Predicting well intervention operation success opportunity for well intervention plan

Abstract

Aspects of the present disclosure provide a predicted intervention success probability of a well intervention plan. A well intervention planning system includes one or more memories storing computer executable code. The system includes a user interface or data connection configured to receive inputs including at least one or more well intervention operations and one or more well conditions. The system includes one or more processors configured to execute computer executable code. The one or more processors are configured to input the one or more well intervention operations and the one or more well conditions to the predictive model; predicting a well intervention operation success probability of the one or more well intervention operations based at least in part on the one or more well conditions using the predictive model; and outputting a predicted well intervention operation success probability for the one or more well intervention operations.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to well interventions, and more specifically, to operational success opportunities for well intervention plans. BACKGROUND

[0002] In the field of oil and gas exploration, efficient extraction of hydrocarbon resources is critical to maximizing production and minimizing costs. As more and more geographic locations are developed for extraction of hydrocarbons, the proportion of production from mature hydrocarbon production increases. As a result, maximizing the capacity and efficiency of existing oil wells becomes increasingly important.

[0003] Well intervention is one method of improving productivity. At some point in the life of all oil and gas wells, parts need to be maintained, repaired, or replaced. At this point, the operator can turn to an intervention specialist. Interventions are divided into two broad categories: light and heavy. During a light intervention, a technician runs a tool or sensor into a live well while controlling pressure at the surface. In a heavy intervention, the drilling crew can stop production from the formation before making major equipment changes.

[0004] Oil well service personnel typically use slickline, wireline, or coiled tubing for light interventions. These systems allow the operator to minimize the potential for plugging the oil well. Operators also require light interventions to change or adjust downhole equipment, such as valves or pumps, or to collect downhole pressure, temperature, and flow data. Heavy interventions, also known as workovers, can require the drilling crew to remove the wellhead and other pressure barriers from the well to allow full access to the wellbore. These operations can require the rig to dismantle and reinstall the wellhead and completion equipment. Heavy interventions can be used to replace parts, such as tubing strings and pumps, that cannot be retrieved through light interventions. Some heavy interventions are performed to plug and abandon original production zones, reconfiguring the well to produce from secondary production zones; these operations are known as re-completions.

[0005] Software can be used to plan well interventions. For example, well intervention software can be used to help select wells to intervene on, to determine well intervention methods that will yield the best results, and / or to generate detailed work plans for the intervention. Using such well intervention software can increase the likelihood of success of well interventions and maximize the productivity of wells. Well intervention software can provide an interactive tool for intervention planning. Intervention plans can include operational plans, inflow simulations, and indicative prices. Operational plans can include a series of operations using intervention techniques, including primary and contingent workflows. Users can use well intervention software to develop operational plans. Operational plans can be the starting point for inflow simulations and indicative prices. One example software is Intervention Advisor (IA) provided by Schlumberger Limited of Houston, Texas.

[0006] For current well intervention software, users manually create intervention plans using interactive well intervention software. Manual planning includes user input of one or more success probabilities for the intervention plan. For example, based on the user’s own expertise and knowledge, the user can input success probabilities for each intervention technique and operation in the well intervention plan. Such manual planning can be time consuming, can require the user’s expertise, can be inaccurate and subjective, and is prone to human error.

[0007] There is a need for further improvements in well intervention planning. SUMMARY

[0008] The present disclosure provides techniques for predicting intervention operation success opportunities for well intervention planning.

[0009] Some aspects provide an automated well intervention planning system. The automated well intervention planning system can include one or more memories storing computer executable code. The automated well intervention planning system can include a user interface or data connection configured to receive input including at least one or more well intervention operations and one or more well conditions. The automated well intervention planning system can include one or more processors configured to execute the computer executable code and to: input the one or more well intervention operations and the one or more well conditions into a prediction model; use the prediction model to predict a well intervention operation success probability for the one or more well intervention operations based at least in part on the one or more well conditions; and output the predicted well intervention operation success probability for the one or more well intervention operations.

[0010] Some aspects provide a method for well intervention planning. The method for well intervention planning can include receiving input from a user or a data connection, the input comprising at least one or more well intervention operations and one or more well conditions; inputting the one or more well intervention operations and the one or more well conditions into a predictive model; using the predictive model, predicting a well intervention operation success probability for the one or more well intervention operations based at least in part on the one or more well conditions; and outputting the predicted well intervention operation success probability for the one or more well intervention operations.

[0011] Some aspects provide a computer-readable medium storing computer executable code for well intervention planning. The computer executable code can include code for receiving input from a user or a data connection, the input comprising at least one or more well intervention operations and one or more well conditions; code for inputting the one or more well intervention operations and the one or more well conditions into a predictive model; code for using the predictive model, predicting a well intervention operation success probability for the one or more well intervention operations based at least in part on the one or more well conditions; and code for outputting the predicted well intervention operation success probability for the one or more well intervention operations.

[0012] For purposes of illustration, certain features are described below in the context of certain examples. BRIEF DESCRIPTION OF DRAWINGS

[0013] The accompanying drawings illustrate several embodiments of systems in accordance with the present disclosure.

[0014] Figure 1 An example well intervention planning system is shown.

[0015] Figure 2 An example home display of a user interface of a well intervention planning system is depicted.

[0016] Figure 3 An example well selection display of a user interface of a well intervention planning system is depicted.

[0017] Figure 4 An example well intervention display of a user interface of a well intervention planning system is depicted.

[0018] Figure 5 An example plan overview display of a user interface of a well intervention planning system is depicted for market and / or environmental selection.

[0019] Figure 6 Another example plan overview display of a user interface of a well intervention planning system is depicted for well symptom selection.

[0020] Figure 7Another example plan overview display depicting intervention type selection of a user interface of a well intervention planning system.

[0021] Figure 8 An example pre-intervention display depicting well completion information of a user interface of a well intervention planning system.

[0022] Figure 9 An example pre-intervention display depicting well completion information of a user interface of a well intervention planning system.

[0023] Figure 10 An example pre-intervention display depicting trajectory information of a user interface of a well intervention planning system.

[0024] Figure 11 An example pre-intervention display depicting operational data information of a user interface of a well intervention planning system.

[0025] Figure 12 An example intervention display depicting a user interface of a well intervention planning system.

[0026] Figure 13A An example intervention display depicting a decision tree of a user interface of a well intervention planning system.

[0027] Figure 13B Another example intervention display depicting a decision tree of a user interface of a well intervention planning system.

[0028] Figure 14 An example intervention display depicting a well intervention plan report of a user interface of a well intervention planning system.

[0029] Figure 15 An example machine learning algorithm for predicting intervention operation success opportunities.

[0030] Figure 16 is a flowchart of an example method of describing predicting intervention operation success opportunities for well intervention planning.

