Fracturing equipment intelligent cross-domain layout system and method based on digital twinning and transfer learning

Abstract

The invention discloses a fracturing equipment intelligent cross-domain layout system and method based on digital twinning and transfer learning, and the system comprises a parameterized model building module which is used for building a basic model of an equipment and environment digital twinning model in a fracturing scene; the scene unit interaction module is used for solving an optimal solution set of the equipment number and the point location in the fracturing well site by taking the fracturing operation chain and the spatial constraint as a comprehensive weighted objective function; and the cross-domain scene coupling module fuses the features of the multi-source domain and the target domain through a cross-domain scene knowledge multiplexing mechanism to output a scene candidate scheme, the optimal layout scheme is output through verification and iteration of an expert system, and the system analyzes and renders to complete the intelligent cross-domain layout of the fracturing scene. According to the method, through digital twin model rehearsal simulation, a cross-domain knowledge reuse mechanism and expert system correction collaborative layout, a well site layout scheme with multiple operation areas and multi-type equipment function linkage is efficiently constructed, the well site construction period is shortened, the pressure test risk before construction is reduced, and the overall efficiency of fracturing mining operation is improved.

Description

Technical Field

[0001] This invention belongs to the fields of well fracturing equipment layout design and computer science, and relates to unconventional oil and gas exploration technology. Specifically, it relates to an intelligent cross-domain layout system and method for fracturing equipment based on digital twins and transfer learning, which is applicable to the layout design of fracturing well sites with multiple operating areas and multiple types of equipment linkage in complex outdoor environments. Background Technology

[0002] Unconventional oil and gas resource development has become a significant trend in the global energy sector. Hydraulic fracturing technology, as a core extraction technology, is crucial for solving the challenges of low-permeability oil and gas reservoirs and improving single-well production and extraction efficiency. Fracturing well sites are complex outdoor environments, involving the coordinated operation of various types of equipment across multiple spatial areas to form a complete fracturing operation chain. However, current fracturing well site construction still relies on the traditional method of manual layout design using two-dimensional drawings and on-site pressure testing. This results in non-production stages such as equipment layout, setup, and removal accounting for 10%-30% of the time, severely restricting operational efficiency and resource development benefits.

[0003] Current research and technology mainly focus on optimizing downhole fracturing process parameters, with insufficient attention paid to optimizing the layout design of surface operation facilities and reducing non-productive time, resulting in a significant technical gap. Furthermore, traditional construction methods heavily rely on the professionalism of practitioners. When faced with scenarios that require consideration of spatial distribution, functional coordination, and operational process connections, a significant amount of time is spent on placing individual equipment points. At the same time, on-site construction and pressure testing are prone to safety risks related to the compatibility parameters of equipment in the operational chain, further extending the operation cycle and reducing overall efficiency.

[0004] Currently, most digital twin cross-domain migration technologies focus on cross-domain modeling of digital twins for individual devices or generating digital twin operation and maintenance solutions for single-area process production lines. Their technical logic is only suitable for application scenarios involving a single object or an independent area. However, for complex industrial scenarios such as fracturing well sites, which involve multiple operating areas, multiple types of equipment with interconnected functions, and the need to form a complete operation chain, there is currently no mature digital twin cross-domain construction technology solution. This technological gap directly leads to low efficiency and insufficient adaptability in the deployment of digital twins for complex cluster equipment scenarios. Summary of the Invention

[0005] To address the aforementioned industry problems and technological gaps, this invention provides an intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning. By constructing virtual multi-regional operation scenarios using digital twin technology, it enables pre-operational simulation of the functional linkage of multiple types of equipment and early risk mitigation. Furthermore, by leveraging cross-domain scenario knowledge reuse mechanisms and expert system correction mechanisms, the transfer learning framework is optimized, enhancing the cross-domain adaptability of equipment layout in complex scenarios. Ultimately, this overcomes the high dependence limitations of traditional construction methods, shortens non-production operation time, improves the overall extraction efficiency of fracturing operations, and provides technical support for the efficient development of unconventional oil and gas resources.

