Oil and gas exploration resource allocation method and device, storage medium and electronic equipment

Abstract

The application discloses an oil and gas exploration resource allocation method and device, a storage medium and an electronic device. The method comprises the following steps: acquiring first asset value data of a target block and second asset value data and spatial value data of a plurality of adjacent transaction blocks of the target block; performing feature extraction on the first asset value data to obtain an initial feature vector of the target block, and constructing a physical barrier network according to the first asset value data, the plurality of second asset value data and the plurality of spatial value data, performing feature extraction on the physical barrier network to obtain a spatial feature vector of the target block; analyzing the initial feature vector and the spatial feature vector of the target block by using a price prediction model to obtain a predicted transaction price interval; and determining an oil and gas exploration and development resource of the target block according to the predicted transaction price interval. The application solves the technical problem that related technologies cannot accurately evaluate the transaction price of a block, resulting in low configuration efficiency of oil and gas exploration resources.

Description

Technical Field

[0001] This application relates to the field of oil and gas resource assessment technology, and more specifically, to an oil and gas exploration resource allocation method and apparatus, storage medium, and electronic equipment. Background Technology

[0002] In the market-oriented bidding and auction process for oil and gas exploration rights, bidders generally rely on traditional valuation methods, such as spatial distance inverse weighting method and Kriging interpolation method, to predict the price of target blocks. The core logic is to use the price and volume information of historically traded blocks as a reference benchmark by attenuating weights according to the Euclidean distance on the ground.

[0003] However, these methods completely ignore the substantial inhibitory effect of the geological and physical constraints of oil and gas resources in underground space on value transmission. For example, a target block may be only 15 kilometers away from a high-priced transaction block, but a large thrust fault zone lies between them, preventing underground fluids from connecting. Traditional methods still overestimate its value due to "proximity," leading companies to misjudge risks and engage in blind bidding. Meanwhile, existing forecasting models mostly use static geological parameters (such as reserves and burial depth) and a single oil price level as inputs, ignoring the dynamic and spatial heterogeneity of oil and gas exploration rights value. This results in valuations that deviate significantly from the logic of actual market transactions, easily leading to major decision-making errors such as "overvaluation leading to wrong purchase" or "undervaluation leading to missed opportunities." Especially during periods of high oil prices, existing forecasting models cannot identify the asymmetric impact of "premium trends" and "dynamic costs," easily resulting in inflated or undervalued valuations. These shortcomings severely restrict the efficiency of allocating limited oil and gas exploration and development resources (such as capital, technology, and land quotas).

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides an oil and gas exploration resource allocation method and apparatus, storage medium, and electronic device to at least solve the technical problem that related technologies cannot accurately assess the transaction price of blocks, resulting in low allocation efficiency of oil and gas exploration resources.

[0006] According to one aspect of the embodiments of this application, a method for allocating oil and gas exploration resources is provided, comprising: acquiring first asset value data of a target block and second asset value data and spatial value data of each of a plurality of adjacent transaction blocks of the target block; extracting features from the first asset value data to obtain an initial feature vector of the target block, and constructing a physical barrier network based on the first asset value data, a plurality of second asset value data and a plurality of spatial value data, extracting features from the physical barrier network to obtain a spatial feature vector of the target block, wherein the physical barrier network is used to reflect the physical barrier status between the target block and each adjacent transaction block; analyzing the initial feature vector and spatial feature vector of the target block using a pre-trained price prediction model to obtain a predicted transaction price range for the target block; and determining the oil and gas exploration and development resources of the target block based on the predicted transaction price range.

[0007] According to another aspect of the embodiments of this application, an oil and gas exploration resource allocation device is also provided, comprising: a data acquisition module, used to acquire first asset value data of a target block and second asset value data and spatial value data of each of a plurality of adjacent transaction blocks of the target block; a data processing module, used to extract features from the first asset value data to obtain an initial feature vector of the target block, and to construct a physical barrier network based on the first asset value data, a plurality of second asset value data and a plurality of spatial value data, and to extract features from the physical barrier network to obtain a spatial feature vector of the target block, wherein the physical barrier network is used to reflect the physical barrier state between the target block and each adjacent transaction block; a data analysis module, used to analyze the initial feature vector and spatial feature vector of the target block using a pre-trained price prediction model to obtain a predicted transaction price range for the target block; and a determination module, used to determine the oil and gas exploration and development resources of the target block based on the predicted transaction price range.

[0008] In an exemplary embodiment, the aforementioned first asset value data includes at least: static geological data and dynamic cost data, wherein the static geological data includes at least one of the following: physical scalar parameters of the target block, geological exploration text, and spatial topology data; and the dynamic cost data includes at least: the crude oil price sequence and dynamic engineering material cost index of the target block during the historical period prior to the listing time. The aforementioned apparatus is used to extract features from the first asset value data to obtain an initial feature vector of the target block, comprising: semantically encoding the geological exploration text using a target language model to obtain a semantic feature vector, and concatenating it with the physical scalar parameters to obtain a static feature vector of the target block, wherein the target language model is fine-tuned using an oil and gas geological text corpus. The resulting oil and gas geology text corpus includes: various oil and gas geology texts and corresponding semantic feature vectors, and the types of oil and gas geology texts include at least one of the following: geological exploration texts, drilling engineering description texts, block structural evaluation texts, and mining rights assessment opinion texts; feature extraction is performed on the crude oil price series using a depth time series model to obtain the dynamic feature vector of the target block; the spatial topology data of the target block is multiplied by the dynamic engineering material cost index, and the obtained dynamic engineering material cost index is mapped through a fully connected layer to obtain the engineering cost constraint vector of the target block, wherein the spatial topology data includes at least: the burial depth of the target layer and the terrain slope; the initial feature vector of the target block is composed of the static feature vector, dynamic feature vector, and engineering cost constraint vector of the target block.

[0009] In an exemplary embodiment, the spatial value data includes at least: historical transaction data of adjacent transaction blocks and surface distance, connectivity status, and geological barrier parameters between the target block and the corresponding adjacent transaction blocks, wherein; the aforementioned apparatus is used to construct a physical barrier network based on the first asset value data, multiple second asset value data, and multiple spatial value data in the following manner: for each adjacent transaction block, feature extraction is performed on the second asset value data and spatial value data of the adjacent transaction block to obtain a comprehensive feature vector of the adjacent transaction block, wherein the comprehensive feature vector includes at least: a static feature vector, a dynamic feature vector, an engineering cost constraint vector, and a spatial feature vector; a geological barrier penalty coefficient between the target block and each adjacent transaction block is determined based on the connectivity status and geological barrier parameters, and a connection weight between the target block and each adjacent transaction block is determined based on the geological barrier penalty coefficient and the surface distance; the initial feature vector of the target block and the comprehensive feature vectors of each of the multiple adjacent transaction blocks are used as node attribute information of graph nodes, and the connection weights are used as edge weights of the edges between each pair of graph nodes to obtain the physical barrier network.

[0010] In an exemplary embodiment, the geological barrier parameters include at least one of the following: the fault displacement, burial depth range, strike, and blocking coefficient of the fault zone between the target block and adjacent transaction blocks, wherein the aforementioned apparatus is used to determine the geological barrier penalty coefficient between the target block and each adjacent transaction block in the following manner based on the connectivity status and the geological barrier parameters: for each adjacent transaction block, determining whether the connectivity status between the target block and the adjacent transaction block is a fully connected state; if the connectivity status between the target block and the adjacent transaction block is a fully connected state, determining the geological barrier penalty coefficient between the target block and the adjacent transaction block as a first value; if the connectivity status between the target block and the adjacent transaction block is not a fully connected state, determining the geological barrier penalty coefficient between the target block and the adjacent transaction block as a second value based on the geological barrier parameters between the target block and the adjacent transaction blocks and a preset parameter weight.

[0011] In an exemplary embodiment, the above-described apparatus is used to extract features from a physical barrier network to obtain a spatial feature vector of a target block, including: determining the attention score between the target block and each neighboring transaction block based on the node attribute information of the graph nodes corresponding to the target block in the physical barrier network and the node attribute information of the graph nodes corresponding to each neighboring transaction block; determining the attention weight between the target block and each neighboring transaction block by combining the edge weights of the edges between the graph nodes corresponding to the target block and the graph nodes corresponding to each neighboring transaction block in the physical barrier network; normalizing the attention weight between the target block and multiple neighboring transaction blocks to obtain the spatial overflow weight of the target block; and aggregating the spatial feature vectors of each of the multiple neighboring transaction blocks according to the spatial overflow weight to obtain the spatial feature vector of the target block.

