Natural gas productivity intelligent prediction method and device for logging small-layer combination test

Abstract

The invention discloses an intelligent natural gas productivity prediction method and device for a logging small layer joint test, and the method comprises the steps: dividing a plurality of target small layers of a target gas well reservoir according to the obtained various logging curve data of a target gas well, constructing corresponding productivity prediction machine learning models, and carrying out the prediction of the productivity of the target gas well. Predicting single-layer daily gas yield and single-layer daily water yield; for the plurality of target small layers, determining an interference coefficient representing interlayer interference of the plurality of target small layers; according to the interference coefficient and the single-layer daily gas production rate and the single-layer daily water production rate of each target small layer, obtaining a combined test productivity prediction result of the logging small layer after the interference loss is deducted; and / or, a deep learning algorithm is adopted to establish a well logging small layer combined test productivity prediction model, and a combined test productivity prediction result corresponding to the target small layer combined test is output. The method is used for improving the productivity prediction efficiency and accuracy of the natural gas combined test well.

Description

Technical Field

[0001] This invention relates to the field of natural gas exploration and development technology, and in particular to a method and device for intelligent prediction of natural gas production capacity based on well logging and small-layer combined testing. Background Technology

[0002] This section is intended to provide background or context for the embodiments of the invention described herein. The description herein is not an admission that it is prior art simply because it is included in this section.

[0003] Traditional natural gas production forecasting methods mainly rely on empirical formulas and simple physical models, analyzing only single reservoirs and ignoring the complexity and inter-layer interference factors in combined well testing scenarios. Existing technologies such as well logging interpretation models, numerical simulations, and artificial intelligence algorithms generally suffer from the following shortcomings: they ignore inter-layer interference caused by pore pressure differences, gas-water permeability differences, depth differences, and gas-water relationships; they cannot accurately reflect the production characteristics of multi-layer reservoirs based on limited data; and single-layer prediction models cannot adapt to multi-layer combined production conditions, leading to significant deviations in prediction results.

[0004] Existing methods cannot solve the problem of integrating multi-layer reservoir parameters and quantifying inter-layer interference in combined well testing. Especially when facing scenarios involving multiple small layers in tight gas and low permeability combined testing, the large differences in reservoir quality and the unclear inter-layer contribution patterns lead to low accuracy and efficiency in production capacity prediction, making it difficult to support scientific mining decisions. Summary of the Invention

[0005] This invention provides an intelligent prediction method for natural gas production capacity in combined well logging tests, to improve the efficiency and accuracy of natural gas production capacity prediction. The method includes:

[0006] Based on the various logging curve data of the target gas well, the reservoir of the target gas well is divided into multiple target sub-layers, and various reservoir sub-layer parameters of each target sub-layer are determined.

[0007] For different types of target sublayers, corresponding production capacity prediction machine learning models are constructed respectively; the production capacity prediction machine learning models are used to predict the daily gas production and daily water production of a single layer using the reservoir sublayer parameters of the target sublayer as input features; the input features of the production capacity prediction machine learning models are preset types of reservoir sublayer parameters corresponding to the reservoir type.

[0008] For multiple target sublayers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship between the multiple target sublayers, an interference coefficient characterizing the interlayer interference of multiple target sublayers is determined; the interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sublayers.

[0009] The natural gas production capacity forecast for the target small-scale strata test is performed as follows:

[0010] Based on the interference coefficient, the daily gas production and daily water production of each target sublayer, the combined testing capacity prediction result of the logging sublayer after deducting interference losses is obtained; and / or, using a deep learning algorithm, a combined testing capacity prediction model of the logging sublayer is established based on various reservoir sublayer parameters, daily gas production, daily water production and the interference coefficient of the target sublayer; through the logging sublayer capacity prediction model, the combined testing capacity prediction result of the corresponding target sublayer is output.

[0011] This invention also provides an intelligent prediction device for natural gas production capacity in well logging tests of small layers, to improve the efficiency and accuracy of natural gas production capacity prediction. The device includes:

[0012] The target sub-layer division module is used to divide the target gas well reservoir into multiple target sub-layers based on the acquired multiple logging curve data of the target gas well, and to determine multiple reservoir sub-layer parameters for each target sub-layer.

[0013] The production capacity prediction machine learning model construction module is used to construct corresponding production capacity prediction machine learning models for different types of target sub-layers. The production capacity prediction machine learning model is used to predict the daily gas production and daily water production of a single layer using the reservoir sub-layer parameters of the target sub-layer as input features. The input features of the production capacity prediction machine learning model are preset types of reservoir sub-layer parameters corresponding to the reservoir type.