[0031] Figure 17 is an example processing system for predicting intervention operation success opportunities for well intervention planning. DETAILED DESCRIPTION

[0032] The present disclosure provides techniques, methods, systems, apparatuses, and computer readable media for predicting intervention operation success probabilities for well intervention planning.

[0033] According to certain aspects, a well intervention can involve various intervention operations associated with various intervention techniques. In some aspects, the various intervention operations and techniques can be performed according to a specified order. The order of well intervention techniques and operations can be referred to as a well intervention main path. The intervention main path can be performed to address one or more specified well symptoms. The well on which the intervention is being performed can be associated with one or more well conditions, such as completion, well trajectory, and well operation data. One or more objectives of the well intervention can be specified. In some aspects, a contingency path can be followed when the main path is unsuccessful.

[0034] In some aspects, as described above, a software tool can be used to create a well intervention plan. In some aspects, using an interactive intervention planning tool can improve the efficiency of intervention planning, provide a standardized method for creating and selecting well intervention plans, and increase the likelihood of successful well intervention.

[0035] The various well intervention operations and related intervention techniques of a well intervention plan can be associated with a probability of success or failure. In certain current systems, a user determines the respective success probabilities of intervention operations and techniques based on their own knowledge and expertise. The user can input the estimated probabilities in a well intervention planning tool. In this case, the user of the well intervention planning tool can need to have experience in well intervention planning. Furthermore, even this estimation of success probabilities of well intervention operations and techniques by an experienced user can still be subjective, time-consuming, and prone to human error. An erroneous estimation of success probabilities can result in a poor well intervention plan, which in turn can lead to an increased risk of well intervention failure and related costs.

[0036] According to certain aspects, a prediction model can be used to predict the success probabilities of intervention plans, operations, and techniques. In some aspects, the prediction model is a statistical model. In some aspects, the prediction model is a machine learning model.

[0037] In some aspects, the prediction model is trained based on historical information of previous well interventions. The historical information can be collected from all users of the well intervention planning system. In some aspects, the machine learning model uses binary classification to predict the success probabilities of well intervention operations and techniques. In some aspects, the machine learning model uses a random forest algorithm. The risk probabilities can be used as key performance indicators in well intervention planning for selecting well intervention operations and techniques. In some aspects, using the prediction model can improve well intervention planning even for users without experience in well intervention planning. With the prediction model, the prediction of success probabilities of well intervention operations and techniques can be more objective, more accurate, and more efficient.

[0038] In some aspects, the probability of success in predicting well intervention involves data collection, preprocessing and feature labeling of collected data, training and evaluation of the predictive model based on the labeled data, deployment of the predictive model, and user interaction with the predictive model.

[0039] The following description includes embodiments of the best mode currently contemplated for practicing the described implementations. This description should not be construed as limiting, but is merely for the purpose of describing the general principles of the implementations. The scope of the described implementations should be determined with reference to the published claims.

[0040] Example well intervention planning system

[0041] Figure 1 Example well intervention planning system 100 is described.

[0042] As shown, user 105 of the well intervention planning system 100 can input information to the intervention plan generator main service 115 through user interface 110. Figure 1 As shown, and as referenced here Figures 2-11 In more detail, the information input by user 105 into intervention plan generator main service 115 via user interface 110 may include well selection information, intervention objectives, well symptoms, well conditions and / or other information.

[0043] Figure 2-1 Figure 3 shows an example display of the user interface 110 of the well intervention planning system 100, which allows user 105 to operate the system. However, it should be understood that... Figure 2-14 The example shown is merely an illustrative example of the user interface 110. In some respects, Figure 2-14 The display includes selectable options. These selectable options can be in the form of buttons. When user 105 selects (e.g., presses) one of these buttons, user interface 110 can display a new screen, a drop-down menu, or a pop-up window with information associated with the selectable option. Furthermore, although the illustrative example provides selectable options, in some cases, user 105 may directly input information into user interface 110, for example, by manually entering or typing information, instead of selecting a provided option. Additionally, while some information is described as information input in fields, it should be understood that in other examples, this information may be provided as a selectable option.

[0044] Figure 2 An exemplary homepage display 200 is depicted, illustrating a user interface 110. As shown, the homepage display 200 may include selectable options 205 for wells and selectable options 210 for interventions.

[0045] In some respects, the information input by user 105 includes the selection of wells for intervention. For example, user 105 can select well 205 selection options, providing a list of one or more wells that user 105 can choose to intervene in. Figure 3 An example well selection display 300 is depicted for user interface 110. As shown, well selection display 300 may include a list of wells. Each well may be associated with a well identifier (ID). The list of wells is selectable. For example, the list may include selectable options for well ID 1 305, selectable options for well ID 2 310, ... and so on, up to selectable options for well ID N for one or more wells.

[0046] After selecting a well, user 105 can begin inputting information about the well intervention plan for the selected well into the well intervention planning system 100. In an illustrative example, user 105 can select optional options for well intervention 210. Figure 4 An example well intervention display 400 is depicted for user interface 110. As shown, well selection display 400 may include optional options 405 for entering plan overview information, optional options 410 for entering pre-intervention information, and optional options 415 for one or more well intervention plans for the selected well.

[0047] In some respects, the program overview information includes intervention objectives. In an illustrative example, in response to user 105 selecting the optional options for program overview 405, user 105 may be provided with a list of optional market and / or environment options. Figure 5 An exemplary plan overview display 500 depicts a user interface 110. In one example, such as Figure 5 As shown, user 105 can be provided with Y selectable options for markets and / or environments, including market and / or environment 1 505, market and / or environment 2 510, etc., up to selectable options for market and / or environment Y 515. Markets can be geographic regions, countries, and / or other market options. Environments can include whether the well is an onshore or offshore well, such as deepwater, offshore, remote, platform, high-end onshore engineering, high-capacity onshore engineering, and / or other environments.

[0048] In some aspects, the program overview information includes well symptoms that will be addressed by well intervention. In an illustrative example, in response to user 105 selecting optional options in program overview 405, a list of optional options for well symptoms can be provided to user 105. Alternatively, a list of optional options for well symptoms can be provided after user 105 selects a market and / or environment. Alternatively, a list of optional options for market and / or environment can be provided after user 105 selects a well symptom. Alternatively, the list of optional options for well symptoms can be provided along with the selection options for market and / or environment. Figure 6Another example of a plan overview display 600 depicting the user interface 110. In one example, such as Figure 6 As shown, user 105 can be provided with optional options for well symptoms Z, including optional options for symptom 1 605, optional options for symptom 2 610, and so on, up to optional options for symptom Z 615. Well symptoms can include a variety of types of well symptoms. Some non-limiting examples include increased skin coverage, casing malfunction, increased water cut, decreased flow rate, increased annular pressure, significant prediction of bituminous deposits, reduced caliper logging, camera evidence of blockage, attachments malfunction, inability to run (Rih), decreased production, surface equipment deposition, high dogleg severity, changes in downhole pressure, etc.