[0012] In an exemplary embodiment, the above-described apparatus is used to determine the attention score between the target block and each neighboring transaction block based on the node attribute information of the graph nodes corresponding to the target block within the physical barrier network and the node attribute information of the graph nodes corresponding to each neighboring transaction block, by means of the following method: mapping the node attribute information of the graph nodes corresponding to the target block and the node attribute information of the graph nodes corresponding to each neighboring transaction block within the physical barrier network to a high-dimensional latent space using a preset feature weight matrix, thereby obtaining the latent feature vectors of the graph nodes corresponding to the target block and the latent feature vectors of the graph nodes corresponding to each neighboring transaction block, wherein the feature weight matrix includes: static feature vectors, dynamic feature vectors, engineering cost constraint vectors, and spatial feature vectors, each of which... The corresponding feature weight sub-matrices, the feature weight sub-matrices corresponding to the dynamic feature vector, the engineering cost constraint vector, and the spatial feature vector are all determined using the value range after Xavier initialization, and the feature weight sub-matrices corresponding to the static feature vector are determined using the value range after Xavier initialization and amplification. The attention weight parameters of the graph attention network are used to perform a weighted summation of the concatenated feature vectors of the hidden layer feature vectors of the graph nodes corresponding to the target block and the hidden layer feature vectors of the graph nodes corresponding to each neighboring transaction block, to obtain the attention score between the target block and each neighboring transaction block. The attention weight parameters include the attention weights corresponding to the static feature vector, the dynamic feature vector, the engineering cost constraint vector, and the spatial feature vector.

[0013] In an exemplary embodiment, the above-described apparatus is used to train a price prediction model in the following manner: determining an initial prediction model; acquiring multiple sets of training sample data, wherein each set of training sample data includes: a comprehensive feature vector of the transaction block and the actual transaction price; for each training batch in the iterative training process, inputting the comprehensive feature vector from each set of training sample data in the training batch into the prediction model obtained from the previous iteration training to obtain each predicted transaction price range output by the prediction model obtained from the previous iteration training, wherein the predicted transaction price range includes: the predicted transaction price corresponding to each of multiple quantiles; constructing a quantile regression loss function using each predicted transaction price range and the corresponding actual transaction price, and adjusting the model parameters of the prediction model obtained from the previous iteration training according to the quantile regression loss function.

[0014] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, which stores a computer program, wherein the computer-readable storage medium, when executed by a processor, performs the steps in any of the above method embodiments.

[0015] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the steps of any of the above method embodiments through the computer program.

[0016] In this embodiment, the first asset value data of the target block and the second asset value data and spatial value data of each of the multiple adjacent transaction blocks of the target block are obtained; features are extracted from the first asset value data to obtain the initial feature vector of the target block; a physical barrier network is constructed based on the first asset value data, multiple second asset value data, and multiple spatial value data; features are extracted from the physical barrier network to obtain the spatial feature vector of the target block, wherein the physical barrier network is used to reflect the physical barrier status between the target block and each adjacent transaction block; the initial feature vector and spatial feature vector of the target block are analyzed using a pre-trained price prediction model to obtain the predicted transaction price range of the target block; and the oil and gas exploration and development resources of the target block are determined based on the predicted transaction price range. The above technical solution adopts a multimodal feature fusion and intelligent price range prediction method based on geological barrier networks, which realizes accurate modeling of the transmission path of the true value of oil and gas exploration rights, eliminates the misleading valuation caused by surface distance and geological barrier interference, and achieves the goal of guiding oil and gas exploration and development resources to high-quality blocks with excellent geological conditions and strong economic feasibility. In this way, it solves the technical problem that related technologies cannot accurately evaluate the transaction price of blocks, resulting in low allocation efficiency of oil and gas exploration resources. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 This is a flowchart illustrating an optional block exploration method according to an embodiment of this application;

[0019] Figure 2 A schematic diagram illustrating the construction process of an optional physical barrier network according to an embodiment of this application;

[0020] Figure 3 A schematic diagram of an optional spatial feature vector extraction process according to an embodiment of this application;

[0021] Figure 4 A schematic diagram of the training process of an optional price prediction model according to an embodiment of this application;

[0022] Figure 5 This is a structural block diagram of an optional block exploration device according to an embodiment of this application;

[0023] Figure 6 This is a structural block diagram of an optional electronic device according to an embodiment of this application. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first," "second," etc., used in the specification, claims, and drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] To better understand the embodiments of this application, the following is a translation and explanation of some nouns or terms that appear in the description of the embodiments of this application:

[0027] A tectonic unit is a crustal structural unit formed by specific geological evolution processes within a regional tectonic context, possessing a relatively uniform structural morphology, stress field characteristics, deformation history, and hydrocarbon accumulation conditions. It is the basic unit in geology used to delineate and describe hydrocarbon accumulation spaces, typically defined based on characteristics such as fault systems, fold morphology, basement properties, stratigraphic response, and tectonic boundaries, possessing a clearly defined spatial extent and geological boundaries. Tectonic units can be classified into different levels, such as first-order, second-order, and third-order, according to their scale and hierarchy.

[0028] Xavier initialization (also known as Glorot initialization) is a well-known method for initializing weights in neural networks. By dynamically adjusting the initialization range of weights based on the number of input and output neurons, it keeps the variance of input and output consistent across each layer. This ensures that the amplitude of the signal does not increase or decrease exponentially with the number of layers during forward and backward propagation, aiming to solve the gradient vanishing and gradient exploding problems in deep network training.

[0029] The quantile loss function is a loss function used for regression tasks, particularly suitable for estimating specific quantiles of the target variable (such as the median, 90th percentile, etc.), unlike traditional mean prediction. Quantile loss does not treat all errors equally; instead, it assigns different penalty weights to positive and negative errors based on the selected quantile τ (0 < τ < 1): when the predicted value is lower than the true value (positive residual), the loss weight is τ; when the predicted value is higher than the true value (negative residual), the loss weight is 1-τ.

[0030] According to an embodiment of this application, an oil and gas exploration resource allocation method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0031] Figure 1 This is a flowchart illustrating an oil and gas exploration resource allocation method according to an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0032] Step S102: Obtain the first asset value data of the target block and the second asset value data and spatial value data of each of the multiple adjacent transaction blocks of the target block;

[0033] Step S104: Extract features from the first asset value data to obtain the initial feature vector of the target block, and construct a physical barrier network based on the first asset value data, multiple second asset value data and multiple spatial value data. Extract features from the physical barrier network to obtain the spatial feature vector of the target block. The physical barrier network is used to reflect the physical barrier status between the target block and each neighboring transaction block.

[0034] Step S106: Analyze the initial feature vector and spatial feature vector of the target block using the pre-trained price prediction model to obtain the predicted transaction price range of the target block.

[0035] Step S108: Determine the oil and gas exploration and development resources of the target block based on the predicted transaction price range.

[0036] Optionally, in this embodiment of the application, the target block is an oil and gas exploration right block that has not yet completed market transactions and is in the preparation stage of "bidding, auction and listing" or the stage of waiting for bidding, and has not yet formed an actual transaction price.

[0037] Optionally, in this embodiment, the aforementioned adjacent transaction block refers to a transaction block that is geographically adjacent to the target block, geologically comparable, and has already been traded in the market. It can be understood as a block located within a certain radius of the target block, which, according to GIS (Geographic Information System) spatial analysis, has a surface proximity relationship, shares similar structural units, and possesses complete transaction price records and geological evaluation data, and has been listed for sale.

[0038] Optionally, in this embodiment of the application, the aforementioned first asset value data is quantifiable asset data that exists before the target block is traded, including static geological data and dynamic cost data. Among them: Static geological data refers to structured and unstructured data that are relatively stable at the time of listing and do not change rapidly over time, describing the geological background conditions of the target block. It typically includes, but is not limited to: physical scalar parameters of the target block, geological exploration texts, and spatial topological data. Physical scalar parameters are numerical geological parameters with clear physical dimensions, such as block area (km²), target layer depth (m), porosity (%), permeability (mD), gas saturation (%), etc. Geological exploration texts are unstructured natural language descriptive reports, such as drilling reports, geological evaluation reports, structural analysis conclusions, etc. The content may include subjective interpretation information such as "microfracture development", "risk of fault leakage", "belonging to the Changning syncline structural zone"; Spatial topological data is GIS (Geographic Information System) data related to spatial geometry and connectivity, including: polygonal boundary coordinates of the target block, topographic elevation, spatial adjacency with surrounding blocks, and the presence of regional fault zones / aquitards and other physical barrier structures. Dynamic cost data refers to external macroeconomic and engineering cost factors that fluctuate over time and affect the economic viability of developing the target block. It reflects the market environment and development constraints at the time of listing and typically includes, but is not limited to, the crude oil price series and dynamic engineering material cost index of the target block in the historical period before the listing time. The crude oil price series is the time series data of international crude oil futures settlement prices for a certain historical period (such as the past 12 months) before the listing time, used to capture oil price trends, fluctuations and sentiment. The dynamic engineering material cost index is a comprehensive index reflecting the price changes of key materials required for oil and gas drilling (such as steel pipes, cement, fracturing sand, diesel, etc.) relative to the base period. For example, "current index = 120" means that the material cost has increased by 20% year-on-year, which is directly related to the pressure of rising development costs and is used to quantify "high consumption deduction".

[0039] Optionally, in this embodiment of the application, the aforementioned second asset value data is the asset value information actually reflected in the market transaction of the adjacent transaction blocks, which also includes: static geological data and dynamic cost data. The static geological data includes, but is not limited to: physical scalar parameters, geological exploration texts and spatial topology data of the adjacent transaction blocks; while the dynamic cost data includes, but is not limited to: crude oil price series and dynamic engineering material cost index of the adjacent transaction blocks in the historical time period before the corresponding listing time.