[0006] To achieve the above objectives, the embodiments of the present invention are implemented through the following technical solutions:

[0007] One aspect of this invention provides an intelligent cross-domain layout system for fracturing equipment based on digital twins and transfer learning, comprising:

[0008] (1) Parametric model building module: Construct the basic elements required for scene layout design, namely, establish a digital twin model of multiple operation areas including equipment digital twin model and environment digital twin model. The equipment digital twin model is associated with the fusion and verification of fracturing operation feature attribute parameters, functional linkage constraints and management. The environment digital twin model includes fracturing operation external space boundary threshold and terrain adaptation parameters.

[0009] (2) Scene Unit Interaction Module: Under the constraints of quantity matching, operation chain relationship and spatial constraints of natural environment, the equipment digital twin model calculates the multi-objective solution set of the optimal number of equipment and equipment placement in the scene with five factors as comprehensive weighted objectives: topology, information, cost, safety and efficiency.

[0010] (3) Cross-domain scenario coupling module: Through the cross-domain scenario knowledge reuse mechanism, the historical layout experience data of fracturing operations in different source domains are called, and the unique constraint parameters of the target domain operation scenario are combined to solve the comprehensive weighted objective optimization function and generate a set of candidate layout schemes. The candidate layout schemes are then corrected for constraint weights and verified for feasibility through an expert system. After iterative optimization, the optimal layout scheme is output. The digital twin model analyzes the optimal layout scheme to complete the rendering of three-dimensional scene data, thereby realizing the intelligent cross-domain layout of the multi-unit collaborative cluster of fracturing.

[0011] Another aspect of this invention provides a method for intelligent cross-domain deployment of fracturing equipment based on digital twins and transfer learning, comprising the following steps:

[0012] Step 1: Encode and establish the feature data and label classification of all physical entities, including the feature data of physical entities of operating equipment and physical entities of the operating environment;

[0013] Step 2: Establish a digital twin model corresponding to the physical entity, and parameterize the constraints of the model, including limitations on the number of devices, terrain limitations, and the correlation and coupling constraints between various features;

[0014] Step 3: Correct the constructed digital twin model and classify it to establish a 3D model library to provide basic elements for scene construction and layout; the user selects the scene construction method, including two methods: manual construction of single-source scenes and cross-domain scene migration construction, i.e., select step 4 or step 5;

[0015] Step 4: Select the manual construction method for a single-source scenario of the digital twin model, based on physical boundary constraints. Weighted composite objective Based on the interaction relationships between devices and between devices and the environment, the layout and positioning coordinates of each model are set one by one, thereby completing the construction of all elements in the digital twin scene. If all interaction logic within the scene complies with the work guidelines, the manual layout design of the single-source scene is completed. The system analyzes and renders to obtain a digital twin fracturing well site; conversely, it is necessary to reconstruct the interaction between equipment and between equipment and the environment, and optimize the scene layout scheme until it meets the operating criteria.

[0016] Step 5: Select the cross-domain scenario migration construction method for the digital twin model. The user inputs the requirements and constraints of the target domain scenario. The digital twin model extracts the feature knowledge of the target scenario. Through the cross-domain scenario knowledge reuse mechanism, candidate layout schemes are obtained by training and iterating from the historical layout data of fracturing operations in each source domain. The expert system adjusts the constraint weights of candidate layout schemes until they meet the operational criteria. This layout scheme is then considered the optimal layout scheme for the fracturing equipment in the target domain. The system analyzes and instantiates the coordinates of twin model points, visualizes and renders the target fracturing well site, and realizes intelligent cross-domain layout of fracturing equipment with multiple regions, multiple units, and functional linkage.

[0017] As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. This invention constructs a pre-simulation and risk prediction mechanism for the operation chain of fracturing equipment based on digital twin technology. Before operation, the entire fracturing operation process is simulated through digital twin scenarios, and the simulation model is used to predict the downtime and other faults during operation, effectively eliminating potential operational risks, greatly reducing non-production construction time such as physical well site construction and evacuation, and thus improving overall operational efficiency.

[0019] 2. This invention proposes a dual-mode human-machine collaborative layout and construction mechanism. On the one hand, it supports grassroots industry personnel to build scenarios based on historical operational experience; on the other hand, it allows non-professionals to achieve automatic and intelligent scenario construction by calling pre-stored expert knowledge through the system. Both modes can integrate multi-dimensional weighted objectives such as topology, safety, and efficiency to generate the optimal scenario layout scheme under the current operational environment and equipment constraints, solving the problems of high dependence on personnel expertise and incomplete consideration of objective constraints in traditional two-dimensional drawing construction methods.