[0014] The interference coefficient determination module is used to determine the interference coefficient, which characterizes the interlayer interference of multiple target sublayers, based on the pore pressure difference, the difference in relative permeability of gas and water, the gravity effect caused by the depth difference, and the interlayer gas and water distribution relationship between the multiple target sublayers. The interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sublayers.

[0015] The capacity forecasting module is used to forecast the natural gas production capacity of the target small-layer combined test as follows:

[0016] Based on the interference coefficient, the daily gas production and daily water production of each target sublayer, the combined testing capacity prediction result of the logging sublayer after deducting interference losses is obtained; and / or, using a deep learning algorithm, a combined testing capacity prediction model of the logging sublayer is established based on various reservoir sublayer parameters, daily gas production, daily water production and the interference coefficient of the target sublayer; through the logging sublayer capacity prediction model, the combined testing capacity prediction result of the corresponding target sublayer is output.

[0017] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned intelligent prediction method for natural gas production capacity from well logging small-layer combined testing.

[0018] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent prediction method for natural gas production capacity from well logging small-layer combined testing.

[0019] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-mentioned intelligent prediction method for natural gas production capacity from well logging small-layer combined testing.

[0020] In this embodiment of the invention, based on the acquired logging curve data of the target gas well, multiple target sub-layers of the target gas well reservoir are divided, and multiple reservoir sub-layer parameters for each target sub-layer are determined. For different types of target sub-layers, corresponding production prediction machine learning models are constructed. These production prediction machine learning models are used to predict the daily gas production and daily water production of a single layer, using the reservoir sub-layer parameters as input features. The input features of the production prediction machine learning models are preset types of reservoir sub-layer parameters corresponding to the reservoir type. For multiple target sub-layers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer... The gas-water distribution relationship is used to determine the interference coefficient, which characterizes the inter-layer interference of multiple target sub-layers. The interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sub-layers. The natural gas production capacity prediction of the combined test of the target sub-layers is performed as follows: based on the interference coefficient, the daily gas production and daily water production of each target sub-layer, the combined test production capacity prediction result of the logging sub-layer after deducting interference losses is obtained; and / or, a deep learning algorithm is used to establish a combined test production capacity prediction model of the logging sub-layer based on various reservoir sub-layer parameters, daily gas production, daily water production and the interference coefficient of the target sub-layer; the combined test production capacity prediction result of the corresponding target sub-layer is output through the logging sub-layer production capacity prediction model. This invention addresses the reservoir description distortion problem caused by neglecting the heterogeneity of reservoir layers in traditional methods by utilizing multiple reservoir layer parameters for each target layer. It also solves the problem of existing single-layer prediction models being incompatible with reservoir quality differences by constructing corresponding production prediction machine learning models for different types of target layers, thus improving the accuracy of single-layer prediction for gas well reservoir layers. Furthermore, it determines the interference coefficient characterizing interlayer interference among multiple target layers based on the pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship, thus solving the problem of neglecting key interference factors such as interlayer crossflow and water intrusion risk in existing technologies. A mechanism-driven method is used to establish an interference loss correction function, subtracting interference loss from the theoretical production capacity of a single layer and then superimposing the effective production capacity. Finally, a data-driven method is used to construct an end-to-end model based on deep learning algorithms, integrating layer characteristic parameters, single-layer production capacity prediction values, and interference coefficients to directly output the combined testing results. This achieves high-precision intelligent prediction of combined logging layer production capacity, thereby improving the efficiency and accuracy of natural gas combined testing well production capacity prediction. Attached Figure Description

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

[0022] Figure 1 This is a flowchart illustrating a method for intelligent prediction of natural gas production capacity based on well logging small-layer combined testing, as described in an embodiment of the present invention.

[0023] Figure 2 This is a specific example diagram of a method for intelligent prediction of natural gas production capacity based on well logging small-layer combined testing in an embodiment of the present invention;

[0024] Figure 3 This is a specific example diagram of a well logging layer division and value acquisition flowchart in an embodiment of the present invention;

[0025] Figure 4 This is a specific example diagram of a single-layer capacity prediction machine learning method in an embodiment of the present invention;

[0026] Figure 5A This is a schematic diagram illustrating a specific method for calculating interlayer interference coefficients in an embodiment of the present invention.

[0027] Figure 5B This is a schematic diagram illustrating the composition of an interlayer interference coefficient in an embodiment of the present invention;

[0028] Figure 6 This is a structural example diagram of a natural gas production capacity intelligent prediction device for well logging small-layer combined testing in an embodiment of the present invention;

[0029] Figure 7 This is a schematic diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0031] In this document, the terms "and" and / or merely describe a relationship, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the term "at least one" in this document means any combination of at least two of any one or more of a plurality. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0032] In the description of this specification, the terms "comprising," "including," "having," and "containing" are open-ended, meaning they include but are not limited to. The references to "an embodiment," "a specific embodiment," "some embodiments," and "for example," etc., indicate that a specific feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. The order of steps involved in the various embodiments is used to illustrate the implementation of this application; the order of steps is not limited and can be adjusted appropriately as needed.