[0040] Optionally, in the embodiments of this application, the aforementioned spatial value data is topological and physical constraint information reflecting the geographical and geological spatial relationship between the target block and each adjacent transaction block, including but not limited to historical transaction data of adjacent transaction blocks, surface distance between the target block and adjacent transaction blocks, connectivity status, and geological barrier parameters. The aforementioned historical transaction data refers to transaction indicators that quantify the market value of adjacent transaction blocks at historical transaction points, including at least one of the following: historical total transaction price (i.e., the total amount of actual transactions in history), unit area transaction price (i.e., the total amount of actual transactions in history divided by the confirmed area of ​​the block), and transaction premium rate (i.e., the relative excess ratio between the total amount of actual transactions in history and its geological and economic assessment benchmark price); the aforementioned surface distance is the Euclidean straight-line distance between the surface center points of the target block and adjacent transaction blocks; connectivity status is a binary or hierarchical geological connectivity judgment obtained through GIS spatial intersection analysis, used to indicate whether the target block and adjacent transaction blocks have physical channels for oil and gas fluid migration or resource enrichment underground; the geological barrier parameter is a continuous penalty coefficient obtained through GIS spatial intersection analysis, used to quantify the physical inhibition strength of underground geological structures on value transmission.

[0041] Optionally, in the embodiments of this application, the above-mentioned physical barrier network is a graph structure topology model constructed based on the laws of oil and gas geology, used to reflect the physical barrier status between the target block and each adjacent transaction block.

[0042] Optionally, in the embodiments of this application, the above-mentioned oil and gas exploration and development resources are economic resources and strategic asset allocation bases derived from the predicted transaction price range and can be used to guide exploration and development decisions. They include the following dimensions: (1) Investment decision threshold resources, namely, different prices determined by the predicted transaction price range, such as the floor price percentile - the bidding floor that cannot be lowered, the fair center - the reference for reasonable bidding, and the red line warning - the critical point of loss risk; (2) Risk hedging resource allocation, namely, engineering cost deduction items and geological risk warnings. Relevant enterprises can decide whether to invest in more expensive well types (such as ultra-deep wells and horizontal wells), whether to use anti-leakage technology, and whether to increase the scale of fracturing based on this resource allocation; (3) Strategic block priority ranking resources, namely, when relevant enterprises know the predicted transaction price range of multiple unsold target blocks, they can prioritize bidding for blocks with "low engineering deductions" according to their budget to achieve the optimal allocation of resources in space.

[0043] For example, step S102 above can be understood as follows: First, spatial topological data is obtained by calling the geographic information system interface, 3D geological model, or satellite remote sensing data; geological exploration text is obtained from publicly available resources such as oilfield company archives and geological databases; and physical scalar parameters are obtained through core analysis, well logging curves, and seismic interpretation reports. Next, the spatial topological data, geological exploration text, and physical scalar parameters are subjected to a series of preprocessing operations (such as noise reduction, sentence segmentation, word segmentation, stop word filtering, coordinate system 1, topology verification, GIS spatial intersection analysis, outlier removal, missing value filling, normalization, and stationarity verification) to form static data in the first asset value data, reflecting the relatively stable long-term geological characteristics of the target block. Second, the crude oil price series of the target block during the historical period before the listing time is obtained from futures exchanges, energy information platforms, or mining rights trading databases; and the dynamic engineering material cost index is obtained from commodity exchanges. Then, the target block's data before the listing time is processed... After a series of preprocessing operations on the historical crude oil price series and dynamic engineering material cost index, dynamic data is formed in the first asset value data. Finally, the second asset value data of each adjacent transaction block is obtained from the public transaction records. Its structure is consistent with the first asset value data of the target block, including physical scalar parameters, geological exploration text, spatial topology data, crude oil price series at the time of transaction, and dynamic engineering material cost index at the time of transaction. This data is used to compare and analyze the asset value of the target block. Spatial value data is obtained through GIS system, geological structure map, pipeline distribution map or oilfield development planning documents to reflect the impact of the spatial relationship between adjacent blocks and the target block on the value.

[0044] For example, step S104 above can be understood as follows: vectorizing and mapping various types of data in the first asset value data of the target block, and concatenating the feature vectors of various types of data to form a multidimensional initial feature vector of the target block; then, based on the second asset value data and spatial value data of multiple neighboring transaction blocks, constructing a physical barrier network with the target block and multiple neighboring transaction blocks of the target block as nodes and the connectivity state between each block as edges; finally, considering that each neighboring transaction block has a different degree of influence on the predicted transaction price of the target block, calculating the attention weight of each node attribute in the physical barrier network, and aggregating the attention weights of all nodes as the spatial feature vector of the target block.

[0045] For example, step S106 above can be understood as follows: by dimensionally aligning and concatenating the initial feature vector of the target block with the spatial feature vector to form a high-dimensional fusion feature vector, which is used as the input of the pre-trained price prediction model. The price prediction model uses the actual transaction price of historical transaction blocks as the label, minimizes the loss function between the predicted transaction price and the actual transaction price, and outputs the predicted transaction price range.

[0046] For example, step S108 above can be understood as follows: To achieve profitability, the predicted transaction price range is used as the potential revenue range. Further, dynamic engineering costs, geological risks, and market fluctuations during the development cycle are considered to quantify the expected net revenue of the target block, thereby determining the scale of investment in oil and gas exploration and development resources for that target block. If the calculated upper limit of the predicted transaction price range is insufficient to support the scale of investment in oil and gas exploration and development resources, the target block may be deemed to lack economic exploitation value.

[0047] Through the embodiments of this application, a multimodal feature fusion and intelligent price range prediction method based on geological barrier networks is adopted to achieve accurate modeling of the transmission path of the true value of oil and gas exploration rights. This eliminates the misleading valuation caused by surface distance and geological barrier interference, and achieves the goal of guiding oil and gas exploration and development resources to be efficiently allocated to high-quality blocks with excellent geological conditions and strong economic feasibility. This solves the technical problem that related technologies cannot accurately evaluate the transaction price of blocks, resulting in low allocation efficiency of oil and gas exploration resources.

[0048] As an optional approach, feature extraction is performed on the first asset value data to obtain the initial feature vector of the target block, including:

[0049] Step S21: Semantically encode the geological exploration text using the target language model to obtain a semantic feature vector, and concatenate it with physical scalar parameters to obtain the static feature vector of the target block. The target language model is obtained by fine-tuning the oil and gas geological text corpus, which includes: various oil and gas geological texts and their corresponding semantic feature vectors. The types of oil and gas geological texts include at least one of the following: geological exploration text, drilling engineering description text, block structure evaluation text, and mining rights assessment opinion text.

[0050] Step S22: Use a deep time series model to extract features from the crude oil price series to obtain the dynamic feature vector of the target block;

[0051] Step S23: Multiply the spatial topology data of the target block with the dynamic engineering material cost index, and map the obtained dynamic engineering material cost index through a fully connected layer to obtain the engineering cost constraint vector of the target block. The spatial topology data includes at least: the target layer burial depth and the terrain slope.

[0052] Step S24: The initial feature vector of the target block is composed of the static feature vector, dynamic feature vector, and engineering cost constraint vector of the target block.

[0053] Optionally, in the embodiments of this application, the target language model mentioned above is a special model for geological oil and gas semantic recognition obtained by performing domain adaptation training and fine-tuning on the basis of a general language model through an oil and gas geological text corpus. It includes, but is not limited to, BERT (Bidirectional Encoder Representations from Transformers), RoBERTa (Robustly Optimized BERT Pretraining Approach), GPT (Generative Pre-trained Transformer), etc. The text corpus mentioned above includes: various oil and gas geological texts and corresponding semantic feature vectors, and the types of the oil and gas geological texts include at least one of the following: geological exploration texts, drilling engineering description texts, block structure evaluation texts, and mineral rights assessment opinion texts.

[0054] Optionally, in the embodiments of this application, the above-mentioned deep time series model is a model built on a deep learning architecture for processing time series data (such as crude oil prices, production, inventory, etc.), which includes, but is not limited to, LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), and TCN (Temporal Convolutional Network), used to capture dynamic features such as long-term dependence, periodicity, and trend in the data through nonlinear transformation.

[0055] Optionally, in the embodiments of this application, the above-mentioned spatial topology data is a set of parameters that reflect the spatial geometric characteristics of the target block, extracted based on geographic information systems or three-dimensional geological modeling technology, including at least: the depth of the target layer, the terrain slope, etc.

[0056] Optionally, in the embodiments of this application, the aforementioned dynamic engineering material cost index is a set of material cost weights that vary with time and space, constructed based on market price fluctuations, changes in engineering demand, and specific conditions of the block (such as geological complexity and construction difficulty). It includes, but is not limited to, steel cost index, cement cost index, fuel cost index, and drilling tool and pipeline cost index.