[0020] 3. This invention breaks through the limitations of traditional "single-scene layout design" and constructs a digital twin cross-domain layout architecture adapted to the characteristics of "multi-region, multi-unit, and functional linkage" in fracturing operations. It combines expert system to correct parameters for secondary iteration, forming a closed-loop optimization of "algorithm optimization - manual verification - scene adaptation". It takes into account the spatial distribution characteristics of equipment, functional collaboration attributes and work process correlation, which not only improves the cross-domain learning performance of scene layout, but also effectively suppresses negative transfer phenomenon and the problem of blurred transfer boundary, solving the technical pain point of insufficient adaptability of existing transfer learning frameworks in complex industrial cluster equipment scenarios. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of the intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning provided in the embodiments of this application;

[0023] Figure 2 This is a system architecture diagram of an intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning, provided in an embodiment of this application.

[0024] Figure 3 This is a schematic diagram of the cross-domain layout method for transfer learning networks provided in the embodiments of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, this invention adopts the following technical solution.

[0026] This invention proposes an intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning. The architecture of its layout scheme design consists of three main implementation modules: a parameterized model establishment module, a scene unit interaction module, and a cross-domain scene coupling module. Figure 2 As shown. The implementation process of this method is detailed below:

[0027] Parametric model building module: The establishment of equipment model and environment model is the process of building digital twin models of two different objects: physical operating equipment and physical natural environment;

[0028] In this embodiment of the invention, key features of physical operation equipment and physical natural environment are quantitatively analyzed and extracted, including their geometric features, physical features, behavioral features and rule features. A digital twin model is established for its main features. A digital twin model is also established for the correlation and fusion constraint relationship between the features. After the established digital twin model is corrected and verified by introducing parameter correction parameters, the equipment twin model and the twin environment model are compiled into the model library for classification, storage and management.

[0029] Scene Unit Interaction Module: The scene unit interaction module enables the interaction and mutual constraints between basic element units in the digital twin scene, thereby realizing the collaborative implementation of unit equipment in multiple regions and the functional linkage between multiple types of facility clusters to form a complete fracturing operation chain.

[0030] like Figure 2 As shown, the set of device types for: Where EE represents energy equipment, ME represents monitoring equipment, LE represents living equipment, DE represents detection equipment, RME represents raw material equipment, FE represents fracturing equipment, OE represents operational equipment, and OTE represents other equipment. The decision variables for each type of equipment are... binary variables Indicates whether device type e is placed at position i, i.e. Indicates placement, that is Indicates not to place, among which The quadruple [x, y, z, r] represents the position state of a scene at coordinates (x, y, z) with a rotation angle r around the z-axis. This represents the set of locations for the selected equipment.

[0031] like Figure 2 As shown, the constraints mainly include, but are not limited to, three main constraints: matching constraints between the number of devices, natural environment constraints, and the relationship between the number of devices and their placement.

[0032] The device quantity matching constraint mentioned in the scene unit interaction module is specifically expressed as follows: for any device type They all ,in For device e to device The quantity requirement factor. Taking the quantity configuration of fracturing equipment, operating equipment, and energy equipment as an example, assuming one fracturing truck ( If two sets of piping connections for the fracturing equipment (OE) and one set of power supply for the energy equipment (EE) are required, then the demand factor of the fracturing equipment (FE) on the working equipment (OE) is... Demand factor for energy equipment (EE) Other cases are set according to the matching relationship between devices.

[0033] The natural environment constraints mentioned in the scene unit interaction module are specifically represented as follows: for position ,set up Let i be the environmental constraint function for location i. When the environmental constraint is triggered, the device must not be placed at that location; this is determined by the formula: Taking rivers or houses as an example, in a twin scene, placing devices is prohibited when the device is less than 5 units away from the river / house. ,in Given the distance from location i to the river / house, the following conditions must be met. The equipment cannot be placed at that time, that is .

[0034] The relationship between the number of devices and their placement in the scene unit interaction module is specifically expressed as: the total number of device type e. It equals the sum of the number of items placed at its location, and is determined by the formula: ,in This is the location storage coefficient for device e, and the specific situation is determined based on the device's own functional attributes.