[0033] The acquisition, storage, use, and processing of data in this application comply with relevant regulations. The information collected in this application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation interfaces are provided for users to choose to authorize or refuse.

[0034] It should be noted that in the embodiments of this application, certain existing solutions in the industry, such as software, components, and models, may be mentioned. For example, some existing software tools, components, algorithm models, or solutions well-known in other technical fields may be cited. These should be considered exemplary, and their purpose is only to illustrate the feasibility of implementing the technical solution of this application. These mentions should be understood as typical examples, and their core purpose is to illustrate and verify the rationality and feasibility of implementing the technical solution proposed in this application. However, this does not mean that the applicant has already used or necessarily used the solution. Such citations do not imply that the applicant has actually adopted these existing solutions, or that it will necessarily adopt these methods in its technical implementation process in the future. In other words, these mentions are only illustrative in nature, helping to understand the connection and transcendence of the innovation points of this application with the prior art, and do not constitute an endorsement or reliance statement on a specific prior art product.

[0035] In natural gas exploration and development, combined well testing (i.e., opening multiple layers simultaneously for production) has become an important means to improve exploration and development efficiency. Accurately predicting the production capacity of reservoirs in combined well testing is crucial for rationally planning production schemes, improving recovery rates, and reducing development costs. Traditional natural gas production capacity prediction methods mainly rely on empirical formulas and simple physical models. These methods often only analyze single reservoirs, ignoring the complexity of reservoirs and inter-layer interference factors in combined well testing scenarios. In combined well testing, the pore pressure and relative permeability of gas and water differ among different layers, and the gravity caused by depth differences and the gas-water relationship between layers all significantly affect production capacity. Traditional methods cannot comprehensively consider these factors, resulting in low accuracy of prediction results.

[0036] While logging technology has advanced to provide more reservoir information, existing methods for predicting the production capacity of multiple sub-layers in a combined well test lack effective data integration and analysis. Currently, there is a lack of a comprehensive method for predicting natural gas production capacity levels that can intelligently and accurately consider multi-layer reservoir parameters and inter-layer interference characteristics in a combined well test, making it difficult to meet the actual needs of combined well testing operations in natural gas exploration and development.

[0037] Currently, there are three main methods for predicting reservoir productivity: well logging interpretation models, numerical simulation, and artificial intelligence algorithms. However, most of these methods are designed for single-layer prediction and have certain limitations. Specifically:

[0038] (1) Ignoring inter-layer interference: Single-layer prediction methods usually only consider the characteristics of a single reservoir, while ignoring the mutual influence and interference between different reservoirs. Under multi-layer reservoir conditions, the fluid flow between different reservoirs may affect each other. For example, the production of one reservoir may affect the pressure distribution of adjacent reservoirs, thereby affecting its production capacity.

[0039] (2) Data limitations: Single-layer prediction methods are often based on limited geological and logging data, which makes it difficult to fully reflect the productivity characteristics of multi-layer reservoirs. Under multi-layer reservoir conditions, there may be significant differences in reservoir quality, fluid properties, pressure distribution, etc., in different layers, and single-layer prediction methods are difficult to accurately capture these differences.

[0040] (3) Applicability of prediction models: Prediction models established by single-layer prediction methods are not applicable to multi-layer reservoir conditions. Under multi-layer reservoir conditions, the complexity of the reservoir increases, and more factors need to be considered, such as inter-layer interference and reservoir heterogeneity, while single-layer prediction models cannot accurately reflect these factors.

[0041] (4) Prediction bias: Due to the neglect of inter-layer interference and data limitations, the prediction results of single-layer prediction methods may have significant biases under multi-layer combined testing reservoir conditions. This bias may lead to overestimation or underestimation of reservoir productivity, thereby affecting oil and gas field development decisions. For example, the existing multi-layer combined production capacity prediction methods for tight sandstone gas reservoirs generally have a complex calculation process and require high data quality and processing capabilities, which limits their application in rapid on-site decision-making.

[0042] In summary, natural gas reservoirs such as tight gas and low-permeability reservoirs are mostly composed of multiple small layers, and the reservoir quality of each layer varies greatly. The contribution of different layers to the production capacity of oil and gas wells and the variation law of the contribution are unclear. Therefore, traditional production capacity prediction methods are difficult to obtain accurate prediction results.

