Systems and methods for generating an interference prediction for a target well are disclosed herein. A computing system generates a plurality of interference metrics for a plurality of interference events. For each well, the computing system generates a graph based representation of the well and its neighboring wells. The computing system generates a predictive model using a graph-based model by generating a training data set and learning, by the graph-based model, an interference value for each interference event based on the training data set. The computing system receives, from a client device, a request to generate an interference prediction for a target well. The computing system generates, via the predictive model, an interference metric based on the one or more metrics associated with the target well.

Apparatus and methods for autonomously shifting a downhole sliding sleeve. A shift tool includes a shifter arm, an artificial neural network, and a control circuit. The artificial neural network is trained to identify engagement of the shifter arm with a shifting feature of a sliding sleeve. The control circuit is configured to extend the shifter arm at a first pressure for seeking engagement with the shifting feature of the sliding sleeve, and responsive to the artificial neural network recognizing engagement of the shifter arm with the shifting feature of the sliding sleeve, extend the shifter arm at a second pressure for shifting the sliding sleeve.

Drilling operations may be monitored to detect and quantify potential drilling dysfunctions. Using a Bayesian network, potential improvements to drilling operation may be made depending upon the type of dysfunction detected. Suggestions for improved drilling performance may comprise increasing, decreasing, or maintaining one or both of RPM and weight on bit. Suggestions may be presented to an operator as a cone having an apex at the current RPM and weight on bit drilling parameters, with suggestions for modifications to one or both of the RPM and weight on bit corresponding to a cone extending from that apex.

The present disclosure relates to a prediction method for constant production decline of a water-producing gas well in a highly heterogeneous reservoir. The prediction method mainly includes: collecting related data of a target water-producing gas well, fitting to obtain a water-drive constant and a water invasion constant, fitting dynamic reserves by adopting a Blasingame plotting method, conducting fitting by adopting a dual-medium model to obtain an elastic storativity ratio and an interporosity flow coefficient, calculating a reservoir heterogeneity coefficient, obtaining a flowing bottomhole pressure at the later stage of stable production, calculating formation pressure of a new day through quantitative production of the target water-producing gas well with 1 day as an iteration stride, performing iteration until the formation pressure is less than or equal to the formation pressure at the end of stable production, and drawing a prediction curve about constant production decline of the target water-producing gas well.

A method for performing wellbore correlation across multiple wellbores includes predicting a depth alignment across the wellbores based on a geological feature of the wellbores. Predicting a depth alignment includes selecting a reference wellbore, defining a control point in a reference signal of a reference well log for the reference wellbore, and generating an input tile from the reference signal, the control points, and a number of non-reference well logs corresponding to non-reference wellbores. The well logs include changes in a geological feature over a depth of a wellbore. The input tile is input into a machine-learning model to output a corresponding control point for each non-reference well log. The corresponding control point corresponds to the control point of the reference log. Based on the corresponding control points output from the machine-learning model, the non-reference well logs are aligned with the reference well log to correlate the multiple wellbores.

The systems and methods generally apply a machine learning algorithm that is configured to automatically reconcile and update field costs. The algorithm may also predict when field cost data has not captured all or substantially all of the costs based on historical data. The systems and methods may either flag the missing cost or automatically populate at least a portion or up to all of any missing field costs. Thus, machine learning and data analytic techniques may be implemented to reconcile field costs by using, for example, invoice and financial information.

A method can include receiving input for a drilling operation that utilizes a bottom hole assembly and drilling fluid; generating a set of offset drilling operations using historical feature data, where the historical feature data are processed by computing feature distances; performing an assessment of the offset drilling operations as characterized by at least feature distance-based similarity between the drilling operation and the offset drilling operations; and outputting at least one recommendation for selection of one or more of a component of the bottom hole assembly and the drilling fluid based on the assessment.

A method for a hyperparameter optimization for an automated ensemble machine learning model includes: generating an initial population of a plurality of machine learning (ML) models with a plurality of randomly chosen hyperparameters; calculating a loss function for each of the plurality of machine learning models; creating a new population of ML models and generating a base learner model using the hyperparameters of the best model. The method for creating the new population include the steps of: (a) selecting multiple best models with least errors as parents from a previous generation; (b) creating an offspring of the new population of ML models with a crossover probability and a mutation probability; and (c) repeating the steps (a) and (b) until a number of generations is reached and reporting the hyperparameters of the best model.

Systems and methods presented herein are configured to train a neural network model using a first set of photographs, wherein each photograph of the first set of photographs depicts a first set of objects and include one or more annotations relating to each object of the first set of objects; to automatically create mask images corresponding to a second set of objects depicted by a second set of photographs; to enable manual fine tuning of the mask images corresponding to the second set of objects depicted by the second set of photographs; to re-train the neural network model using the second set of photographs, wherein the re-training is based at least in part on the manual fine tuning of the mask images corresponding to the second set of objects depicted by the second set of photographs; and to identify one or more individual objects in a third set of photographs using the re-trained neural network model.

A method and apparatus for evaluating light-weight cement (LWC) bond conditions in production well in presence or absence of the tubing. The apparatus includes a tool string that can be lowered into the casing or the tubing. The tool string includes a segmented transducer ring matrix to excite the surrounding medium resulting in vibrations. The tool string further includes an impedance measurement circuit that can determine natural resonance frequencies of the structural components. The impedance measurement circuit can determine certain non-harmonic resonance mode shapes for mechanical impedance measurements that are sensitive to the LWC bond conditions. A machine learning model can be used for segmented impedance measurements to correct effects from the tubing eccentricity inside the casing. Lab calibration generates bond index mapping table and the field logging impedance measurement data can be processed into the bond indexes to further evaluate LWC bond conditions.