[0035] In summary, to obtain the number of devices that satisfy all constraints in the scene unit interaction module... and placement The Pareto optimal solution set and the weighted comprehensive objective function formula are as follows:

[0036] ;

[0037] The topological dimension weights in the above formula Information dimension weight Cost dimension weight Security dimension weight Weight of benefit dimension ;

[0038] Topology target: This indicates the rationality of the topological connections between devices (such as connectivity, path length, etc.); Information target: This indicates the efficiency of information transmission between devices (such as latency and bandwidth utilization); cost target: This represents the total cost of equipment deployment (such as purchase, transportation, and installation costs); security objectives: This indicates the safety level of equipment operation (such as risk value, probability of accident, etc.); benefit objectives: This represents the benefits of equipment operation (such as production capacity, revenue, etc.). Each objective function... Normalization is performed, i.e., mapping to the interval ([0,1]). In practical applications, it needs to be adjusted according to the monotonicity of the function. It can be "inverted" or the sign can be adjusted to adapt to the minimization / maximization objective. For example, the smaller the cost objective, the better, and the larger the benefit objective, the better.

[0039] Cross-domain scene coupling module: This module is a transfer learning framework optimized based on a cross-domain scene knowledge reuse mechanism. Under similar constraints, it solves the optimal adaptive layout scheme for similar scene units. Specifically, the digital twin model extracts knowledge from elements of a single-job scene from multiple source domains, uses this knowledge as a dataset for training, and updates the cross-domain shared weights. Simultaneously, knowledge is extracted from the feature weights of the target scene, combining user needs and environmental elements. The knowledge parameters obtained by weighted geometric calculation are then input into a transfer learning network (TL-Net) to output candidate solutions for the scene. After data verification and expert monitoring to verify whether it meets the construction guidelines, the optimal layout scheme is obtained. The target work scenario is obtained by analyzing scene data through a digital twin model and rendering it. Conversely, the weighted parameters are adjusted to calculate new knowledge parameters, and the above transfer training process is repeated until the criteria are met.

[0040] The specific implementation process of this invention embodiment is as follows: Figure 1 The solution process for the intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning consists of the following steps:

[0041] Step S110: The digital twin model is digitally modeled based on the mechanistic characteristics of all physical entities, including two different objects: the physical operating environment and the physical natural environment. This involves analyzing the key features of the physical operating equipment and the physical natural environment, extracting the main characteristic factors affecting the collaborative layout, including geometric features, physical features, behavioral features, and rule features; compressing secondary features, including operational attributes, functional features, and inherent setting attributes. The main characteristic factors are then digitally processed and symbolized into their respective mathematical models for easy storage and retrieval.

[0042] Step S120: A digital twin model is established based on the feature data of physical entities, where physical entities include operating equipment and natural environment physical entities. During the modeling process, the data of each model are parameterized, including equipment quantity restrictions, terrain restrictions, correlation constraints, coupling constraints, etc. between various features.

[0043] Step S130: Perform parameter correction analysis on the feature data and label data of the established digital twin model. This involves using a normalization function to process the data and unify the dimensions of each data point to complete model correction. Simultaneously, the expert system verifies and revises the 3D model. After completing model correction and verification, the twin device model and twin environment model are categorized, stored, and managed to build a scene element model library, providing basic elements for subsequent scene construction and layout.

[0044] In the method of this invention, the multi-unit collaborative cluster scenario layout and construction scheme based on digital twins mainly includes two main methods: a single-source manual construction method and a multi-source scenario cross-domain migration layout method. The specific implementation methods are as follows:

[0045] If the single-source manual scenario setup method is selected, proceed to step S140: The digital twin model is mainly built based on the user-defined interaction relationships between devices and the interaction logic between devices and the environment to create a single operational scenario. After the user confirms that all interaction logic within the scenario conforms to the operational criteria, the single-source manual scenario layout design is completed. The system analyzes the design scheme to obtain the digital twin fracturing well site. Otherwise, it is necessary to rebuild the interaction relationships between devices and between devices and the environment until they conform to the operational criteria. The detailed implementation process of step S140 is as follows:

[0046] Step S141: Enter the single-source layout scheme construction design mode. The user inputs physical boundary constraints, such as the size of the work site, fixed obstacle constraints, safety margin settings, etc. Then, the user selects the optimization objective factor, sets the weight of each objective, and calculates the weighted comprehensive objective function for this scenario. ;

[0047] Step S142: The user selects a digital twin model of the natural environment from the model library to build a work twin scene model, sets the surrounding parameters of the environment model according to the physical characteristics of each environment model, and saves the digital twin natural environment scene model as a scene map.