[0043] To address the aforementioned problems, this invention provides an intelligent method for predicting natural gas production capacity in combined well logging tests, thereby improving the efficiency and accuracy of natural gas production capacity prediction. Figure 1 This is a flowchart illustrating a method for intelligent prediction of natural gas production capacity from a well logging small-layer combined test, as described in an embodiment of the present invention. Figure 1 The method may include:

[0044] Step 101: Based on the obtained logging curve data of the target gas well, divide the target gas well reservoir into multiple target sub-layers and determine the multiple reservoir sub-layer parameters for each target sub-layer;

[0045] Step 102: Construct corresponding production capacity prediction machine learning models for different types of target sub-layers; the production capacity prediction machine learning models are used to predict the daily gas production and daily water production of a single layer using the reservoir sub-layer parameters of the target sub-layer as input features; the input features of the production capacity prediction machine learning models are preset types of reservoir sub-layer parameters corresponding to the reservoir type.

[0046] Step 103: For multiple target sublayers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship between the multiple target sublayers, determine the interference coefficient characterizing the interlayer interference of the multiple target sublayers; the interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sublayers.

[0047] Step 104: Predict the natural gas production capacity of the target small-layer combined test as follows:

[0048] Based on the interference coefficient, the daily gas production and daily water production of each target sublayer, the combined testing capacity prediction result of the logging sublayer after deducting interference losses is obtained; and / or, using a deep learning algorithm, a combined testing capacity prediction model of the logging sublayer is established based on various reservoir sublayer parameters, daily gas production, daily water production and the interference coefficient of the target sublayer; through the logging sublayer capacity prediction model, the combined testing capacity prediction result of the corresponding target sublayer is output.

[0049] In this embodiment of the invention, based on the acquired logging curve data of the target gas well, multiple target sub-layers of the target gas well reservoir are divided, and multiple reservoir sub-layer parameters for each target sub-layer are determined. For different types of target sub-layers, corresponding production prediction machine learning models are constructed. These production prediction machine learning models are used to predict the daily gas production and daily water production of a single layer, using the reservoir sub-layer parameters as input features. The input features of the production prediction machine learning models are preset types of reservoir sub-layer parameters corresponding to the reservoir type. For multiple target sub-layers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer... The gas-water distribution relationship is used to determine the interference coefficient, which characterizes the inter-layer interference of multiple target sub-layers. The interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sub-layers. The natural gas production capacity prediction of the combined test of the target sub-layers is performed as follows: based on the interference coefficient, the daily gas production and daily water production of each target sub-layer, the combined test production capacity prediction result of the logging sub-layer after deducting interference losses is obtained; and / or, a deep learning algorithm is used to establish a combined test production capacity prediction model of the logging sub-layer based on various reservoir sub-layer parameters, daily gas production, daily water production and the interference coefficient of the target sub-layer; the combined test production capacity prediction result of the corresponding target sub-layer is output through the logging sub-layer production capacity prediction model. This invention addresses the reservoir description distortion problem caused by neglecting the heterogeneity of reservoir layers in traditional methods by utilizing multiple reservoir layer parameters for each target layer. It also solves the problem of existing single-layer prediction models being incompatible with reservoir quality differences by constructing corresponding production prediction machine learning models for different types of target layers, thus improving the accuracy of single-layer prediction for gas well reservoir layers. Furthermore, it determines the interference coefficient characterizing interlayer interference among multiple target layers based on the pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship, thus solving the problem of neglecting key interference factors such as interlayer crossflow and water intrusion risk in existing technologies. A mechanism-driven method is used to establish an interference loss correction function, subtracting interference loss from the theoretical production capacity of a single layer and then superimposing the effective production capacity. Finally, a data-driven method is used to construct an end-to-end model based on deep learning algorithms, integrating layer characteristic parameters, single-layer production capacity prediction values, and interference coefficients to directly output the combined testing results. This achieves high-precision intelligent prediction of combined logging layer production capacity, thereby improving the efficiency and accuracy of natural gas combined testing well production capacity prediction.

[0050] In practice, the first step is to proceed to step 101: based on the various logging curve data of the target gas well, divide the target gas well reservoir into multiple target sub-layers and determine the various reservoir sub-layer parameters for each target sub-layer.

[0051] In this embodiment, the various logging curve data include natural gamma, resistivity, sonic transit time, neutron porosity, and density curves; the various reservoir layer parameters include porosity, permeability, gas saturation, water saturation, clay content, and mineral composition obtained through logging interpretation models.

[0052] Based on the acquired logging data of various target gas wells, the reservoir of the target gas well is divided into multiple target sub-layers, and various reservoir sub-layer parameters for each target sub-layer are determined, including:

[0053] Based on the distribution characteristics of marker layers, sedimentary cycles, and lithological abrupt change interfaces in the regional geological stratification standard, vertical stratigraphy is divided according to the changes in well logging curve morphology, and the top and bottom depths, thicknesses, and various reservoir sub-layer parameters of each target sub-layer are determined.

