracking fluid flows through perforations to create fractures in a reservoir at each new depth-specific fracturing treatment of the plurality of new depth-specific hydraulic fracturing treatments.
2 . The method of claim 1 , wherein the set of hyper-parameters includes at least learning_rate, gamma, max_depth, and max_leaves.
3 . The method of claim 1 , wherein the fracking operations dataset includes, for each fracking well: well survey data including landing depth of true vertical depth, azimuth, and deviation, in-situ stresses and orientation, and rock mechanical properties.
4 . The method of claim 1 , wherein the machine learning algorithm is an XGBoost algorithm.
5 . The method of claim 1 , wherein incrementally adjusting the breakdown pressure includes increasing the breakdown pressure in increments of 200 pounds per square inch (psi).
6 . The method of claim 1 , wherein incrementally adjusting the breakdown pressure includes increasing the breakdown pressure in increments of 400 pounds per square inch (psi).
7 . The method of claim 1 , wherein predicting whether the new depth-specific hydraulic fracturing treatment can be initiated or not includes using a prediction function based on previous predictions.
8 . The method of claim 7 , wherein the prediction function includes learning rate multipliers applied to previous predictions.
9 . The method of claim 1 , wherein the plurality of depth-specific fracturing treatments comprises 100 depth-specific fracturing treatments.
10 . The method of claim 9 , wherein the plurality of new depth-specific fracturing treatments comprises 6 depth-specific fracturing treatments.
11 . A method for fracturing a reservoir, the method comprising:
collecting hydraulic fracturing field data for a plurality of depth-specific fracturing treatments, the hydraulic fracturing field data representing historical information for a group of fracking wells from fields, and comprises, for each depth-specific fracturing treatment of the plurality of depth-specific fracturing treatments, a stress regime, a rock type, a porosity, a number of perforations, a perforation interval, a perforation diameter, a breakdown pressure, and a fracture initiation classification, wherein the stress regime represents one of a normal fault or a strike-slip, and the rock type represents one of shale, sandstone, or carbonate;
training, by a computer, a machine learning model for predicting breakdown pressures and fracture initiation classifications for new depth-specific fracturing treatments using a regression algorithm and most of the hydraulic fracturing field data, wherein the machine learning model can handle missing values;
predicting, by the computer, breakdown pressures and fracture initiation classifications for new depth-specific fracturing treatments using the trained machine learning model;
selecting a hydraulic fracturing design of a wellbore by selecting a casing, a treatment tubing, a wellhead, and a pumping schedule based on the predicted breakdown pressures and fracture initiation classifications for the new depth-specific fracturing treatments; and
injecting a fracking fluid containing proppants into the wellbore based on the pumping schedule such that the fracking fluid flows through perforations to create fractures in the reservoir at the new depth-specific fracturing treatments.
12 . The method of claim 11 , comprising testing the trained machine learning model using the rest of the hydraulic fracturing field data, and then using the trained and tested machine learning model to predict the breakdown pressures and fracture initiation classifications for the new depth-specific fracturing treatments.
13 . The method of claim 11 , comprising in response to determining that a predicted fracture initiation classification for at least one of the new depth-specific fracturing treatments represents no fracture initiation, increasing the breakdown pressure, and updating the fracture initiation classification prediction.
14 . The method of claim 11 , wherein at least one of the predicted fracture initiation classifications for the new depth-specific fracturing treatments represents no fracture initiation, and at least one of the predicted fracture initiation classifications for the new depth-specific fracturing treatments represents fracture initiation.
15 . The method of claim 14 , comprising, for the at least one of the predicted fracture initiation classifications representing no fracture initiation:
increasing the breakdown pressure by at least 200 pounds per square inch; and
updating the fracture initiation classification prediction.
16 . The method of claim 11 , wherein the hydraulic fracturing field data for each depth-specific fracturing treatment comprises a tensile strength, an ultimate compressive strength, a landing depth, a well azimuth, a deviation, a vertical stress, a maximum horizontal stress, a minimum horizontal stress, and a reservoir pressure.
17 . The method of claim 11 , wherein the predicted breakdown pressures represent an internal casing pressure when fractures initiate from a perforation tunnel.
18 . The method of claim 11 , wherein the plurality of depth-specific fracturing treatments comprises 100 depth-specific fracturing treatments, and the new depth-specific fracturing treatments comprises 6 depth-specific fracturing treatments.
19 . The method of claim 11 , comprising tuning a set of hyper-parameters of the machine learning model, the set of hyper-parameters comprising a learning_rate, a gamma, a max_depth, and max_leaves.
20 . A method comprising:
receiving, at a computer, a fracking operations dataset comprising historical information for a group of fracking wells from fields, the fracking operations dataset representing a plurality of depth-specific fracturing treatments and comprising, for each depth-specific fracturing treatment, a stress regime, a rock type, a porosity, a number of perforations, a perforation interval, a perforation diameter, a breakdown pressure, and a fracture initiation classification, wherein the stress regime represents one of a normal fault or a strike-slip, and the rock type represents one of shale, sandstone, or carbonate;
for each depth-specific fracturing treatment of a plurality of new depth-specific fracturing treatments:
predicting, by the computer using a machine learning algorithm trained on the fracking operations dataset, a breakdown pressure at the respective depth-specific fracturing treatment;
predicting, by the computer using the machine learning algorithm trained on the fracking operations dataset, a fracture initiation classification representing whether hydraulic fracture can be initiated at the respective depth-specific fracturing treatment; and
in response to the fracture initiation classification being indicative of an unsuccessful fracture initiation at the respective depth-specific fracturing treatment, updating the breakdown pressure prediction and the fracture initiation classification prediction until the fracture initiation classification is indicative of a successful fracture initiation;
selecting a hydraulic fracturing design of a wellbore by selecting a casing, a treatment tubing, a wellhead, and a pumping schedule based on the predicted breakdown pressures and successful fracture initiation classifications for the plurality of new depth-specific fracturing treatments; and
injecting a fracking fluid into the wellbore to initiate fractures at the plurality of new depth-specific fracturing treatments.