[0048] Step S143: Based on the natural environment scene model, the user selects the equipment in each stage and, based on the interrelationship between the equipment, sequentially selects the layout coordinates of the model in the scene. This enables the pre-setting of the positioning of functional areas within the macro-scene layout;

[0049] Step S144: The system calculates and solves the pre-position of the functional area in the scene based on the surrounding parameters of each environmental model, determines whether the interaction between the device and the environment conforms to the operating criteria, and provides adjustment suggestions.

[0050] Step S145: Based on the suggestions, the user sets the layout and positioning coordinates of the functional area according to the interaction relationships between devices and between devices and the environment, and calls a series of device models that implement the function from the model library to fill the corresponding functional area, thus completing the layout point design scheme for all models. ;

[0051] Step S146: In the digital twin model, the expert review subsystem calculates the impact factors of all functional connection logic and safety margins within the scenario. After confirming that all logic within the scenario conforms to the operational guidelines, it completes the layout design for the single-operation scenario. The system analyzes and renders to obtain a digital twin fracturing well site; otherwise, it proceeds to steps S143 to S146 to reset the layout points until they meet the operating criteria.

[0052] If the cross-domain migration layout method in a multi-source scenario is selected, proceed to step S150: The digital twin model deconstructs and encodes the new user input requirements and the constraints of the new working environment, and uses the scenario knowledge after data preprocessing and analysis, i.e., the network shared parameter weights. The input transfer learning network (TL-Net) performs cross-domain transfer of equipment layout points in a work scenario. It obtains the optimal scenario solution through four core stages: multi-source domain knowledge extraction and encoding, cross-domain transfer of weight parameters, adaptive fine-tuning of the target domain, and iterative verification based on expert feedback. Finally, the digital twin model analyzes the scene data and renders the target task scene. The transfer learning network structure diagram is as follows: Figure 3 As shown, the detailed implementation process of step S150 is as follows:

[0053] Step S150.1: Multi-source domain knowledge extraction and encoding. The system selects and calls source domain datasets containing equipment layout locations, scene constraints, and environmental features from multiple source domains (such as "Source Domain 1", "Source Domain 2", etc., for different work scenarios). The dataset is extracted using features extracted from a classifier (such as the feature extraction function of Fast Fourier Transform, FFT). The knowledge encoding of each source domain data is performed based on the "location-parameter-constraint" association, and the source domain knowledge representation is generated and input into the transfer learning network for pre-training.

[0054] The above-mentioned encoding of the dataset is specifically represented as follows: This is an identifier for the number of source domains, i.e., the first... The datasets for each source domain are:

[0055] ;

[0056] In the formula, For source domain The scene condition feature vector (s represents the source domain, and t represents the target domain). Set of equipment layout points ( Represents the source domain No. In the first scenario (Layout coordinates of the device class) For the set of scene constraints ( The number of constraints. For the first The first source domain The sample at the th Under each constraint, the model outputs a violation function indicating the degree to which the constraint is violated; it returns 0 if the layout satisfies the constraint, and a positive value otherwise. This represents the total number of scene samples contained in the k-th source domain.

[0057] Step S150.2: TL-Net Pre-training and Weight Aggregation. TL-Net takes multi-source domain knowledge representations as input, learns the general mapping relationship between "scene features - constraints - device layout", and obtains the pre-trained weights of each source domain by minimizing the joint objective of "layout prediction error loss" and "constraint violation loss". And aggregate the pre-trained weights from multiple source domains to obtain cross-domain shared weights. It includes core parameters for extracting general features of equipment and modeling constraint relationships, and will The data is migrated to the corresponding layer of the target domain (fracture operation well site) model and used as the initial weight of the target domain model, inheriting the common knowledge of multiple source domains.