[0057] For example, step S21 above can be understood as follows: The geological exploration text is input into the target language model. The target language model identifies core information related to geological characteristics (such as reservoir type, structural features, oil content, etc.) from the text and understands the meaning of terms through context (e.g., low permeability may refer to absolute permeability less than a certain threshold, or it may refer to low relative permeability). Terms with the same meaning are mapped to a unified numerical range. The geological exploration text is semantically encoded into a fixed-dimensional semantic feature vector (e.g., 128-dimensional, 256-dimensional) to facilitate subsequent mathematical operations. For example, if the input geological exploration text is "This block is an anticline structure, with a sandstone reservoir thickness of approximately 20 meters and an oil saturation of 65%", the target language model outputs a 128-dimensional semantic feature vector. Each dimension of the vector may implicitly contain information such as structural type, reservoir lithology, formation thickness, and oil content. Physical scalar parameters exist in numerical form, but each scalar parameter has a different scale. They can be normalized and then directly concatenated with the semantic feature vector according to the data dimension to obtain the static feature vector of the target block, which is used to describe the inherent geological properties of the target block that do not change over time.

[0058] For example, step S22 above can be understood as: conducting in-depth analysis of the crude oil price sequence of the target block in the historical period before the listing time through a deep time series model, capturing the complex patterns of crude oil price changes over time (such as long-term trends, short-term fluctuations, and the impact of sudden events), and encoding this dynamic information into a fixed-dimensional dynamic feature vector. The dynamic feature vector value is updated as the time window slides (such as updated once a day), reflecting the impact of crude oil price factors changing over time on the target block.

[0059] For example, step S23 above can be understood as follows: First, the spatial topology data of the target block is multiplied by the dynamic engineering material cost index to simulate the effect of spatial topology factors on the increase or decrease of engineering material costs. For example, if the transportation distance of block B is twice that of block A, and the current steel cost index is 1.2 (an increase of 20%), then the weighted cost of block B is 2 × 1.2 = 2.4 (the dynamic engineering material cost index is magnified by 2 times). Then, the weighted dynamic engineering material cost index is input into the fully connected layer for nonlinear transformation, transforming the coupling information of the original spatial topology factors and the dynamic engineering material cost index into higher-dimensional and fixed-dimensional engineering cost constraint features. Each dimension can represent an engineering cost constraint condition, used to quantify the cost constraints in the development process of the target block.

[0060] For example, step S24 above can be understood as: concatenating the three types of feature vectors—static feature vector, dynamic feature vector, and engineering cost constraint vector—of the target block in a fixed order to form a higher-dimensional comprehensive vector, as shown in the following formula: In the formula, This represents the composite vector of target block i. This represents the static feature vector of target block i. This represents the dynamic feature vector of target block i. This represents the engineering cost constraint vector of target block i, avoiding information loss due to dimensionality reduction or fusion.

[0061] This application utilizes a target language model to extract semantic feature vectors representing heterogeneity, risk level, and resource potential from oil and gas geological texts, overcoming the limitations of traditional models in shallow keyword matching of unstructured texts. Simultaneously, a deep time-series model is employed to capture the nonlinear trends and periodic fluctuations of international crude oil prices, generating dynamic feature vectors reflecting market sentiment at the time of listing. Furthermore, spatial topological constraints such as target layer depth and terrain slope are explicitly cross-multiplied with a dynamic engineering cost index. Through nonlinear mapping of fully connected layers, a quantifiable engineering cost constraint vector representing the combined effects of mining difficulty and cost is constructed, achieving a joint penalty mechanism of geological difficulty plus material cost. Finally, the above three types of feature vectors are concatenated into an initial feature vector, forming a three-dimensional value encoding system that integrates geological essence cognition, market trend perception, and engineering cost penalty, improving the accuracy and comprehensiveness of the target block's value representation.

[0062] As an optional approach, based on the first asset value data, multiple second asset value data, and multiple spatial value data, the following method is used: Figure 2 The process shown constructs a physical barrier network, including:

[0063] Step S31: For each adjacent transaction block, feature extraction is performed on the second asset value data and spatial value data of the adjacent transaction block to obtain the comprehensive feature vector of the adjacent transaction block. The comprehensive feature vector includes at least: static feature vector, dynamic feature vector, engineering cost constraint vector and spatial feature vector.

[0064] Step S32: Determine the geological barrier penalty coefficient between the target block and each adjacent transaction block based on the connectivity status and geological barrier parameters, and determine the connection weight between the target block and each adjacent transaction block based on the geological barrier penalty coefficient and the surface distance.

[0065] Step S33: The initial feature vector of the target block and the comprehensive feature vectors of multiple neighboring transaction blocks are used as the node attribute information of the graph nodes, and the connection weights are used as the edge weights of the edges between each pair of graph nodes to obtain the physical barrier network.

[0066] Optionally, the aforementioned connectivity refers to whether there is a physical channel in the underground geological structure between the target block and the adjacent transaction block that allows for the free migration of oil and gas fluids or the conduction of energy. This includes, but is not limited to, whether the target block and the adjacent transaction block belong to the same structural unit, whether they are separated by regional fault zones or aquitards, whether they share the same reservoir system, whether they are in the same hydrocarbon accumulation system, or whether they are effectively blocked by structural boundaries such as thrust faults, strike-slip faults, and regional unconformities.

[0067] Optionally, the above-mentioned geological barrier parameters are numerical indicators that quantify the intensity of the inhibition of the spatial transmission of oil and gas value by underground geological structures. They include, but are not limited to, the fault displacement, burial depth range, strike, and sealing coefficient of the fault zone between the target block and the adjacent transaction block.

[0068] For example, step S31 above can be understood as follows: First, semantically encode unstructured text descriptions such as geological exploration texts and structural evaluation texts of neighboring blocks based on the target language model, and output static feature vectors to reflect their geological endowment and engineering risks; second, use a deep time series model to perform time series modeling on the historical crude oil price series before and after the transaction time of neighboring blocks, extract dynamic feature vectors to represent the market environment and sentiment cycle; third, perform multiplication and cross operations on the spatial topology data of the block and the dynamic engineering material cost index at the transaction time, and generate an engineering cost constraint vector through a fully connected layer nonlinear mapping to quantify the joint impact of mining difficulty and cost pressure; finally, calculate the geometric spatial features of neighboring blocks based on GIS spatial coordinates, construct spatial feature vectors to represent their geographical location attributes; finally, concatenate the above four types of vectors in a fixed dimension order to form a comprehensive feature vector with a unified dimension, as shown in the following formula: In the formula, This represents the combined feature vector of neighboring block j. This represents the static feature vector of neighboring block j. This represents the dynamic feature vector of neighboring block j. This represents the engineering cost constraint vector of neighboring block j. This represents the spatial feature vector of neighboring block j, realizing the structured fusion of four-dimensional information: geological essence, market dynamics, engineering cost, and spatial location.

[0069] For example, step S32 above can be understood as follows: The connectivity status and geological barrier parameters between the target block and each adjacent transaction block are obtained through GIS intersection analysis. Combined with the preset fault zone classification and tectonic unit attribution relationship, the geological barrier penalty coefficient between the target block and each adjacent transaction block is dynamically determined through rule mapping and a hierarchical judgment mechanism. The connection weight between the target block and each adjacent transaction block is determined based on the geological barrier penalty coefficient and the surface distance, using the following specific formula: In the formula, This represents the connection weight between target block i and its neighboring transaction block j. This represents the geological barrier penalty coefficient between target block i and adjacent traded block j, with a value range of [value range missing]. , This represents the surface distance between target block i and the adjacent traded block j.

[0070] For example, step S33 above can be understood as follows: mapping the target block and multiple neighboring transaction blocks to nodes in a graph network, using the initial feature vector of the target block and the comprehensive feature vector of multiple neighboring transaction blocks as the attribute information of the nodes, wherein the feature vector of the target block includes geological parameters, macro market environment and unknown transaction value fields, and the feature vector of the neighboring transaction blocks includes complete real historical transaction features; subsequently, based on the connection weights calculated in step S32, a directed or undirected edge is established between each pair of nodes, and the weight is used as the unique topological attribute of the edge (i.e., edge weight), thus constructing a physically meaningful physical barrier network.

[0071] In the embodiments of this application, high-value blocks isolated by underground geological barriers such as fault zones and tectonic boundaries are mistakenly transmitted to target blocks, leading to inflated valuations. This solution accurately identifies the geological connectivity and geological barrier states through GIS spatial intersection analysis, and dynamically calculates the geological barrier penalty coefficient (range [0.01, 1.0]) based on the above states. This coefficient is then used as a hard constraint to incorporate into the edge weight calculation of the graph network to construct a physical barrier network. This ensures that the value transmission between adjacent blocks only occurs between geologically connected blocks, fundamentally eliminating the misjudgment that spatial proximity equals value proximity.

[0072] As an optional approach, the geological barrier penalty coefficient between the target block and each neighboring transaction block is determined based on connectivity and geological barrier parameters, including:

[0073] Step S41: For each adjacent transaction block, determine whether the connectivity between the target block and the adjacent transaction blocks is fully connected.

[0074] Step S42: If the connectivity between the target block and the adjacent transaction block is fully connected, determine the geological barrier penalty coefficient between the target block and the adjacent transaction block as the first value.

[0075] Step S43: If the connectivity between the target block and the adjacent transaction blocks is not fully connected, the geological barrier penalty coefficient between the target block and the adjacent transaction blocks is determined as a second value based on the geological barrier parameter between the target block and the adjacent transaction blocks and the preset parameter weight, wherein the second value is less than the first value.