[0054] In the above embodiments, after obtaining the logging curve data of the target gas well, the reservoir sub-layers are finely divided according to the regional geological stratification standard. The regional geological stratification standard includes the comprehensive identification of the distribution characteristics of marker layers, the sedimentary cycle law and the lithological transition interface. The vertical stratigraphic structure is analyzed by the changes in the shape of the logging curve. Specifically, the amplitude change of the natural gamma curve indicates the boundary between mudstone and sandstone layers, the abnormal abrupt change of the resistivity curve reflects the lithological transition interface, and the combination characteristics of sonic transit time and neutron porosity identify the sedimentary cycle.

[0055] Based on the above analysis, the top boundary depth, bottom boundary depth, and thickness parameters of each target sub-layer were determined. After completing the sub-layer division, the reservoir key parameters of each sub-layer were calculated using the well logging interpretation model, including porosity parameters, permeability parameters, gas saturation parameters, water saturation parameters, clay content parameters, and mineral composition parameters.

[0056] The logging data includes natural gamma curves, resistivity curves, sonic transit time curves, neutron porosity curves, and density curves. The resistivity curves include multiple types of measurements such as deep lateral resistivity and medium induced resistivity. The sonic transit time curves reflect the formation's sonic wave propagation characteristics, and the density curves provide information on rock volume density.

[0057] In specific implementation, after step 101: based on the acquired logging curve data of the target gas well, the target gas well reservoir is divided into multiple target sub-layers, and multiple reservoir sub-layer parameters of each target sub-layer are determined, step 102: for different types of target sub-layers, corresponding production capacity prediction machine learning models are constructed respectively; the production capacity prediction machine learning model is used to predict the daily gas production and daily water production of a single layer using the reservoir sub-layer parameters of the target sub-layer as input features; the input features of the production capacity prediction machine learning model are preset types of reservoir sub-layer parameters corresponding to the reservoir type.

[0058] In this embodiment, an adaptive production capacity prediction machine learning model is independently constructed to address the differences in the target reservoir types after division. For the first type of reservoir, porosity parameters, permeability parameters, gas saturation parameters, effective thickness parameters, and formation pressure parameters are selected as input features of the model. Daily gas production and daily water production from historical single-layer test or production data are used as output labels. Random forest or gradient boosting tree algorithms are used to train the model and learn the mapping relationship between features and production capacity.

[0059] For the second type of reservoir, considering its low porosity and low permeability characteristics and gas-water co-containment features, in addition to the basic parameters of porosity, permeability, gas saturation, and thickness, the resistivity and sonic transit time ratio are added to quantify the impact of pore throat structure on production capacity. Using historical data of daily gas production and daily water production of the same type of single layer as labels, XGBoost or adaptive algorithms are used to train the model independently. Both types of models optimize parameters through historical sample data. When predicting the target well, the corresponding sub-layer feature parameters are input into the trained model, and the predicted values ​​of theoretical daily gas production and daily water production of the single layer are directly output.

[0060] The different types of target layers include Class I and Class II reservoirs classified according to industry standards. For Class I reservoirs, the machine learning model for predicting production capacity uses porosity, permeability, gas saturation, thickness, and formation pressure as input features. For Class II reservoirs, the machine learning model for predicting production capacity uses porosity, permeability, gas saturation, and thickness as input features, with the resistivity-to-sonic transit time ratio added to adapt to the low porosity and permeability characteristics. The model establishes a mapping relationship between features and production capacity using random forest, gradient boosting tree, or XGBoost algorithms, and outputs predicted daily gas and water production for a single layer.

[0061] In one embodiment, it also includes:

[0062] Based on the set lower limits of porosity, permeability, and gas saturation, the effective target sublayers are obtained by excluding the dense layers and mudstone layers in the target sublayers.

[0063] Based on electrical resistivity, physical properties porosity and permeability, and gas-bearing properties and gas saturation, the effective target layers are divided into Class I reservoirs and Class II reservoirs.

[0064] In the above embodiments, after the extraction of reservoir sublayer parameters is completed, based on the set lower limits of porosity, permeability, and gas saturation, and in conjunction with core analysis data verification, non-reservoir rock layers such as tight layers and mudstone layers in the target sublayers are excluded, and effective target sublayers are obtained. Subsequently, according to the reservoir classification method specified in industry standards, based on multiple dimensions of indicators such as electrical resistivity parameters, physical property porosity and permeability parameters, and gas-bearing property gas saturation parameters, the effective target sublayers are scientifically divided into Class I reservoirs and Class II reservoirs.

[0065] In one embodiment, the input features of the production capacity prediction machine learning model constructed for the first type of reservoir are porosity, permeability, gas saturation, thickness, and formation pressure; the input features of the production capacity prediction machine learning model constructed for the second type of reservoir are porosity, permeability, gas saturation, thickness, resistivity, and sonic transit time ratio.