[0058] Specifically, for each source domain The TL-Net training is completed by optimizing the following objective function (i.e., minimizing the weighted sum of the multi-source domain layout prediction loss and the multi-source domain constraint violation loss), thus obtaining the optimal parameter weights of the model in the k-th source domain pre-training stage (pre). :

[0059] ;

[0060] In the formula, The set of learnable parameters for the model. The weighting coefficients for the prediction loss of the k-th source domain layout point are: The weight coefficients of the k-th source domain constraint loss satisfy... , ; For the first The source domain layout point location mean square error prediction loss function. For the network prediction of the first Equipment locations; For the first The constraints of the source domain violate the loss function. For the forward inference function of TL-Net, It is an FFT-based feature extraction function;

[0061] By aggregating the shared parameter portion of pre-trained weights from multiple source domains, cross-domain shared weights can be obtained. :

[0062] ;

[0063] In the formula, For the number of source domains, For source domain A subset of shared parameters in pre-trained weights. This is the weighted aggregation function, typically an average or attention-weighted aggregation function.

[0064] Step S150.3: Target domain migration and iterative fine-tuning optimization of the target. Initial dataset for the target domain fracturing scenario. Includes initial fracturing scene features Scenario-specific constraints Due to the limited initial data volume, the "frozen core parameters" approach was adopted. +Fine-tuning the number of adaptation layers The strategy optimizes the joint objective of "target domain layout prediction error - constraint violation degree - transfer regularization term" to obtain a preliminary fit of model weights. .

[0065] Shared weights during pre-training During the migration to the target domain, the parameters of the adaptation layer are fine-tuned in the m-th round. The optimization objective is:

[0066] ;

[0067] In the formula, , For the first The layout prediction loss and constraint violation loss of the target domain are in the same form as the source domain loss function;

[0068] For the transfer weight regularization term, Initialize parameters for the source domain adaptation layer;

[0069] , These are the weight coefficients for the target domain constraint loss in the m-th round and the weight coefficients for the transfer regularization term, which can be adjusted as needed.

[0070] In addition, such as Figure 3 The first iteration of the target domain Round dataset is , These are the locations to be predicted. The TL-Net weights are represented as follows: ,in Weights (shared parameter dimensions) for multi-source domain shared features extraction ), Adaptation layer weights for the target domain (adaptation parameter dimensions) ).

[0071] Step S150.4: Expert feedback-driven iterative optimization. Utilize the preliminary model to infer candidate layouts for fracturing scenarios. Through simulation verification using "equipment-environment interaction simulation and digital twin model analysis," and combined with expert system feedback (such as new constraints on physical boundaries and safety regulations), the feature weights of the target domain input data are adjusted and constraints are supplemented to generate an iterative dataset. This process of "adaptive fine-tuning - simulation verification - expert feedback - data iteration" is repeated until the layout meets the requirements of operational specifications and data verification, thus obtaining stable target domain model weights. and optimal layout scheme .

[0072] like Figure 3 In the middle, for the first Candidate layout schemes for digital twin scenarios obtained through round-robin reasoning Corrected feature weights are obtained through "device-environment interaction simulation" and "expert monitoring". Then the first The input features of the target domain are:

[0073] ;

[0074] New constraints based on expert feedback Integrate into the target domain constraint set:

[0075] ;

[0076] Repeat the iteration until the layout change is less than the threshold. :

[0077] ;

[0078] The final output is the optimal layout for the fracturing scenario that satisfies all constraints:

[0079] ;

[0080] The constraints in the above equation are: That is, the layout satisfies the set of constraints of all target domains. This is the final iteration round; where For the forward inference function of TL-Net, These are the shared weight parameters obtained from pre-training in the source domain. Indicates the process After rounds of iteration, the parameters of the target domain adaptation layer converge.

[0081] Step S150.5: Optimal layout design of the target area fracturing well site The coordinates of the equipment twin model and the environment twin model are mapped by parsing the digital twin model, and the digital twin fracturing well site adapted to the target domain is obtained by data visualization rendering based on the optimal layout scheme.