[0049] In some aspects, the program overview information includes the intervention type. In an illustrative example, in response to user 105 selecting the options in program overview 405, a list of options for the well intervention type can be provided to user 105. Alternatively, a list of options for the well intervention type can be provided after user 105 selects the market and / or environment and / or selects the well symptom. Alternatively, a list of options for the market and / or environment and / or well symptom can be provided after user 105 selects the well intervention type. Alternatively, the list of options for the well intervention type can be provided together with the selection options for the market and / or environment and / or well symptom. Figure 7 Another example of a plan overview display 700 depicting the user interface 110. In one example, such as Figure 7 As shown, user 105 can be provided with selectable options for well intervention types L, including options for intervention type 1 705, options for intervention type 2 710, and so on, up to options for intervention type L 715. Well intervention types can include a variety of well intervention types. Some non-limiting illustrative examples include sand consolidation, sand inflow assessment, sand production prevention, sand screen repair, barrier diagnosis, bent pipe, deformed pipe, cement assessment, debris in the well, debris on top of completion attachments, etc.

[0050] In some respects, pre-intervention information includes well completion information, trajectory information, and / or operational data. Figure 8 An example pre-intervention display 800 depicts a user interface 110. In the illustrative example, user 105 can select optional options for pre-intervention information 410, and can be provided with optional options for well completion information 805, optional options for trajectory information 810, and optional options for operational data information 815.

[0051] In some aspects, completion information includes completion information about the selected well. In some aspects, completion information includes downhole equipment information, personnel information, fluid information, and / or fittings information about the selected well. In some aspects, in response to user 105 selecting selectable options for completion information 805, user 105 may be provided with a list of selectable options for inputting downhole equipment information, personnel information, fluid information, and / or fittings information about the selected well. In an illustrative example, in response to user 105 selecting selectable options for completion information 805 (or additionally, in response to selectable sub-options for fittings information), user 105 may be provided with a set of fields for inputting completion information, such as casing information, piping information, and / or perforation information. Figure 9 Another example of a pre-intervention display 900 depicts the user interface 110. In one example, such as Figure 9 As shown, user 105 may be provided with fields for inputting casing information 905, fields for inputting piping information 910, and / or fields for inputting perforation information 915. Fields for inputting casing information 905 may include fields for inputting the start and end measurement depths (MD) of the casing, fields for inputting the outer diameter (OD) of the casing, fields for inputting the casing weight, and / or fields for inputting the inner diameter (ID) of the casing. Fields for inputting piping information 910 may include fields for inputting the start and end measurement depths of the piping, fields for inputting the outer diameter of the piping, fields for inputting the weight of the piping, and / or fields for inputting the inner diameter of the piping. In some aspects, these fields may include fields for inputting casing thickness and / or piping thickness. Fields for inputting perforation information 915 may include, for each perforation, fields for inputting the perforation name, fields for inputting the top measurement depth, and fields for inputting the bottom measurement depth. Completion information may also include one or more fields for inputting other completion information (e.g., various completion information). In some respects, completion information includes information indicating whether the well is a cased well or an open hole well. In other respects, it may also display visualizations of the completion (e.g., graphic images or models).

[0052] In some respects, well trajectory information includes trajectory information about the selected well. In an illustrative example, in response to user 105 selecting an optional option for well trajectory information 810, user 105 may be provided with a set of fields for inputting well trajectory information, such as one or more measurement points, measurement depth, well angle, azimuth, and true vertical depth (TVD). Figure 10 Another example of a pre-intervention display 1000 is depicted in the user interface 110. In such... Figure 10In the illustrative example shown, user 105 may be provided with fields for inputting a measurement depth 1005, a field for inputting an angle 1010 at the measurement depth, a field for inputting an azimuth angle 1015 at the measurement depth, and a field for inputting the TVD at the measurement depth. In some aspects, well trajectory information includes information indicating whether the well is a vertical, horizontal, or deviated well. In some aspects, a visualization of the well trajectory (e.g., a graphic image or model) may also be displayed.

[0053] In some aspects, well operation information includes operational data about the selected well. In some aspects, operational data information includes pressure, temperature, and / or fluid density information about the selected well. In some aspects, in response to the user 105 selecting selectable options for well operation data information 815, a set of fields can be provided to the user 105 for inputting well operation data information. Figure 11 Another example of a pre-intervention display 1100 depicts the user interface 110. In one example, such as Figure 11 As shown, user 105 may be provided with areas for inputting pressure information 1105, areas for inputting temperature information 1110, and / or areas for inputting fluid density information 1115. In some aspects, well operation information includes fields for inputting minimum well limits, fields for inputting total depth (TD), and / or fields for inputting seafloor depth. In some aspects, the fields for inputting pressure information 1105 include the maximum pressure at the total depth. In some aspects, the fields for inputting temperature information 1110 include fields for inputting the highest temperature at the total depth, fields for inputting surface temperature, and / or fields for inputting seafloor temperature. In some aspects, the fields for inputting fluid density information 1115 include fields for inputting the fluid in the well and / or fields for inputting the fluid density. In some aspects, well operation information includes fields for indicating the presence of acidic gas and fields for indicating the concentration of acidic gas.

[0054] In some systems, the intervention plan for the selected well is manually created by user 105. For example, in some systems, in response to selecting optional options for well intervention 415, user 105 may be required to manually generate an intervention plan by inputting each intervention operation and associated intervention techniques (e.g., for assessment, perforation, development, cement assessment, ultrasonic logging tools, plugging, grinding, cleaning, etc.) as well as related interventions, symptoms, and tools. In some aspects, user 105 can add intervention operations and techniques to the decision-making process to plan well interventions. In some current systems, user 105 can further manually add the probability of success for each intervention operation.

[0055] In the example shown, in response to user 105 selecting an optional option for well intervention 415, a list of optional options can be provided to user 105.Figure 12 Example intervention display 1200 depicts user interface 110. (e.g.) Figure 12 As shown, the intervention display 1200 includes selectable options for decision tree information 1205, pricing information 1210, data visualization 1215, and / or reporting 1220. In response to user 105 selecting an option for pricing information 1210, pricing information associated with the cost of implementing the well intervention plan can be provided to user 105. The selectable option for reporting 1220 allows user 105 to select one or more options to configure reporting for the well intervention plan.

[0056] Figure 13A A well intervention display 1300A is shown for a user interface 110 used for a sample well intervention plan decision tree. As shown, in response to user 105 selecting selectable options for decision tree information 1205, a list of selectable options for creating a decision tree to plan well intervention operations can be provided to user 105. In the illustrative example, the list of selectable options may include a list of intervention techniques, including intervention technique A 1305, intervention technique B 1310, intervention technique C 1315, and so on, up to intervention technique X 1320. User 105 can use the decision tree to determine which intervention operations and techniques are included in the final well intervention plan. The probability of success of the intervention operation may be a key parameter used by the user when making a decision.