[0066] In the above embodiments, suitable machine learning models for production capacity prediction are constructed for the first type of reservoir and the second type of reservoir after division.

[0067] The input features of the first type of reservoir model are porosity parameters, permeability parameters, gas saturation parameters, thickness parameters, and formation pressure parameters. The production capacity of high porosity and permeability pure gas reservoirs is learned through random forest or gradient boosting tree algorithms.

[0068] The second type of reservoir model, in addition to porosity, permeability, gas saturation, and thickness parameters, adds resistivity and acoustic transit time ratio features to capture the control effect of the pore throat structure in the same layer of low-porosity gas-water permeability on production capacity. The model is trained independently using algorithms such as XGBoost. Both types of models use historical daily gas production and daily water production of a single layer as output labels, and predict the theoretical production capacity of a single layer by inputting the feature parameters of the target well corresponding to the reservoir type.

[0069] In one embodiment, corresponding capacity prediction machine learning models are constructed for different types of target sub-layers, including:

[0070] By employing random forest, gradient boosting tree, or XGBoost algorithms, a corresponding production capacity prediction machine learning model is constructed by establishing a mapping relationship between various reservoir sublayer parameter features and production capacity of the target sublayer.

[0071] In the above embodiments, adaptive production capacity prediction machine learning models are independently constructed to address the differences in physical properties of different types of reservoir sublayers. For the first type of reservoir, a random forest or gradient boosting tree algorithm is used to construct the model. By inputting porosity parameters, permeability parameters, gas saturation parameters, effective thickness parameters, and formation pressure parameters, a mapping relationship is established between the model and the daily gas production and daily water production of a single layer. For the second type of reservoir, an XGBoost algorithm is used to construct the model. In addition to porosity parameters, permeability parameters, gas saturation parameters, and thickness parameters, the resistivity and acoustic transit time ratio feature is added as input to establish a mapping relationship with the production capacity of the gas and water layers.

[0072] Both types of models optimize the mapping relationship through training on historical production samples. The specific training process includes: collecting the feature parameters of small layers of the same type of reservoir in historical gas wells and the corresponding actual daily gas production and daily water production data; using the feature parameters as input variables and the actual production data as output labels; and learning the nonlinear relationship between features and production capacity through algorithm iteration. When the trained model is applied to the target well, the feature parameters of the small layer to be predicted are input into the corresponding type of model, and the predicted values ​​of the theoretical daily gas production and daily water production of a single layer are directly output.

[0073] In specific implementation, after step 102: constructing corresponding production capacity prediction machine learning models for different types of target sub-layers, step 103: for multiple target sub-layers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and inter-layer gas-water distribution relationship between the multiple target sub-layers, determine the interference coefficient characterizing the inter-layer interference of the multiple target sub-layers; the interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sub-layers.

[0074] In one embodiment, for multiple target sublayers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship between the multiple target sublayers, an interference coefficient characterizing the interlayer interference of the multiple target sublayers is determined, including:

[0075] Based on the pore pressure difference, the interlayer fluid channeling trend is calculated by an extended form of Darcy's law, and a pressure difference interference coefficient is generated.

[0076] Based on the difference in relative permeability between air and water, the differences in flow capacity of different sub-layers under a fixed water saturation are analyzed by obtaining the air-water relative permeability curves of each sub-layer, and the interference coefficient of the difference in relative permeability between air and water is calculated.

[0077] The depth difference interference coefficient is calculated based on the gravitational potential energy difference determined by the gravitational effect caused by the depth difference.

[0078] Based on the interlayer air-water distribution relationship, the interference coefficient of the interlayer air-water relationship is calculated using a crossflow risk assessment model.

[0079] The interference coefficients characterizing the interlayer interference of multiple target sublayers are obtained by weighting the interference coefficients of pressure difference, gas-water relative permeability difference, depth difference, and interlayer gas-water relationship.

[0080] In the above embodiments, the process of quantitatively calculating the interlayer interference coefficient for combined test layer arrays is achieved through decoupling a four-dimensional physical mechanism:

[0081] First, based on the pore pressure difference mechanism, static or dynamic pressure data of each sublayer are collected to obtain the pore pressure distribution. Then, the interlayer fluid flow trend caused by the pressure difference is calculated by the extended form of Darcy's law. The pressure difference interference coefficient is defined and normalized.

[0082] Next, based on the gas-water relative permeability difference mechanism, the gas phase relative permeability and water phase relative permeability curves of each sub-layer were obtained through core experiments or empirical models. Under the condition of fixed water saturation, the gas-water flow capacity ratio difference of different sub-layers was analyzed, and the relative permeability difference interference coefficient was calculated.