[0076] For example, step S41 above can be understood as: judging the connectivity status between the target block and the adjacent transaction block based on the GIS spatial intersection analysis results, including judging whether there is a Class I or Class II regional fault zone between the target block and the adjacent transaction block, and whether the target block and the adjacent transaction block are in the same structural unit.

[0077] For example, step S42 above can be understood as follows: if there is no Class I or Class II regional fault zone between the target block and the adjacent transaction block, and the target block and the adjacent transaction block are located in the same Class III structural unit, then it is determined that the target block and the adjacent transaction block are completely connected, that is, they are not isolated by physical barriers such as regional faults, aquitards, structural boundaries or lithological barriers, and share the same reservoir system in the underground geological structure, possessing fluid connectivity and structural continuity in the process of oil and gas migration, accumulation and hydrocarbon accumulation. Therefore, its geological barrier penalty coefficient is set to the first value;

[0078] For example, step S43 above can be understood as follows: if there is a small or large fault zone between the target block and the adjacent transaction block, or if they belong to different third-level units, then it is determined that the target block and the adjacent transaction block are not fully connected. A second value within a preset range can be output based on the geological barrier parameter by a pre-trained deep learning model (such as a lightweight classification network or a rule-embedded graph neural network), serving as the geological barrier penalty coefficient between the target block and the adjacent transaction block. The smaller the second value, the larger the fault zone between the target block and the adjacent transaction block, the closer the fault distance, and the weaker the connectivity between the adjacent transaction block and the target block.

[0079] Through the embodiments of this application, multi-dimensional geological barrier parameters (fault displacement, burial depth range, strike, and sealing coefficient) are introduced, and a hierarchical assignment mechanism for geological barrier penalty coefficients based on connectivity and barrier parameters is established. This achieves a leap from traditional empirical zoning judgments (such as binary division of "connectivity" or "barrier") to fine-grained quantitative modeling driven by geophysical processes. It transforms the qualitative judgments that the oil and gas industry has long relied on expert experience into calculable, calibrable, and reproducible quantitative geological barrier penalty coefficients, greatly improving the model's ability to distinguish complex geological scenarios and its engineering applicability.

[0080] As an optional solution, according to... Figure 3 The process shown extracts features from the physical barrier network to obtain the spatial feature vector of the target block, including:

[0081] Step S51: Based on the node attribute information of the graph node corresponding to the target block in the physical barrier network and the node attribute information of the graph node corresponding to each neighboring transaction block, determine the attention score between the target block and each neighboring transaction block, and combine the edge weights of the edges between the graph node corresponding to the target block and the graph node corresponding to each neighboring transaction block in the physical barrier network to determine the attention weight between the target block and each neighboring transaction block.

[0082] Step S52: Normalize the attention weights between the target block and multiple neighboring transaction blocks to obtain the spatial overflow weight of the target block.

[0083] Step S53: Aggregate the spatial feature vectors of multiple neighboring transaction blocks according to the spatial overflow weight to obtain the spatial feature vector of the target block.

[0084] For example, step S51 above can be understood as: obtaining the node attribute information of the graph nodes corresponding to the target block within the physical barrier network. Node attribute information of graph nodes corresponding to each adjacent transaction block The hidden feature vector is mapped to d, and this hidden feature vector is passed through an activation function layer to determine the attention score between the target block and each neighboring transaction block. The edge weights of the edges between the graph nodes corresponding to the target block and the graph nodes corresponding to each neighboring transaction block within the physical barrier network are then calculated. Multiplying the attention score above, the attention weight between the target block and each neighboring transaction block is determined, as shown in the following formula: In the formula, This represents the attention score between the target block and each neighboring block. This represents the attention weight between the target block and each neighboring transaction block. The attention weight combines geological similarity (i.e., attention score) with physical connectivity. Even if the geological features between the target block and each neighboring transaction block are very similar, the attention weight value will be limited if they are not connected (i.e., the edge weight is very low).

[0085] For example, step S52 above can be understood as: normalizing the attention weights between the target block and multiple neighboring transaction blocks to ensure that the sum of all weights is 1, thus obtaining the spatial overflow weight of the target block, as shown in the following formula: In the formula, This represents the spatial overflow weight of the neighboring transaction block j to the target block i, and N represents the total number of neighboring transaction blocks to the target block i. This represents the normalization layer, which is used to transform the attention weights between the target block and each neighboring transaction block into a set of probability weight distributions that sum to 1.

[0086] For example, step S53 above can be understood as: linearly weighting and fusing the spatial feature vectors of each neighboring historical transaction block according to their value transmission strength to the target block (i.e., spatial spillover weight), thereby generating the spatial feature vector of the target block in the geological-economic space, as shown in the following formula: The spatial feature vector of the target block retains only historical value information with real transmission capability after physical gating, providing a high-fidelity and interpretable input representation for subsequent multimodal fusion and quantile prediction.

[0087] Through the embodiments of this application, by inputting the node attributes and edge weights of the target block and each of the neighboring transaction blocks into the graph attention network, the attention score not only reflects the feature similarity between different neighboring blocks, but also forces them to comply with the physical constraints of geological accessibility. Even if two blocks are highly similar in reservoir features, if they are isolated by strong fault zones, their attention weights are still significantly suppressed, thus more accurately describing the spatial feature vector of the target block.

[0088] As an optional approach, based on the node attribute information of the graph nodes corresponding to the target block within the physical barrier network and the node attribute information of the graph nodes corresponding to each neighboring transaction block, the attention score between the target block and each neighboring transaction block is determined, including:

[0089] Step S61: Using a preset feature weight matrix, the node attribute information of the graph nodes corresponding to the target block in the physical barrier network and the node attribute information of the graph nodes corresponding to each neighboring transaction block are mapped to a high-dimensional latent space to obtain the latent feature vectors of the graph nodes corresponding to the target block and the latent feature vectors of the graph nodes corresponding to each neighboring transaction block. The feature weight matrix includes: feature weight sub-matrices corresponding to the static feature vector, dynamic feature vector, engineering cost constraint vector and spatial feature vector. The feature weight sub-matrices corresponding to the dynamic feature vector, engineering cost constraint vector and spatial feature vector are all determined by the value range after Xavier initialization. The feature weight sub-matrices corresponding to the static feature vector are determined by the value range after Xavier initialization and amplification.

[0090] Step S62: The attention weight parameters of the graph attention network are used to perform a weighted summation of the concatenated feature vectors of the hidden feature vectors of the graph nodes corresponding to the target block and the hidden feature vectors of the graph nodes corresponding to each neighboring transaction block, so as to obtain the attention score between the target block and each neighboring transaction block. The attention weight parameters include the attention weights corresponding to the static feature vector, dynamic feature vector, engineering cost constraint vector and spatial feature vector.

[0091] Optionally, the aforementioned feature weight matrix is ​​a learnable linear transformation parameter matrix used to project the node attribute vectors of each node in the physical barrier network to a high-dimensional latent space. It includes, but is not limited to, the sub-weight matrices corresponding to the static feature vector, dynamic feature vector, engineering cost constraint vector, and spatial feature vector. The sub-matrix corresponding to the static feature vector is initialized by amplifying the range by a factor of M after Xavier initialization to strengthen the dominant role of geological background parameters in value assessment. The sub-matrices corresponding to the dynamic feature vector, engineering cost constraint vector, and spatial feature vector are all initialized using the value range of standard Xavier initialization to ensure that temporal information, engineering cost perturbation, and geometric relationships have stable gradient propagation capabilities in the early stage of training.

[0092] For example, step S61 above can be understood as follows: using a k-dimensional feature weight matrix in the graph attention network, the node attribute information of the graph node corresponding to the target block and the node attribute information of the graph node corresponding to each neighboring transaction block are linearly transformed and mapped to a k-dimensional high-dimensional latent space to obtain the hidden layer feature vector of the graph node corresponding to the target block and the hidden layer feature vector of the graph node corresponding to each neighboring transaction block, as shown in the following formula: , In the formula, and Let i and j represent the initial feature vectors of the target block i and the neighboring transaction block j, respectively. Let represent a k-dimensional weight matrix, where and , The aforementioned hidden layer feature vectors retain the original information of various feature vectors of the target block and each neighboring transaction block.

[0093] For example, step S62 above can be understood as: extracting the hidden feature vectors of the graph nodes corresponding to the target blocks within the physical barrier network. Hidden feature vectors of graph nodes corresponding to each neighboring transaction block An activation function layer is used to determine the attention score between the target block and each neighboring transaction block according to the following formula: In the formula, The attention score represents the relationship between target block i and neighboring traded blocks j, used to characterize the geological similarity between the target block and its neighboring traded blocks. This represents a non-linear activation function used to preserve weak negative gradients and enhance the model's ability to distinguish between low-similarity node pairs. This represents an attention weight parameter to be trained, used to perform a weighted summation of the node attribute information of the target block i and the neighboring transaction blocks j after splicing, and to determine which dimensions of features are more important for value transmission.