[0057] Figure 13B A well intervention display 1300B is shown as the user interface 110 of an example well intervention plan decision tree. In response to user 105 selecting one or more of selectable intervention techniques 1305-1320, an intervention action can be added to the well intervention display 1300B in the decision. Figure 13B As shown, based on the selected intervention technique, intervention 1 1325, intervention 2 1330, and so on, up to intervention A 1335, can be added to the decision tree. In the illustrative example, user 105 can select an intervention from the decision tree to input and / or view information associated with that intervention. Figure 13B As shown, in response to user 105 selecting an optional intervention operation 1 1325, information associated with intervention operation 1 1325 can be provided to user 105. As shown, this information may include relevant intervention techniques (e.g., intervention technique A 1305 in the illustrated example), predicted intervention success rate 1340, intervention duration 1345, intervention cost 1350, and / or intervention configuration 1355. According to certain aspects of this disclosure, the intervention success rate 1340 may be generated by a predictive model, as referenced herein. Figure 15 To be discussed in more detail.

[0058] In an illustrative example of a decision tree, the decision tree can indicate intervention objectives such as flow reduction and skin reduction. The decision tree can include a first intervention operation, such as a cable gauge cutter operation, with an 88% predicted operation success rate, a first duration, and a first cost. The decision tree can include a second intervention operation, such as a cable flow scanner operation, with an 80% predicted operation success rate, a second duration, and a second cost. The decision tree can include a third intervention operation, such as a cable flow-caliper imaging detector (PFCS-A) operation, with an 87% predicted operation success rate, a third duration, and a third cost. The decision tree can include a fourth intervention operation, such as a cable charging operation, with a 90% predicted operation success rate, a fourth duration, and a fourth cost. The decision tree can provide an overall predicted operation success rate, overall duration, and overall cost of 55.12%.

[0059] Figure 14 A well intervention plan report display 1400 is shown for a user interface 110 used for an example well intervention plan report. As shown, the well intervention plan report can display intervention objectives 1405, path information 1410, and inflow information 1445. Path information 1410 can include one or more intervention operations 1415. For each intervention operation 1415, path information 1410 can include associated intervention techniques 1420, estimated costs 1425 of operation 1415, estimated duration of operation 1415 1430, predicted probability of success of intervention operation 1415 1435, and / or exit point 1440 of operation 1415. Exit point 1440 can indicate the implementation cost and / or duration at which the operation can be aborted. In some cases, an emergency path may be followed when the operation is aborted. Inflow information 1445 can include estimated, predicted, or simulated inflow values ​​of one or more fluids associated with the well. Figure 14 As shown, inflow information 1445 may include water inflow 1450, oil inflow 1455, and / or gas inflow 1460 information. Inflow information allows for the assessment of inflows even without planned well intervention, compared to inflows following a well intervention. In some respects, well intervention plans may be implemented or not, at least in part, based on differences in inflows.

[0060] According to aspects of this disclosure, machine learning is used to train a predictive model of the probability of success of an intervention. In some aspects, machine learning algorithms can be used to generate predictive models that predict the probability of success of an intervention. In some aspects, the predictive model is trained to predict the probability of success of the intervention based on the current state, such as based on input from user 105. In some aspects, training involves reward parameters that maximize or minimize an objective function.

[0061] In some respects, machine learning algorithms can be modeled as Markov decision processes (MDPs), which can be reinforcement learning algorithms, deep learning algorithms, supervised learning algorithms, unsupervised learning algorithms, semi-supervised learning algorithms, Q-learning algorithms, value reinforcement algorithms, polarity reinforcement algorithms, deep convolutional networks (DCNs), or combinations thereof.

[0062] In some examples, neural networks are used to perform machine learning. Neural networks can be designed with a variety of connection patterns. In a feedforward network, information is passed from lower layers to higher layers, with each neuron in a given layer communicating with neurons in higher layers. Hierarchical representations can be built into successive layers of a feedforward network. Neural networks may also have recurrent or feedback (also known as top-down) connections. In a recurrent connection, the output from a neuron in a given layer can be passed to another neuron in the same layer. Recurrent architectures can be helpful in recognizing patterns across multiple chunks of input data that are passed sequentially to the neural network. Connections from neurons in a given layer to neurons in lower layers are called feedback (or top-down) connections. Networks with many feedback connections can be helpful when recognizing high-level concepts can help discern specific low-level features of the input. Individual nodes in an artificial neural network can mimic biological neurons by taking input data and performing simple operations on the data. The results of performing simple operations on the input data are selectively passed to other neurons. Weights are associated with each vector and node in the network, and these values ​​constrain how input data is associated with output data. The weights can initially be determined by an iterative stream of training data through the network (e.g., weights are established during the training phase, where the network learns how to identify specific classes based on the characteristics of its typical input data). Different types of artificial neural networks can be used to implement machine learning, such as recurrent neural networks (RNNs), multilayer perceptron (MLP) neural networks, convolutional neural networks (CNNs), etc. These multilayer architectures can be trained layer by layer and can be fine-tuned using backpropagation.

[0063] In one example, the Random Forest machine learning algorithm is used to predict the probability of success of an intervention. The Random Forest algorithm is a supervised classification ensemble machine learning algorithm. Supervised learning can be useful when a large amount of data is available for model training. Through supervised training, the training data contains input and target values; the algorithm learns a pattern that maps input values ​​to output values, and uses this pattern to predict future output values ​​based on the input values.

[0064] The Random Forest algorithm involves creating many decision trees during the model training phase. Each decision tree can be built using random subsets of data from the training dataset to measure a random subset of features in each partition. This randomness introduces variability among the individual decision trees, reducing the risk of overfitting and improving overall predictive power. A process known as "bagging" involves training weak models on different subsets of the training data and sampling each subset with replacements. For predictions, the Random Forest algorithm aggregates the results of all decision trees either by majority voting (e.g., for classification tasks) or by averaging (e.g., for regression tasks) the predictions of the weak models. This collaborative decision-making process, supported by multiple insightful decision trees, provides stable and accurate results.

[0065] Figure 15 An example random forest algorithm 1500 for predicting the success probability of intervention operations is described. Figure 15 As shown, the algorithm can begin with a training data collection phase 1505. In some aspects, the training data collection phase 1505 includes data ingestion, parsing, and contextualization. In some aspects, the training data collection phase continues even after the predictive model has been deployed, in order to fine-tune and improve the model. In some aspects, the training data includes historical information 1510 associated with previous well interventions. In some aspects, historical well intervention information is collected from many (e.g., all) users of the well intervention planning system 100, which may be associated with many different wells. In some aspects, the training data includes cable and slipline operation data. Operational logging for cable and slipline operations can be collected, processed, and stored as an optimized data structure. In some aspects, the training data is stored in a training data repository.