[0083] Subsequently, based on the mechanism of gravity, the gravitational potential energy difference caused by the small interlayer depth difference was calculated, and the vertical position difference was converted into a gravity interference coefficient according to the hydrostatic pressure formula, so as to clearly distinguish the inhibitory or promoting effect of gravity on interlayer flow.

[0084] Finally, based on the interlayer gas-water distribution relationship mechanism, the gas saturation was interpreted by well logging to divide the pure gas layer, gas-water co-layer and water layer. Combined with the gas-water interface location to identify high-risk distribution patterns of inverted differentiation or isolated water body, a crossflow risk assessment model was established to quantify the gas-water relationship interference coefficient.

[0085] Finally, by weighted superposition, the interference coefficients of pressure difference, relative permeability difference, gravity, and gas-water relationship are combined to generate a comprehensive interference coefficient. This coefficient serves as a physical parameter to quantitatively characterize the mutual interference intensity between the combined test layers, and it fully characterizes the physical intensity of the mutual interference between the combined test layers.

[0086] In specific implementation, after step 103: for multiple target sublayers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship between the multiple target sublayers, determine the interference coefficient characterizing the interlayer interference of the multiple target sublayers, then proceed to step 104:

[0087] The natural gas production capacity forecast for the target small-scale strata test is performed as follows:

[0088] Based on the interference coefficient, the daily gas production and daily water production of each target sublayer, the combined testing capacity prediction result of the logging sublayer after deducting interference losses is obtained; and / or, using a deep learning algorithm, a combined testing capacity prediction model of the logging sublayer is established based on various reservoir sublayer parameters, daily gas production, daily water production and the interference coefficient of the target sublayer; through the logging sublayer capacity prediction model, the combined testing capacity prediction result of the corresponding target sublayer is output.

[0089] In one embodiment, based on the interference coefficient, the daily gas production and daily water production of each target sub-layer, the predicted combined production capacity of the logging sub-layer after deducting interference losses is obtained, including:

[0090] An interference loss correction function is established based on the interference coefficient; the inter-layer interference loss is deducted from the predicted values ​​of daily gas production and daily water production of each target sub-layer according to the interference loss correction function, and the effective daily gas production and effective daily water production of each sub-layer are obtained.

[0091] The effective daily gas production and effective daily water production of each sub-layer are superimposed to generate the combined test production capacity prediction result of the well logging sub-layer.

[0092] In the above embodiments, an interference loss correction function is established based on seepage theory. A production capacity conversion model is constructed by correlating the comprehensive interference coefficient with the theoretical daily gas production and theoretical daily water production of a single layer. The interference loss is corrected for the theoretical production capacity of each target sub-layer using this function. Specifically, the theoretical daily gas production and theoretical daily water production of a single layer are input into the correction function, and the production capacity loss caused by inter-layer interference is calculated by combining the comprehensive interference coefficient. After deducting the loss, the effective daily gas production and effective daily water production of each sub-layer are output. This correction function quantifies the interference effect through the principle of seepage mechanics, and the function form is determined by fitting experimental data based on actual geological conditions and fluid characteristics.

[0093] After completing the correction of all small layer interference losses, the effective daily gas production of each small layer is summed to generate the prediction result of the total daily gas production of the combined test, and the effective daily water production of each small layer is summed to generate the prediction result of the total daily water production of the combined test.

[0094] In this embodiment, the interference loss correction function is established based on seepage theory. It constructs a capacity conversion model by associating the comprehensive interference coefficient with the theoretical daily gas production and theoretical daily water production of a single layer. The interference loss is corrected for the theoretical capacity of each target sub-layer using this function, and the effective daily gas production and effective daily water production after deducting the influence of inter-layer interference are calculated. Finally, the corrected effective daily gas production of each sub-layer is superimposed to obtain the predicted result of the total daily gas production of the combined test, and the effective daily water production of each sub-layer is superimposed to obtain the predicted result of the total daily water production of the combined test.

[0095] In one embodiment, a deep learning algorithm is used to establish a well logging test production capacity prediction model for the target sub-layer based on various reservoir sub-layer parameters, daily gas production of a single layer, daily water production of a single layer, and the aforementioned interference coefficient. This model includes:

[0096] Using the reservoir layer parameters, daily gas production, daily water production, and interference coefficient of historical gas wells as input features, and the actual combined test daily gas production and actual combined test daily water production as output labels, a training sample set is constructed.

[0097] A multilayer perceptron, convolutional neural network, or long short-term memory network is used to train a sample set to obtain a well logging small-layer combined test production capacity prediction model.