[0094] Through the embodiments of this application, by structurally expanding the initialization range of the weight matrix corresponding to the static feature vector, the model is forced to assign higher initial activation intensity to the geological features in the early stages of training. At the same time, the weight matrices corresponding to the dynamic features, engineering costs, and spatial features are still initialized using the standard Xavier method to maintain their relatively balanced learning ability, thereby constructing a priori feature importance that is clear in its hierarchy during the model initialization stage.

[0095] As an optional approach, the training process for the price prediction model is as follows: Figure 4 As shown, it includes:

[0096] Step S71: Determine the initial prediction model;

[0097] Step S72: Obtain multiple sets of training sample data, wherein each set of training sample data includes: the comprehensive feature vector of the transaction block and the actual transaction price;

[0098] Step S73: For each training batch in the iterative training process, input the comprehensive feature vector from each group of training sample data in the training batch into the prediction model obtained from the previous iteration training to obtain the predicted transaction price ranges output by the prediction model obtained from the previous iteration training. The predicted transaction price ranges include the predicted transaction prices corresponding to multiple quantiles. Construct a quantile regression loss function using each predicted transaction price range and the corresponding actual transaction price, and adjust the model parameters of the prediction model obtained from the previous iteration training according to the quantile regression loss function.

[0099] For example, step S71 above can be understood as follows: the initial prediction model can be a trainable neural network prediction model (such as multilayer perceptron, fully connected neural network, and multi-head attention mechanism, etc.), which uses a fully connected feedforward neural network as the backbone prediction structure, contains several hidden layers, each layer is composed of linear transformation and nonlinear activation functions (such as ReLU, Sigmoid, and Tanh, etc.), used to realize nonlinear combination of input features and high-order interaction modeling; the output layer is a multi-head linear unit, the number of output heads is equal to the preset number of quantiles, each output head independently outputs the price prediction value of the corresponding quantile, realizing the synchronous output of the price range in a single forward propagation, and the initial prediction model is constructed by combining the above modules.

[0100] For example, step S72 above can be understood as follows: taking the comprehensive feature vector of the transaction blocks in the historical time period as input data, obtaining the actual total transaction price (unit: RMB) or unit area transaction price of the above transaction blocks from the public database of the mining rights trading platform as data labels, performing a series of preprocessing operations on the above data labels (such as data cleaning, missing value filling and normalization processing, etc.), and forming multiple sets of training sample data by corresponding each input data with its data label one by one.

[0101] For example, step S73 above can be understood as follows: the training process of the initial prediction model is implemented based on an end-to-end supervised learning framework guided by physical mechanisms. Furthermore, during each round of iterative training, each set of training sample data can be analyzed as follows:

[0102] Step 1: Calculate the spatial overflow weight of the traded block, and then sum the historical transaction prices of neighboring traded blocks around the traded block based on this weight to obtain the spatial benchmark value of the traded block. The formula for calculating the spatial benchmark value can be written as: In the formula, Indicates spatial benchmark value, This represents the standardized transaction price per unit area of ​​the historically adjacent transaction block y (unit: 10,000 yuan / km²). This represents the area of ​​the transaction block x that has been confirmed. This indicates the space overflow weight of the transaction block z. This represents the set of historical neighboring blocks y surrounding block x.

[0103] Step 2: Considering the cyclical rise and fall of crude oil prices in the international crude oil trading market, based on the spatial benchmark value, the time series data of international crude oil prices within a preset time period before the listing time of the transaction block is extracted and substituted into the oil price trend deviation formula: In the formula, This indicates the degree of deviation in oil price trends, determining whether the current crude oil market is in a period of market fluctuation characterized by appreciation (or loss). This represents the average international crude oil price over a preset period before the listed block. This represents the average international crude oil price over the three years prior to the listing of the traded block. Based on the oil price trend deviation, the pre-set macroeconomic market impact coefficient, and the spatial benchmark value, the macroeconomic value-added item is calculated using the following formula: In the formula, This represents the macroeconomic value-added component, reflecting the long-term market trend within that cycle. This represents the macroeconomic market impact coefficient, used to control the extent to which oil price sentiment affects valuations. By introducing a macroeconomic value-added term, it can prevent the initial forecasting model from experiencing block valuation deviations due to momentary disturbances such as geopolitical events and speculative trading, ensuring that price forecasts are always anchored to industry cycle patterns.

[0104] Step 3: Utilize the engineering cost constraints of the traded blocks to deduct the additional investment losses incurred due to high burial depth, high engineering material costs, and high mining difficulty. Determine the engineering cost deduction constraints according to the following formula: In the formula, This represents the engineering cost deduction constraint, used to characterize the impact of engineering difficulty on the value of the spatial benchmark. This indicates the normalized depth of the target layer in the transaction block. The greater the depth, the closer the value is to 1 (the more is deducted). , This represents the normalized terrain slope of the traded block. The steeper the slope, the closer the value is to 1. , The code represents the geological risk level of the transaction block. It is divided according to the degree of fracture development and leakage risk. High-risk blocks can be assigned a value of 1.0, and ordinary blocks can be assigned a value of 0.5. This represents the dynamic engineering material cost index. It deducts engineering costs from the constraint. In conjunction with spatial benchmark values, the fully connected layer with a Sigmoid activation function is input to calculate the engineering cost deduction item, as shown in the following formula: , In the formula, This represents a deduction from project costs, with a value range of [0, 0.6]. These represent the learnable weights and biases of the fully connected layer, respectively. By introducing an engineering cost deduction term, the mining difficulty of the target block and the pressure of rising dynamic engineering costs are transformed into a systematic discount on the spatial benchmark value, thereby avoiding inflated valuations and investment losses due to neglecting development economics.

[0105] The fourth step is to aggregate the aforementioned spatial benchmark value, macro-value-added items, and engineering cost deduction items to form the final comprehensive feature vector, as shown in the following formula: In the formula, This represents the comprehensive feature vector. Inputting this comprehensive feature vector into the initial prediction model yields the initial predicted price of the traded block.

[0106] Fifth, using the initial price prediction range and the actual historical transaction prices of each transaction block, construct the quantile loss function: In the formula, This indicates the preset quantile. This indicates the actual price tag of the block in question. This represents the initial predicted price of the transaction block output by the initial prediction model. Different quantiles correspond to different penalty weights, forcing the initial prediction model not only to predict the central price value, but also to more precisely fit the lower limit (safe floor price) and upper limit (risk red line) of the price distribution. Using the upper and lower limits as the boundaries of the transaction price, the predicted transaction price range is obtained.

[0107] The sixth step is to calculate the gradient of the quantile target loss function with respect to the model parameters of the prediction model obtained from the previous iteration of training, adjust the model parameters according to the gradient and the preset learning rate, and update all model parameters in the prediction model obtained from the previous iteration of training simultaneously.

[0108] Step 7: Determine whether the prediction model obtained from the previous round of iteration training meets the preset iteration termination conditions (the number of iterations reaches the preset threshold or the loss function reaches the minimum value). If it meets the conditions, proceed to step 6; otherwise, return to step 1 to start the next round of iteration training.

[0109] The eighth step is to use the prediction model obtained in the current iteration training process as the completed price prediction model, and obtain the predicted transaction price range calculated in this iteration.

[0110] Through the embodiments of this application, the value of historical transaction blocks is weighted and aggregated under physical constraints to obtain a spatial benchmark value. Long-term cyclical market sentiment is captured through oil price trend deviation to form a macro-value-added item. Engineering cost deduction items are derived by cross-modeling burial depth, slope, geological risk, and dynamic material cost index in an engineering scenario. These three factors work synergistically to form a quantifiable component with clear geological, economic, and engineering physical significance, improving the accuracy of value representation of transaction blocks. During the training phase, by constructing a multi-quantile loss function, the model can not only provide the most probable fair value (e.g., median) but also simultaneously output a conservative floor price (e.g., 10th quantile) and a loss threshold (e.g., 90th quantile), forming a complete price risk range.

[0111] Through the steps described in this application embodiment, starting from the static geological data, unstructured geological text, and dynamic engineering cost parameters of the target block, the initial feature vector of the target block is accurately generated by using a target language model finely tuned with oil and gas-specific corpus to extract geological semantics, a deep time-series model to capture oil price trends, and a fully connected layer to fuse dynamic engineering material cost indices. Simultaneously, based on spatial data such as transaction prices, surface distances, fault zone displacements, sealing coefficients, and structural unit connectivity of neighboring already transacted blocks, a physical barrier network reflecting the actual underground physical barrier relationships is constructed. In this network, if two blocks are separated by a strong fault zone, their connection weights are dynamically penalized to near 0, meaning that even if the surface distance is very close, their value transmission will be cut off by physical barriers. Furthermore, the graph attention network uses static geological feature weights amplified by Xavier during the feature mapping stage. The initialization strategy forces the model to prioritize fundamental geological endowments that determine oil and gas enrichment, such as porosity and reservoir type, rather than short-term oil price fluctuations. A geological gating attention mechanism is used to jointly modulate the attention score by combining the barrier penalty coefficient and surface distance, enabling the model to automatically learn an effective value transmission rate—that is, how much geological barriers the premium of historically high-priced blocks can penetrate and truly affect the value of the target block. Finally, the system inputs the initial feature vector of the target block and the spatially aggregated comprehensive feature vector into the quantile regression prediction model, outputting three price ranges: the lower quantile represents the safe bidding floor, the median is the fair value center, and the upper quantile is the loss threshold. Simultaneously, the system calculates the macroeconomic oil price trend appreciation and engineering extraction cost deductions, clearly decomposing the final valuation into interpretable and auditable physical components of spatial benchmarks, macroeconomic sentiment in the crude oil market, and exploration costs. This increases the accuracy of block value estimation and improves the efficiency of allocating limited oil and gas exploration and development resources.