[0066] Cable operations involve using cables to lower tools and instruments into the wellbore to perform various types of well intervention tasks. Cables can consist of a core conductor, typically made of steel, surrounded by several layers of armored wire for reinforcement and protection. One type of cable is the E-line, which contains an electrical conductor that transmits real-time data and control signals between the surface and downhole tools. Another type is the slide wire, a simpler, non-conductive cable with no electrical components, primarily used for mechanical operations. Cable operations can include logging, which uses tools to measure the physical properties of the formation and fluids in the wellbore using tools that record various data, such as resistivity, porosity, and gamma rays. Cable operations may include perforation, the placement and firing of perforating guns to create holes in the casing and cement, allowing reservoir fluids to enter the wellbore. Another type of cable operation is setting and retrieval, installing or removing plugs, chokes, safety valves, packers, and other downhole equipment. Another type of cabled operation is retrieval, which involves recovering lost or stuck equipment from the wellbore. Another type of cabled operation is mechanical servicing, performing tasks such as moving casing, cutting fittings, cutting paraffin, setting packers, adjusting valves, and other interventions requiring mechanical force. Operational data associated with wireless operations may include, but is not limited to, depth and pressure, temperature, and fluid type.

[0067] Depending on some aspects, the training data collection phase 1505 includes a data preprocessing phase 1515. In some respects, the table can span combinations of available operational data, such as jobs, runs, passes, wells, and business environments. Data preprocessing can be performed using various techniques, such as imputation to handle missing values, outlier removal, unit standardization, data classification, class rebalancing, rescaling, and dummy encoding.

[0068] In some respects, data preprocessing stage 1515 can begin by handling missing values, for example, by imputing or removing missing values, to ensure a complete and reliable dataset. Next, because random forests use numerical inputs, categorical variables in the data can be encoded using techniques such as one-hot encoding or label encoding to convert the categorical variables into numerical format. Data preprocessing stage 1515 can further include data scaling and normalization, which can help improve the efficiency of the predictive model training process.

[0069] In some respects, the training data collection phase 1505 includes a feature generation phase 1520. The feature generation phase 1520 may include evaluating the importance of features in the dataset and selecting relevant features for model training. These features can be mapped to labels on the training data. In some respects, the user 105 can select and input features.

[0070] The training data collection phase 1505 may also include addressing imbalanced data, for example by adjusting class weights or employing resampling to ensure balanced representations during training.

[0071] like Figure 15 As shown, the algorithm can proceed to the prediction model training phase 1525. The model training phase 1525 may include using the prediction model to predict the probability of success of a well intervention and comparing the prediction with the actual outcome of the well intervention. When a specified intervention is achieved for a slipline operation or through a user questionnaire for a cable operation, the intervention can be marked as successful (e.g., "Objective achieved"). In some aspects, interventions are marked by users of the system. A binary classification prediction model can predict true / false outputs (e.g., successful or unsuccessful), which can be scored against labels, and different algorithms can be optimized by comparing the scores.

[0072] As shown, a decision tree 1530 is constructed for model training. A decision tree is a classification problem because it may begin with a binary decision. For example, a decision might be the success of an intervention. Each node in the decision tree may involve a further binary decision related to the previous binary decision. Observations (data) that meet the node criteria can follow a "yes" path (e.g., represented by a checkmark in the node of decision tree 1530), while observations that do not meet the criteria follow an alternative "no" path (e.g., represented by an "X" in the node of decision tree 1530). Therefore, the nodes in decision tree 1530 are used to split the data into subsets. The decision tree attempts to find the optimal split for the subset of data, which can be trained using the Classification and Regression Tree (CART) algorithm. Metrics such as Gini impurity, information gain, or mean squared error (MSE) can be used to evaluate the quality of the split. When multiple decision trees are ensembled in a random forest algorithm, they predict more accurate results.

[0073] The Random Forest algorithm can be associated with a set of configured hyperparameters, which can be specified before the model training and testing phases. For example, hyperparameters can include node size, number of trees, and number of sampled features.

[0074] To ensure that each decision tree 1530 in the ensemble brings a unique perspective, random feature selection can be performed. During the training of each decision tree 1530, a random subset of features can be selected. This randomness ensures that each tree focuses on different aspects of the data, cultivating a distinct set of predictors within the ensemble.

[0075] By injecting randomness through feature packing, the diversity of the dataset is increased, and the correlation between decision trees is reduced. The packing stage 1535 can be used to test (e.g., validate) the predictive model. The Random Forest algorithm consists of a series of decision trees, each tree in the set consisting of data samples drawn from the training set with replacements. These data samples can be called bootstrap samples. A portion of the training samples can be left out as test data, called out-of-bag (OOB) samples. Packing is a technique that involves creating multiple bootstrap samples from the original dataset, allowing for replacement sampling of data instances, thus producing different subsets of data for each decision tree, and introducing variability into the training process, making the predictive model more robust.

[0076] The determination of the prediction 1540 will vary depending on the type of problem. For regression tasks, the predictions from each decision tree will be averaged, while for classification tasks, the majority vote (i.e., the most common categorical variable (pattern) across all decision trees) will determine the predicted class. This voting mechanism ensures a balanced collective decision-making process. Finally, out-of-bounds (OOB) samples are used for cross-validation to finalize the prediction.

[0077] Once the predictive model has been tested and validated, it can be deployed in the model deployment phase 1545. User 105 can then access the predictive model for predicting the probability of success of intervention operations, which can be used to make decisions during well intervention planning. For example, as described herein, when user 105 uses well intervention planning system 100 to create a well intervention plan, user 105 inputs information into user interface 110. This input can be fed as features into the predictive model, and the predictive model outputs the probability of success for one or more well intervention operations. This prediction can be displayed to user 105. For example, in response to a high probability of success, user 105 can select a well intervention operation for the well intervention plan. In response to a low probability of success, user 105 can specify an emergency well intervention operation, an exit point for a well intervention operation, and / or select a different well intervention operation for the well intervention plan.

[0078] Example operation of predicting the success probability of well intervention plan

[0079] Figure 16 This is a flowchart depicting an example operation 1600 that illustrates the predicted probability of success of an intervention operation used in a well intervention plan. In some aspects, aspects of operation 1600 can be performed by a well intervention plan system, which may include local, remote, cloud, virtualized, and / or distributed hardware components.

[0080] As shown, operation 1600 may include, at operation 1605, receiving input from a user or data connection, the input including at least one or more well intervention operations and one or more well conditions. In some aspects, input is received at the well intervention planning system from a defined component within the well intervention planning system. In some aspects, input is received at the well intervention planning system via a data connection from another device or system remote from the well intervention planning system. In some aspects, input is received at the well intervention planning system from a user of the well intervention planning system via a user interface.