[0082] Through the hierarchical mechanism of TL-Net described above, which includes "multi-source pre-training - cross-domain weight transfer - target domain iterative fine-tuning - expert feedback enhancement - intelligent scenario adaptation", it is possible to achieve efficient transfer, reuse and adaptive optimization of multi-source scenario layout knowledge to the fracturing scenario domain. Even when target domain data is scarce or constraints are complex, it can still achieve the optimal layout location that meets the scenario's "operation specifications, physical boundaries and safety requirements" through iterative learning, thus providing support for the intelligent layout of fracturing well site equipment.

[0083] The above description is merely a specific exemplary case in this invention. The feature of the example is the cross-domain migration method of equipment that "forms a complete operation chain by linking multiple work areas and multiple types of equipment in complex scenarios". The implementation case takes into account spatial distribution characteristics, functional synergy attributes and work process correlation. Those skilled in the art can clearly understand that the scope of protection of this invention is not limited to this. The above-described cross-domain layout method is applicable to other case implementation processes with the characteristics of "multiple work areas and multiple equipment linkage".

[0084] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0085] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An intelligent cross-domain layout system for fracturing equipment based on digital twins and transfer learning, characterized in that, include: (1) Parametric model building module: Construct the basic elements required for scene layout design, namely, establish a digital twin model of multiple operation areas including equipment digital twin model and environment digital twin model. The equipment digital twin model is associated with the fusion and verification of fracturing operation feature attribute parameters, functional linkage constraints and management. The environment digital twin model includes fracturing operation external space boundary threshold and terrain adaptation parameters. (2) Scene Unit Interaction Module: Under the constraints of quantity matching, operation chain relationship and spatial constraints of natural environment, the equipment digital twin model calculates the multi-objective solution set of the optimal number of equipment and equipment placement in the scene with five factors as comprehensive weighted objectives: topology, information, cost, safety and efficiency. (3) Cross-domain scenario coupling module: Through the cross-domain scenario knowledge reuse mechanism, the historical layout experience data of fracturing operations in different source domains are called, and the unique constraint parameters of the target domain operation scenario are combined to solve the comprehensive weighted objective optimization function and generate a set of candidate layout schemes. The candidate layout schemes are then corrected for constraint weights and verified for feasibility through an expert system. After iterative optimization, the optimal layout scheme is output. The digital twin model analyzes the optimal layout scheme to complete the rendering of three-dimensional scene data, thereby realizing the intelligent cross-domain layout of the multi-unit collaborative cluster of fracturing.

2. The intelligent cross-domain layout system for fracturing equipment based on digital twins and transfer learning as described in claim 1, characterized in that: The constraints include equipment quantity matching constraints, natural environment constraints, and the correlation constraints between equipment quantity and placement location; the equipment quantity matching constraints satisfy the formula: In the formula, E represents the set of device types: , respectively representing energy equipment, monitoring equipment, living equipment, testing equipment, raw material equipment, fracturing equipment, operating equipment, and other equipment; the natural environmental constraints satisfy: In the formula Indicates whether device type e is placed at position i; the relationship between the number of devices and their placement positions satisfies: In the formula Let be the location storage coefficient for device e.

3. The intelligent cross-domain layout system for fracturing equipment based on digital twins and transfer learning according to claim 2, characterized in that: The number of devices satisfying all constraints and placement The weighted composite objective function is: In the formula The weighting coefficients for each objective dimension: topology, information, cost, security, and efficiency. For each objective, there is a normalization function.

4. A method for intelligent cross-domain layout of fracturing equipment based on digital twins and transfer learning, characterized in that: Includes the following steps: Step 1: Encode and establish the feature data and label classification of all physical entities, including the feature data of physical entities of operating equipment and physical entities of the operating environment; Step 2: Establish a digital twin model corresponding to the physical entity, and parameterize the constraints of the model, including limitations on the number of devices, terrain limitations, and the correlation and coupling constraints between various features; Step 3: Correct the constructed digital twin model and classify it to establish a 3D model library to provide basic elements for scene construction and layout; the user selects the scene construction method, including two methods: manual construction of single-source scenes and cross-domain scene migration construction, i.e., select step 4 or step 5; Step 4: Select the manual construction method for a single-source scenario of the digital twin model, based on physical boundary constraints. Weighted composite objective Based on the interaction relationships between devices and between devices and the environment, the layout and positioning coordinates of each model are set one by one, thereby completing the construction of all elements in the digital twin scene; if all interaction logic within the scene complies with the operating guidelines, the manual layout scheme design for the single-source scene is completed. The system analyzes and renders to obtain a digital twin fracturing well site; conversely, it is necessary to reconstruct the interaction between equipment and between equipment and the environment, and optimize the scene layout scheme until it meets the operating criteria. Step 5: Select the cross-domain scenario migration construction method for the digital twin model. The user inputs the requirements and constraints of the target domain scenario. The digital twin model extracts the feature knowledge of the target scenario. Through the cross-domain scenario knowledge reuse mechanism, candidate layout schemes are obtained by training and iterating from the historical layout data of fracturing operations in each source domain. The expert system adjusts the constraint weights of candidate layout schemes until they meet the operational criteria. This layout scheme is then considered the optimal layout scheme for the fracturing equipment in the target domain. The digital twin model is used to analyze and instantiate the coordinates of the twin model points, and the target fracturing well site is obtained through visualization rendering, realizing the intelligent cross-domain layout of fracturing equipment with multiple regions, multiple units, and functional linkage.