[0112] In one implementation, a region can be partitioned in memory for loading the prediction model, which may include a structure data storage area and a parameter storage area. The structure data storage area stores structure-related code, and the parameters referenced by it can be accessed via pointers pointing to the addresses of specific parameters in the parameter storage area. During the training of the price prediction model, frequent parameter updates may be required; in this case, updating the parameter values ​​in the parameter storage area is sufficient.

[0113] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0114] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / random access memory (RAM), magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0115] According to another aspect of the embodiments of this application, an oil and gas exploration resource allocation apparatus for implementing the oil and gas exploration resource allocation method provided in the above embodiments is also provided, and details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0116] Figure 5 This is a structural block diagram of an optional oil and gas exploration resource allocation device according to an embodiment of this application, such as... Figure 5 As shown, the oil and gas exploration resource allocation device includes:

[0117] The data acquisition module 52 is used to acquire the first asset value data of the target block and the second asset value data and spatial value data of each of the multiple adjacent transaction blocks of the target block;

[0118] The data processing module 54 is used to extract features from the first asset value data to obtain the initial feature vector of the target block, and to construct a physical barrier network based on the first asset value data, multiple second asset value data and multiple spatial value data. The physical barrier network is used to extract features from the physical barrier network to obtain the spatial feature vector of the target block. The physical barrier network is used to reflect the physical barrier status between the target block and each adjacent transaction block.

[0119] Data analysis module 56 is used to analyze the initial feature vector and spatial feature vector of the target block using a pre-trained price prediction model to obtain the predicted transaction price range of the target block.

[0120] Module 58 is used to determine the oil and gas exploration and development resources of the target block based on the predicted transaction price range.

[0121] In an exemplary embodiment, the aforementioned first asset value data includes at least: static geological data and dynamic cost data, wherein the static geological data includes at least one of the following: physical scalar parameters of the target block, geological exploration text, and spatial topology data; and the dynamic cost data includes at least: the crude oil price sequence and dynamic engineering material cost index of the target block during the historical period prior to the listing time. The aforementioned apparatus is used to extract features from the first asset value data to obtain an initial feature vector of the target block, comprising: semantically encoding the geological exploration text using a target language model to obtain a semantic feature vector, and concatenating it with the physical scalar parameters to obtain a static feature vector of the target block, wherein the target language model is fine-tuned using an oil and gas geological text corpus. The resulting oil and gas geology text corpus includes: various oil and gas geology texts and corresponding semantic feature vectors, and the types of oil and gas geology texts include at least one of the following: geological exploration texts, drilling engineering description texts, block structural evaluation texts, and mining rights assessment opinion texts; feature extraction is performed on the crude oil price series using a depth time series model to obtain the dynamic feature vector of the target block; the spatial topology data of the target block is multiplied by the dynamic engineering material cost index, and the obtained dynamic engineering material cost index is mapped through a fully connected layer to obtain the engineering cost constraint vector of the target block, wherein the spatial topology data includes at least: the burial depth of the target layer and the terrain slope; the initial feature vector of the target block is composed of the static feature vector, dynamic feature vector, and engineering cost constraint vector of the target block.

[0122] In an exemplary embodiment, the spatial value data includes at least: historical transaction data of adjacent transaction blocks and surface distance, connectivity status, and geological barrier parameters between the target block and the corresponding adjacent transaction blocks, wherein; the aforementioned apparatus is used to construct a physical barrier network based on the first asset value data, multiple second asset value data, and multiple spatial value data in the following manner: for each adjacent transaction block, feature extraction is performed on the second asset value data and spatial value data of the adjacent transaction block to obtain a comprehensive feature vector of the adjacent transaction block, wherein the comprehensive feature vector includes at least: a static feature vector, a dynamic feature vector, an engineering cost constraint vector, and a spatial feature vector; a geological barrier penalty coefficient between the target block and each adjacent transaction block is determined based on the connectivity status and geological barrier parameters, and a connection weight between the target block and each adjacent transaction block is determined based on the geological barrier penalty coefficient and the surface distance; the initial feature vector of the target block and the comprehensive feature vectors of each of the multiple adjacent transaction blocks are used as node attribute information of graph nodes, and the connection weights are used as edge weights of the edges between each pair of graph nodes to obtain the physical barrier network.

[0123] In an exemplary embodiment, the geological barrier parameters include at least one of the following: the fault displacement, burial depth range, strike, and blocking coefficient of the fault zone between the target block and adjacent transaction blocks, wherein the aforementioned apparatus is used to determine the geological barrier penalty coefficient between the target block and each adjacent transaction block in the following manner based on the connectivity status and the geological barrier parameters: for each adjacent transaction block, determining whether the connectivity status between the target block and the adjacent transaction block is a fully connected state; if the connectivity status between the target block and the adjacent transaction block is a fully connected state, determining the geological barrier penalty coefficient between the target block and the adjacent transaction block as a first value; if the connectivity status between the target block and the adjacent transaction block is not a fully connected state, determining the geological barrier penalty coefficient between the target block and the adjacent transaction block as a second value based on the geological barrier parameters between the target block and the adjacent transaction blocks and a preset parameter weight.

[0124] In an exemplary embodiment, the above-described apparatus is used to extract features from a physical barrier network to obtain a spatial feature vector of a target block, including: determining the attention score between the target block and each neighboring transaction block based on the node attribute information of the graph nodes corresponding to the target block in the physical barrier network and the node attribute information of the graph nodes corresponding to each neighboring transaction block; determining the attention weight between the target block and each neighboring transaction block by combining the edge weights of the edges between the graph nodes corresponding to the target block and the graph nodes corresponding to each neighboring transaction block in the physical barrier network; normalizing the attention weight between the target block and multiple neighboring transaction blocks to obtain the spatial overflow weight of the target block; and aggregating the spatial feature vectors of each of the multiple neighboring transaction blocks according to the spatial overflow weight to obtain the spatial feature vector of the target block.

[0125] In an exemplary embodiment, the above-described apparatus is used to determine the attention score between the target block and each neighboring transaction block based on the node attribute information of the graph nodes corresponding to the target block within the physical barrier network and the node attribute information of the graph nodes corresponding to each neighboring transaction block, by means of the following method: mapping the node attribute information of the graph nodes corresponding to the target block and the node attribute information of the graph nodes corresponding to each neighboring transaction block within the physical barrier network to a high-dimensional latent space using a preset feature weight matrix, thereby obtaining the latent feature vectors of the graph nodes corresponding to the target block and the latent feature vectors of the graph nodes corresponding to each neighboring transaction block, wherein the feature weight matrix includes: static feature vectors, dynamic feature vectors, engineering cost constraint vectors, and spatial feature vectors, each of which... The corresponding feature weight sub-matrices, the feature weight sub-matrices corresponding to the dynamic feature vector, the engineering cost constraint vector, and the spatial feature vector are all determined using the value range after Xavier initialization, and the feature weight sub-matrices corresponding to the static feature vector are determined using the value range after Xavier initialization and amplification. The attention weight parameters of the graph attention network are used to perform a weighted summation of the concatenated feature vectors of the hidden layer feature vectors of the graph nodes corresponding to the target block and the hidden layer feature vectors of the graph nodes corresponding to each neighboring transaction block, to obtain the attention score between the target block and each neighboring transaction block. The attention weight parameters include the attention weights corresponding to the static feature vector, the dynamic feature vector, the engineering cost constraint vector, and the spatial feature vector.

[0126] In an exemplary embodiment, the above-described apparatus is used to train a price prediction model in the following manner: determining an initial prediction model; acquiring multiple sets of training sample data, wherein each set of training sample data includes: a comprehensive feature vector of the transaction block and the actual transaction price; for each training batch in the iterative training process, inputting the comprehensive feature vectors from each set of training sample data in the training batch into the initial prediction model to obtain each predicted transaction price range output by the initial prediction model, wherein the predicted transaction price range includes: the predicted transaction price corresponding to each of multiple quantiles; constructing a quantile regression loss function using each predicted transaction price range and the corresponding actual transaction price, and adjusting the model parameters of the prediction model obtained from the previous round of iterative training according to the quantile regression loss function.

[0127] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0128] According to another aspect of the embodiments of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein the program executes the steps in any of the above method embodiments when it is run.

[0129] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, ROMs, RAMs, portable hard drives, magnetic disks, or optical disks.

[0130] According to another aspect of the embodiments of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor is configured to perform the steps of any of the method embodiments described above via the computer program. In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0131] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0132] According to another aspect of the embodiments of this application, a computer program product is also provided, the computer program product including a computer program / instructions containing program code for performing the method shown in the flowchart.