[0081] In some aspects, one or more condition inputs include at least one of completion information, well trajectory information, or well operation information. In some aspects, completion information includes at least one of: casing information, tubing information, perforation information, or whether the well is a cased well or an open hole well. In some aspects, well trajectory information includes at least one of: whether the well is a horizontal well, a vertical well, or an inclined well; the angle at the measurement depth; the azimuth at the measurement depth; or the actual vertical depth at the measurement depth. In some aspects, well operation information includes at least one of: pressure information, temperature information, or fluid density information.

[0082] In some aspects, operation 1600 may include, at operation 1608, generating a predictive model based on historical well intervention data. Operation 1600 may include collecting historical well intervention data from multiple wells to generate a training dataset. Operation 1600 may include inputting missing values ​​from the historical well intervention data to generate the training dataset. Operation 1600 may include converting non-numerical data in the historical well intervention data into numerical data to generate the training dataset. Operation 1600 may include co-scaling the data, normalizing the data, removing outliers from the data, and / or labeling features in the historical well intervention data to generate the training dataset. In some aspects, the predictive model is a machine learning model. Training the machine learning model may include generating multiple bootstrap samples from the training dataset; sampling instances of the training dataset with replacements to generate multiple random subsets of the training dataset; and generating multiple decision trees, wherein the multiple decision trees are associated with random subsets of the training dataset.

[0083] Operation 1600 may include, at operation 1610, inputting one or more well intervention operations and one or more well conditions into the prediction model.

[0084] Operation 1600 may include, at operation 1615, using a predictive model, at least in part based on one or more well conditions, predicting the probability of success of one or more well intervention operations. In some aspects, the predictive model is based on historical well intervention data. In some aspects, the historical well intervention data includes well intervention data collected from multiple wells. In some aspects, the historical well intervention data includes slipline and cable operation data associated with previous well intervention operations. In some aspects, the predictive model is a random forest machine learning model comprising multiple decision trees, each of which is associated with a random subset of the historical well intervention data. In some aspects, each of the multiple decision trees comprises multiple nodes, each node being associated with a subset of the historical well intervention data corresponding to a feature and a second-order binary prediction associated with the probability of success of the well intervention operation. In some aspects, predicting the probability of success of one or more well intervention operations includes generating a probability of success of the well intervention operation based on input from each of the multiple decision trees; and aggregating the probability of success of the well intervention operation from the multiple decision trees. In some aspects, aggregation includes selecting the probability of success of the well intervention operation generated by the highest number of the multiple decision trees. In some aspects, the predictive model is a statistical model. In some respects, a predictive model is another type of learning model or another type of predictive model.

[0085] Operation 1600 may include, at operation 1620, outputting the predicted success probability of one or more well intervention operations. In some aspects, the predicted success probability of one or more well intervention operations is output to a user via a display, such as a local display of a well intervention planning system, a remote display, or a user device. In some aspects, the predicted success probability of one or more well intervention operations is output to a well intervention operation selection component of the well intervention planning system.

[0086] In some aspects, operation 1600 may further include, at operation 1622, selecting or receiving a selection of a well intervention operation for a well intervention plan in response to a predicted success probability of one or more well intervention operations. In some aspects, the well intervention operation is automatically selected by the well intervention planning system. In some aspects, the well intervention operation is selected by a user of the well intervention planning system.

[0087] In some aspects, operation 1600 may further include, at operation 1624, performing a well intervention based on the well intervention plan. In some aspects, performing a well intervention may include performing the well intervention operation according to the well intervention plan. In some aspects, the well intervention plan is executed automatically by a well intervention planning system. In some aspects, the well intervention is performed by a user of the well intervention planning system.

[0088] Example system for predicting the success probability of well intervention programs

[0089] Based on certain aspects, a well intervention planning system 1700 is provided that predicts the probability of success of intervention operations, such as... Figure 17 As shown. The Well Intervention Planning System 1700 can run on a single computing device or across multiple devices.

[0090] As shown, the well intervention planning system 1700 may include one or more user interfaces 1710 on one or more devices including the well intervention planning system 1700, which allow users to interact with the well intervention planning system 1700. The one or more user interfaces 1710 may be... Figure 1 Examples of user interface 110 are provided. In some examples, user interface 1710 may include a graphical user interface (GUI) that displays and / or accepts touchscreen input from the user. User interface 1710 may include one or more input / output (I / O) interfaces that allow one or more I / O devices (e.g., keyboard, display, mouse device, pen input, microphone, etc.) to connect to well intervention planning system 1700. In some aspects, user interface 1710 is configured to receive user input, including intervention objectives, well conditions, user preferences, and / or well intervention operations and techniques, as described herein. In some aspects, user interface 1710 is configured to receive selections of well intervention operations for a well intervention plan from the user in response to a predicted probability of success of one or more well intervention operations.

[0091] like Figure 17 As shown, the well intervention planning system 1700 may include a transceiver 1705 and one or more network interfaces. The transceiver 1705 and the network interfaces allow the well intervention planning system 1700 to connect to a network (e.g., such as the Internet, a local area network (LAN), a wireless local area network (WLAN), a wireless wide area network (WWAN), Wi-Fi, etc.) and / or communicate with other devices, such as devices within the well intervention planning system 1700 and / or devices outside the well intervention planning system 1700. In some aspects, the transceiver 1705 and one or more network interfaces may be configured to receive input via a data connection, including intervention targets, well conditions, user preferences, and / or well intervention operations and techniques, as described herein. In some aspects, the transceiver 1705 and one or more network interfaces may be configured to collect historical well intervention planning information and statistics from multiple users of the well intervention planning system 1700. In some aspects, the transceiver 1705 and one or more network interfaces may be configured to receive selections of well intervention operations for a well intervention plan via a data connection in response to a predicted probability of success of one or more well intervention operations.

[0092] As shown, the well intervention planning system may include a processing system comprising one or more processors 1715. The one or more processors 1715 may be located locally or remotely (e.g., via cloud computing resources) of the well intervention planning system 1700. The one or more processors 1715 may include one or more central processing units (CPUs). The CPU may have multiple processing cores. In some aspects, processor 1715 may include an intervention operation success probability prediction model generator 1720 configured to generate an intervention operation success probability prediction model 1750 based on historical well intervention data 1745. In some aspects, processor 1715 may include a well intervention operation success probability predictor processor 1722 configured to provide input to the intervention operation success probability prediction model 1750, receive and output predicted well intervention operation success probabilities from the intervention operation success probability prediction model 1750, and output the predicted well intervention operation success probabilities. In some aspects, processor 1715 may include a well intervention operation selector 1724 configured to select a well intervention operation for the well intervention plan in response to the predicted well intervention operation success probabilities of one or more well intervention operations. In some aspects, processor 1715 may include well intervention plan executor 1726, which is configured to perform well interventions based on well intervention plans.