5. The intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning according to claim 4, characterized in that: The cross-domain scenario knowledge reuse mechanism includes: establishing a source domain-target domain feature mapping relationship library, which contains the operation area division rules, equipment function linkage logic, equipment layout space constraints, and layout target weight allocation rules for different fracturing scenarios. The source domain knowledge is selected and adapted through feature similarity matching and integrated into the target domain layout optimization process.

6. The intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning according to claim 5, characterized in that: Step 5 includes the following steps: Step 5-1: Multi-source knowledge extraction and encoding. The system performs knowledge encoding on the data from each source domain, associating them with "location-parameter-constraint". The datasets for each source domain are: In the formula For source domain The feature vector of the fracturing scene, Set of equipment layout points ( Represents the source domain No. In the first scenario (Layout coordinates of the device class) For the set of scenario constraints, The number of constraints. For the first The first source domain The sample at the th Under a given constraint, the model outputs a violation function that measures the degree to which the constraint is violated. Step 5-2: TL-Net pre-training and weight aggregation. TL-Net learns the general mapping knowledge of the multi-source domain dataset, minimizes the weighted sum of the multi-source domain layout prediction loss and the multi-source domain constraint violation loss, and obtains the pre-trained weights for each source domain. By aggregating the shared parameter portions of pre-trained weights from multiple source domains, cross-domain shared weights can be obtained. In the formula for A subset of shared parameters from the source domain. This is a weighted aggregation function; Step 5-3: Target domain transfer and iterative optimization, TL-Net will share weights Migrate to the target domain and iteratively optimize the model weight parameters of the adaptation layer: ； In the formula, , For the first Loss function for well site layout prediction in the target domain and loss function for constraint violation; For the transfer weight regularization term, corresponding to the first iteration of the target domain. Wheel scene data is To obtain the design of the well site layout to be predicted in the target domain. ; Step 5-4: Expert feedback-driven iterative optimization. The digital twin model is verified through simulation and expert system validation to obtain corrected feature weights. Update the input features of the target domain With the target domain constraint set TL-Net iteratively optimizes the output of the optimal layout scheme for the target domain. ; Step 5-5: Analyzing the optimal layout of the target domain in Step 5-4 using the digital twin model Data is used to instantiate and map the coordinates of the equipment twin model and the environment twin model, and then visualize and render to obtain a digital twin fracturing well site.

7. The intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning according to claim 6, characterized in that: The simulation verification module of the digital twin model is an equipment-environment interaction simulation. It realizes the pre-operation simulation of the physical fracturing well site through the digital twin scenario and detects potential equipment operation safety risks.

8. The intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning according to claim 6, characterized in that: The expert monitoring module of the digital twin model includes scene verification and correction suggestions from the expert subsystem and correction opinions input by the user. The user specifies whether the correction suggestions should be incorporated into the target domain dataset.

9. The intelligent cross-domain layout method for fracturing equipment based on digital twins and transfer learning according to claim 7, characterized in that: The proposed revisions include new constraints on the physical boundaries of fracturing well sites, safety standards, site design preferences, construction process priorities for fracturing operations, weighting of equipment maintenance convenience, and parameters related to the scope of human intervention.

10. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, cause the processor to implement the method described in any one of claims 4 to 9.

Images