[0133] Figure 6 A structural block diagram of an electronic device for implementing embodiments of this application is shown. Figure 6 As shown, the electronic device 60 may include one or more processors 602 (shown as 602a, 602b, ..., 602n in the figure) (processor 602 may include, but is not limited to, a processing device such as a microprocessor or programmable logic device), a memory 604 for storing data, and a transmission device 606 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 6 The structure shown is for illustrative purposes only and does not limit the structure of the computer system described above. For example, electronic device 60 may also include... Figure 6 The more or fewer components shown, or having the same Figure 6 The different configurations shown.

[0134] It should be noted that the aforementioned one or more processors 602 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element of the electronic device 60. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0135] The memory 604 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the oil and gas exploration resource allocation method in this embodiment. The processor 602 executes various functional applications and data processing by running the software programs and modules stored in the memory 604, thereby realizing the oil and gas exploration resource allocation method of the aforementioned application. The memory 604 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 604 may further include memory remotely located relative to the processor 602, and these remote memories can be connected to the electronic device 60 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0136] The transmission device 606 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the electronic device 60. In one example, the transmission device 606 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 606 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0137] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the electronic device 60.

[0138] It should be noted that, Figure 6 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0139] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0140] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for allocating oil and gas exploration resources, characterized in that, include: Obtain the first asset value data of the target block and the second asset value data and spatial value data of each of the multiple adjacent transaction blocks of the target block; Feature extraction is performed on the first asset value data to obtain the initial feature vector of the target block. A physical barrier network is constructed based on the first asset value data, multiple second asset value data, and multiple spatial value data. Feature extraction is performed on the physical barrier network to obtain the spatial feature vector of the target block. The physical barrier network is used to reflect the physical barrier status between the target block and each of the adjacent transaction blocks. The initial feature vector and spatial feature vector of the target block are analyzed using a pre-trained price prediction model to obtain the predicted transaction price range of the target block. The oil and gas exploration and development resources of the target block are determined based on the predicted transaction price range.

2. The method according to claim 1, characterized in that, The first asset value data includes at least: static geological data and dynamic cost data, wherein the static geological data includes at least one of the following: physical scalar parameters of the target block, geological exploration text, and spatial topology data; the dynamic cost data includes at least: crude oil price series and dynamic engineering material cost index of the target block during the historical period prior to the listing time; feature extraction is performed on the first asset value data to obtain an initial feature vector of the target block, including: The geological exploration text is semantically encoded using a target language model to obtain a semantic feature vector, which is then concatenated with the physical scalar parameters to obtain the static feature vector of the target block. The target language model is fine-tuned using an oil and gas geological text corpus, which includes various oil and gas geological texts and their corresponding semantic feature vectors. The types of oil and gas geological texts include at least one of the following: geological exploration text, drilling engineering description text, block structural evaluation text, and mining rights assessment opinion text. The crude oil price series is subjected to feature extraction using a deep time series model to obtain the dynamic feature vector of the target block; The spatial topology data of the target block is multiplied by the dynamic engineering material cost index, and the resulting dynamic engineering material cost index is mapped through a fully connected layer to obtain the engineering cost constraint vector of the target block. The spatial topology data includes at least: the target layer burial depth and the terrain slope. The initial feature vector of the target block is composed of the static feature vector, dynamic feature vector, and engineering cost constraint vector of the target block.

3. The method according to claim 1, characterized in that, The spatial value data includes at least: historical transaction data of the adjacent transaction blocks and surface distance, connectivity, and geological barrier parameters between the target block and its corresponding adjacent transaction blocks, wherein: a physical barrier network is constructed based on the first asset value data, multiple sets of second asset value data, and multiple sets of spatial value data, including: For each adjacent transaction block, feature extraction is performed on the second asset value data and spatial value data of the adjacent transaction block to obtain a comprehensive feature vector of the adjacent transaction block, wherein the comprehensive feature vector includes at least: a static feature vector, a dynamic feature vector, an engineering cost constraint vector, and a spatial feature vector. Based on the connectivity status and the geological barrier parameters, the geological barrier penalty coefficient between the target block and each of the adjacent transaction blocks is determined, and based on the geological barrier penalty coefficient and the surface distance, the connection weight between the target block and each of the adjacent transaction blocks is determined. The initial feature vector of the target block and the comprehensive feature vectors of each of the neighboring transaction blocks are used as the node attribute information of the graph nodes, and the connection weights are used as the edge weights of the edges between each pair of graph nodes to obtain the physical barrier network.

4. The method according to claim 3, characterized in that, The geological barrier parameters include at least one of the following: the fault displacement, burial depth range, strike, and blocking coefficient of the fault zone between the target block and the adjacent transaction block, wherein; Determining the geological barrier penalty coefficient between the target block and each of the adjacent transaction blocks based on the connectivity state and the geological barrier parameters includes: For each of the adjacent transaction blocks, determine whether the connectivity between the target block and the adjacent transaction blocks is fully connected; If the connectivity between the target block and the adjacent transaction block is fully connected, the geological barrier penalty coefficient between the target block and the adjacent transaction block is determined to be a first value. If the connectivity between the target block and the adjacent transaction block is not fully connected, the geological barrier penalty coefficient between the target block and the adjacent transaction block is determined as a second value based on the geological barrier parameter between the target block and the adjacent transaction block and a preset parameter weight, wherein the second value is less than the first value.

5. The method according to claim 1, characterized in that, Feature extraction is performed on the physical barrier network to obtain the spatial feature vector of the target block, including: Based on the node attribute information of the graph node corresponding to the target block in the physical barrier network and the node attribute information of the graph node corresponding to each neighboring transaction block, the attention score between the target block and each neighboring transaction block is determined. In combination with the edge weight of the edge between the graph node corresponding to the target block and the graph node corresponding to each neighboring transaction block in the physical barrier network, the attention weight between the target block and each neighboring transaction block is determined. The attention weights between the target block and the multiple neighboring transaction blocks are normalized to obtain the spatial overflow weight of the target block; The spatial feature vectors of the multiple neighboring transaction blocks are aggregated according to the spatial overflow weight to obtain the spatial feature vector of the target block.

6. The method according to claim 3, characterized in that, Based on the node attribute information of the graph nodes corresponding to the target block within the physical barrier network and the node attribute information of the graph nodes corresponding to each neighboring transaction block, the attention score between the target block and each of the neighboring transaction blocks is determined, including: Using a preset feature weight matrix, the node attribute information of the graph nodes corresponding to the target block in the physical barrier network and the node attribute information of the graph nodes corresponding to each neighboring transaction block are mapped to a high-dimensional latent space to obtain the latent layer feature vectors of the graph nodes corresponding to the target block and the latent layer feature vectors of the graph nodes corresponding to each neighboring transaction block. The feature weight matrix includes: feature weight sub-matrices corresponding to static feature vectors, dynamic feature vectors, engineering cost constraint vectors, and spatial feature vectors. The feature weight sub-matrices corresponding to dynamic feature vectors, engineering cost constraint vectors, and spatial feature vectors are all determined using the value range after Xavier initialization. The feature weight sub-matrices corresponding to static feature vectors are determined using the value range after Xavier initialization and amplification. The attention weight parameters of the graph attention network are used to perform a weighted summation of the concatenated feature vectors of the hidden layer feature vectors of the graph nodes corresponding to the target block and the hidden layer feature vectors of the graph nodes corresponding to each of the neighboring transaction blocks, to obtain the attention score between the target block and each of the neighboring transaction blocks. The attention weight parameters include the attention weights corresponding to the static feature vector, dynamic feature vector, engineering cost constraint vector and spatial feature vector.

7. The method according to claim 1, characterized in that, The training process of the price prediction model includes: Determine the initial prediction model; Obtain multiple sets of training sample data, wherein each set of training sample data includes: the comprehensive feature vector of the transaction block and the actual transaction price; For each training batch in the iterative training process, the comprehensive feature vector from each group of training sample data in the training batch is input into the prediction model obtained from the previous iteration training to obtain each predicted transaction price range output by the prediction model obtained from the previous iteration training. The predicted transaction price range includes the predicted transaction price corresponding to each of multiple quantiles. A quantile regression loss function is constructed using each predicted transaction price range and the corresponding actual transaction price, and the model parameters of the prediction model obtained from the previous iteration training are adjusted according to the quantile regression loss function.

8. An oil and gas exploration resource allocation device, characterized in that, include: The data acquisition module is used to acquire the first asset value data of the target block and the second asset value data and spatial value data of each of the multiple adjacent transaction blocks of the target block; The data processing module is used to extract features from the first asset value data to obtain the initial feature vector of the target block, and to construct a physical barrier network based on the first asset value data, multiple second asset value data and multiple spatial value data, and to extract features from the physical barrier network to obtain the spatial feature vector of the target block, wherein the physical barrier network is used to reflect the physical barrier status between the target block and each of the adjacent transaction blocks; The data analysis module is used to analyze the initial feature vector and spatial feature vector of the target block using a pre-trained price prediction model to obtain the predicted transaction price range of the target block. The determination module is used to determine the oil and gas exploration and development resources of the target block based on the predicted transaction price range.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the device containing the computer-readable storage medium executes the oil and gas exploration resource allocation method according to any one of claims 1 to 7 by running the computer program.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the oil and gas exploration resource allocation method according to any one of claims 1 to 7 through the computer program.

Images