[0093] The processing system may also include memory 1740 and / or storage devices, which may be located locally or remotely (e.g., cloud storage devices) of the well intervention planning system 1700. Memory 1740 may represent random access memory (RAM). Storage devices may be a combination of disk drives, fixed or removable storage devices, such as fixed disk drives, removable memory cards or optical storage, network-attached storage (NAS) or storage area network (SAN). The CPU can retrieve and execute program instructions stored in memory 1740. Similarly, the CPU can retrieve and store application data residing in memory 1740. Figure 17 As shown, memory 1740 can store historical well intervention plan information 1745. (As...) Figure 17 As shown, memory 1740 can store a predictive model 1750 of the probability of success of an intervention operation.

[0094] In some aspects, computing resources, including processing and / or memory resources, can be virtualized. A virtual machine (VM) is a software-based simulation of a physical computer running on host hardware and can provide a virtualized computing environment including an operating system, CPU allocation, memory, storage devices, and network interfaces. In some aspects, the well intervention planning system 1700 includes one or more VMs configured to perform predictions of the success rate of intervention operations for well intervention planning. The VMs utilize the computing resources of the well intervention planning system 1700, such as processor 1715 and / or memory 1740.

[0095] Example Terms

[0096] Examples of implementation methods are described in the following numbered aspects:

[0097] Aspect 1: A method for well intervention planning, the method comprising: receiving input from a user or a data connection, the input including at least one or more well intervention operations and one or more well conditions; inputting the one or more well intervention operations and one or more well conditions into a prediction model; using the prediction model, at least in part based on the one or more well conditions, predicting the probability of success of the one or more well intervention operations; and outputting the predicted probability of success of the one or more well intervention operations.

[0098] Aspect 2: According to the method of aspect 1, one or more status inputs include at least one of the following: well completion information, well trajectory information, or well operation information.

[0099] Aspect 4: According to the method described in Aspect 2, wherein: completion information includes at least one of the following: casing information, pipe information, perforation information, or whether the well is a casing well or an open hole well; well trajectory information includes at least one of the following: whether the well is a horizontal well, a vertical well, or an inclined well; the angle at the measurement depth; the azimuth at the measurement depth; or the actual vertical depth at the measurement depth; and well operation information includes at least one of the following: pressure information, temperature information, or fluid density information.

[0100] Aspect 5: The method described according to any combination of aspects 1-4, wherein the predictive model is a statistical model.

[0101] Aspect 6: The method described according to any combination of aspects 1-4, wherein the predictive model is a machine learning model.

[0102] Aspect 7: The method described according to any combination of aspects 1-6 further includes generating a predictive model based on historical well intervention data.

[0103] Aspect 8: The method according to aspect 7, wherein historical well intervention data includes slipline and cable operation data associated with previous well intervention operations.

[0104] Aspect 9: The method described in aspect 8 further includes collecting historical well intervention data from multiple wells to generate a training dataset.

[0105] Aspect 10: According to the method described in aspect 9, wherein generating the training dataset includes missing values ​​in the input historical well intervention data.

[0106] Aspect 11: The method according to any combination of aspects 8-10, wherein generating the training dataset includes converting non-numerical data in historical well intervention data into numerical data.

[0107] Aspect 12: The method according to any combination of aspects 8-11, wherein generating the training dataset includes at least one of the following: scaling the data, normalizing the data, removing outliers from the data, or labeling features in historical well intervention data.

[0108] Aspect 13: The method according to any combination of aspects 8-12, wherein the prediction model is a machine learning model, and wherein training the machine learning model comprises: generating a plurality of pilot samples from a training dataset; sampling instances of the training dataset with replacements to generate a plurality of random subsets of the training dataset; and generating a plurality of decision trees, wherein the plurality of decision trees are associated with random subsets of the training dataset.

[0109] Aspect 14: The method according to any combination of aspects 1-13, wherein the user interface or data connection is further configured to receive a selection of well intervention operations for a well intervention plan in response to a predicted success probability of one or more well intervention operations.

[0110] Aspect 15: The method according to any combination of aspects 1-13, wherein one or more processors are further configured to select a well intervention operation of the well intervention plan in response to a predicted well intervention operation success probability of one or more well intervention operations.

[0111] Aspect 16: The method according to any combination of aspects 1-15, wherein the predictive model is a random forest machine learning model comprising multiple decision trees, and wherein each of the multiple decision trees is associated with a random subset of historical well intervention data.

[0112] Aspect 17: According to the method of aspect 16, each of the plurality of decision trees includes a plurality of nodes, and each node is associated with a subset of historical well intervention data corresponding to a feature and a second-order binary prediction associated with the probability of success of the well intervention operation.

[0113] Aspect 18: The method according to aspect 17, wherein predicting the success probability of one or more well intervention operations includes: generating a success probability prediction of the well intervention operation from each of a plurality of decision trees based on input; and aggregating the success probability predictions of the well intervention operation from the plurality of decision trees, wherein aggregating includes selecting the success probability predictions of the well intervention operation generated by the highest number of the plurality of decision trees.

[0114] Aspect 19: A system for predicting the probability of successful intervention operations for executing any combination of aspects 1-18 of a well intervention plan.

[0115] Aspect 20: The system according to aspect 19, wherein one or more components of the system are located in the cloud.

[0116] Aspect 21: The system according to aspect 19, wherein one or more components of the system include a virtual machine.

[0117] Aspect 22: An apparatus for predicting the probability of successful intervention operation for executing any combination of aspects 1-18 of a well intervention plan.

[0118] Aspect 23: The apparatus according to aspect 22, wherein one or more components of the apparatus are located in a cloud.

[0119] Aspect 24: The apparatus according to aspect 22, wherein one or more components of the apparatus include a virtual machine.

[0120] Aspect 25: A non-transitory computer-readable medium for performing a well intervention plan of any combination of aspects 1-18, predicting the probability of successful intervention operation.

[0121] Other considerations

[0122] The foregoing description is provided to enable any person skilled in the art to practice the various aspects described herein. The examples discussed herein do not limit the scope, applicability, or aspects set forth in the claims. Various modifications to these aspects will readily be apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects. For example, the function and arrangement of the elements discussed may be altered without departing from the scope of this disclosure. Various processes or components may be appropriately omitted, substituted, or added to the various examples. For example, the described methods may be performed in a different order than described, and various actions may be added, omitted, or combined. Furthermore, features described with respect to some examples may be combined in some other examples. For example, any number of aspects set forth herein may be used to implement an apparatus or practice. Moreover, the scope of this disclosure is intended to cover apparatus or methods practiced using structures, functions, or structures and functions other than or different from the aspects of this disclosure set forth herein. It should be understood that any aspect of the disclosure herein may be embodied by one or more elements of the claims.

[0123] The various illustrative logic blocks, modules, and circuits described in this disclosure may be implemented or performed using a general-purpose processor, a digital signal processor (DSP), an ASIC, a field-programmable gate array (FPGA) or other programmable logic device (PLD), discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. The general-purpose processor may be a microprocessor, but alternatively, it may be any commercially available processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors combined with a DSP core, a system-on-a-chip (SoC), or any other such configuration.

[0124] As used herein, the phrase “at least one of a series of items” refers to any combination of those items, including individual components. As an example, “at least one of a, b, or c” is intended to cover a, b, c, ab, ac, bc, and abc, as well as any combination of multiple identical elements (e.g., aa, aaa, aab, aac, abb, acc, bb, bbb, bbc, cc, and ccc, or any other order of a, b, and c).

[0125] As used herein, the term "determine" encompasses a wide variety of actions. For example, "determine" can include calculation, operation, processing, derivation, investigation, lookup (e.g., searching in a table, database, or other data structure), ascertainment, etc. Furthermore, "determine" can include receiving (e.g., receiving information), accessing (e.g., accessing data in memory), etc. Additionally, "determine" can include parsing, selecting, picking, building, etc.

[0126] The methods disclosed herein include one or more actions for implementing the methods. These method actions may be interchanged without departing from the scope of the claims. In other words, unless a specific order of actions is specified, the order and / or use of a particular action may be modified without departing from the scope of the claims. Furthermore, the various operations of the above methods can be performed by any suitable means capable of performing the corresponding functions. Such means may include various hardware and / or software components and / or modules, including but not limited to circuits, application-specific integrated circuits (ASICs), or processors.

[0127] The following claims are not intended to be limited to the aspects shown herein, but are consistent with the full scope of the language of the claims. In the claims, unless specifically stated otherwise, an element referred to in the singular does not mean “one and only one,” but rather “one or more.” Unless otherwise specifically stated, the term “some” means one or more. Unless the elements of a claim are expressly stated using the phrase “means for…”, they should not be interpreted in accordance with the provisions of 35 U.SC 112(f). All structural and functional equivalents of the elements of all aspects described throughout this disclosure that are known to or will be known hereafter by one of ordinary skill in the art are expressly incorporated herein by reference and are intended to be included in the claims. Furthermore, nothing disclosed herein is intended to be offered to the public, whether or not such disclosure is expressly stated in the claims.

Claims

1. An automated well intervention planning system, the system comprising: one or more memories storing computer executable code; a user interface or data connection configured to receive inputs, the inputs comprising at least one or more well intervention operations and one or more well conditions; and one or more processors configured to execute the computer executable code and to: input the one or more well intervention operations and the one or more well conditions to a prediction model; using the prediction model, predict a well intervention operation success probability for the one or more well intervention operations based at least in part on the one or more well conditions; and output the predicted well intervention operation success probability for the one or more well intervention operations.

2. The system of claim 1, wherein at least one of: the user interface or data connection is further configured to receive a selection of a well intervention operation of a well intervention plan in response to the predicted well intervention operation success probability for the one or more well intervention operations; or the one or more processors are further configured to select a well intervention operation of a well intervention plan in response to the predicted well intervention operation success probability for the one or more well intervention operations.

3. The system of claim 1, wherein the one or more condition inputs comprise at least one of: completion information, wellbore trajectory information, or wellbore operation information.

4. The system of claim 3, wherein: the completion information comprises at least one of: casing information, tubing information, perforation information, or whether the well is a cased well or an open hole well; the well trajectory information comprises at least one of: whether the well is a horizontal well, a vertical well, or a deviated well; an angle at a measured depth; an azimuth at a measured depth; or a true vertical depth at a measured depth; and the well operation information comprises at least one of: pressure information, temperature information, or fluid density information.

5. The system of claim 1, wherein the prediction model is based on historical well intervention data.

6. The system of claim 5, wherein the historical well intervention data comprises well intervention data collected from a plurality of wells.

7. The system of claim 5, wherein the historical well intervention data comprises wireline and cable tool operation data related to previous well intervention operations.

8. The system of claim 5, wherein the prediction model is a random forest machine learning model comprising a plurality of decision trees, and wherein each of the plurality of decision trees is associated with a random subset of the historical well intervention data.

9. The system of claim 8, wherein each of the plurality of decision trees comprises a plurality of nodes, and wherein each node is associated with a subset of the historical well intervention data corresponding to a feature and a binary prediction associated with a well intervention operation success probability.

10. The system of claim 8, wherein predicting the well intervention operation success probability for the one or more well intervention operations comprises: generating a well intervention operation success probability prediction from each of the plurality of decision trees based on the inputs; and aggregating the well intervention operation success probability predictions from the plurality of decision trees, wherein the aggregating comprises selecting the well intervention operation success probability prediction generated by the highest number of the plurality of decision trees.

11. The system of claim 5, wherein the predictive model is a statistical model or a machine learning model.

12. The system of claim 5, wherein the one or more memories and one or more processors are located in the cloud or one or more virtual machines.

13. A method for well intervention planning, the method comprising: receiving input from a user or a data connection, the input comprising at least one or more well intervention operations and one or more well conditions; inputting the one or more well intervention operations and one or more well conditions into a predictive model; using the predictive model, predicting a well intervention operation success probability for the one or more well intervention operations based at least in part on the one or more well conditions; and outputting the predicted well intervention operation success probability for the one or more well intervention operations.

14. The method of claim 13, further comprising generating the predictive model based on historical well intervention data.

15. The method of claim 14, further comprising collecting the historical well intervention data from a plurality of wells to generate a training data set.

16. The method of claim 15, wherein generating the training data set comprises imputing missing values in the historical well intervention data.

17. The method of claim 15, wherein generating the training data set comprises converting non-numerical data in the historical well intervention data to numerical data.

18. The method of claim 15, wherein generating the training data set comprises at least one of: scaling the data, normalizing the data, removing outliers from the data, or labeling features in the historical well intervention data.

19. The method of claim 15, wherein the predictive model is a machine learning model, and wherein training the machine learning model comprises: generating a plurality of bootstrap samples from the training data set; sampling instances of the training data set with replacement to generate a plurality of random subsets of the training data set; and generating a plurality of decision trees, wherein the plurality of decision trees are associated with the random subsets of the training data set.

20. A computer-readable medium storing computer executable code for well intervention planning, the computer executable code comprising: code for receiving input from a user or a data connection, the input comprising at least one or more well intervention operations and one or more well conditions; code for inputting the one or more well intervention operations and one or more well conditions into a predictive model; code for using the predictive model, predicting a well intervention operation success probability for the one or more well intervention operations based at least in part on the one or more well conditions; and code for outputting the predicted well intervention operation success probability for the one or more well intervention operations. ​ ​

Images