[0098] In the above embodiments, a well logging layer combined test production capacity prediction model is constructed using a multilayer perceptron, convolutional neural network, or long short-term memory network. The training samples of this model are formed by integrating multidimensional data from historical gas wells, specifically including the reservoir characteristic parameters of each layer participating in the combined test, the daily gas production and daily water production values ​​of a single layer predicted based on the single-layer model, and the calculated inter-layer interference coefficient as input features. At the same time, the total daily gas production and total daily water production in the actual combined test production of the historical gas well are used as output labels. By pairing the above input features with the output labels to form a training sample set, machine learning algorithms are used to optimize the model parameters to learn the mapping relationship between input features and combined test production capacity. When the trained model is applied to the target gas well, its corresponding layer characteristic parameters, single-layer production capacity prediction results, and interference coefficients are input into the model, and the predicted daily gas production and daily water production values ​​of the combined test of the well are directly output.

[0099] In this embodiment, the deep learning algorithm uses a multilayer perceptron, convolutional neural network, or long short-term memory network to construct a well logging layer combined test production capacity prediction model. The training samples of this model integrate the layer feature parameters of historical gas wells, the predicted daily gas production of a single layer, the predicted daily water production of a single layer, and the interference coefficient as input features, and use the actual combined test daily gas production and the actual combined test daily water production as output labels. Through the trained model, the layer feature parameters, single-layer production capacity prediction results, and interference coefficient of the target gas well are input, and the predicted daily gas production and daily water production of the well combined test are directly output.

[0100] The following three specific embodiments illustrate the specific application of the present invention.

[0101] First specific embodiment:

[0102] The purpose of this specific embodiment is to provide an intelligent prediction method for natural gas production capacity in well logging small-layer combined testing. By comprehensively considering various reservoir parameters and inter-layer interference factors, it utilizes intelligent algorithms to achieve accurate prediction of natural gas production capacity. This specific embodiment aims to solve the above-mentioned problems existing in the prior art and achieve the following functions:

[0103] 1. Based on well logging data and geological standards, finely divide the sub-layers and obtain key parameters.

[0104] 2. Accurately identify and scientifically classify reservoirs (mainly for Class I and Class II reservoirs).

[0105] 3. For different types of reservoirs, machine learning methods are used to predict their single-layer productivity (daily gas production, daily water production).

[0106] 4. Calculate the interlayer interference coefficient caused by pore pressure difference, gas-water relative permeability difference, depth difference (gravity), and interlayer gas-water relationship.

[0107] 5. By combining mechanism-driven (reducing interference losses) and data-driven (machine learning) methods, high-precision and high-efficiency prediction of gas well test production capacity can be achieved.

[0108] Figure 2 This is a specific example diagram of a method for intelligent prediction of natural gas production capacity based on a small-layer well logging test according to an embodiment of the present invention. This embodiment provides a method for intelligent prediction of natural gas production capacity based on a small-layer well logging test, as follows: Figure 2 As shown, it includes the following steps:

[0109] Step 1: Well logging layer division and value determination:

[0110] (1) Obtain logging curve data of the target gas well, including but not limited to natural gamma (GR), resistivity (RT, ILD, ILM), sonic transit time (AC), neutron porosity (CNL), density (DEN), etc.

[0111] (2) Combining regional geological stratification standards (such as marker layers, sedimentary cycles, lithological abrupt change interfaces, etc.), and utilizing the morphological characteristics of well logging curves, the vertical strata of gas wells are finely divided into small layers to determine the top and bottom depths and thicknesses of each small layer.

[0112] (3) Apply the applicable well logging interpretation model to calculate the key reservoir parameters of each sub-layer: porosity ( Permeability (K), gas saturation (Sg), water saturation (Sw), clay content (Vsh), mineral composition, etc.

[0113] (4) Compile the basic data for each sublayer: depth, thickness, layer position, and porosity (…). ), permeability (K), gas saturation (Sg), water saturation (Sw), clay content (Vsh), etc.

[0114] Step 2: As Figure 3 As shown, Figure 3This is a specific example diagram of a well logging layer division and value extraction flowchart in an embodiment of the present invention, which includes reservoir identification and classification, including the following:

[0115] Reservoir identification: Based on the reservoir parameters obtained in step 1 ( ), and in conjunction with core analysis data, set lower limit standards for reservoir properties and gas content, such as (Minimum porosity) (Minimum penetration rate) (Minimum gas saturation) to identify and exclude non-reservoir rock formations (such as tight layers and mudstone layers) and determine effective reservoirs.

[0116] Reservoir Classification: For identified valid reservoirs, reservoir classification standards are formulated according to industry standards. Referring to SY / T6832-2020 Oil and Gas Reservoir Evaluation Methods, the classification is primarily based on electrical properties. , physical properties Based on K and gas-bearing Sg, the reservoir is classified into the well logging classification results of a certain gas field as shown in Table 1.

[0117] Table 1

[0118]

[0119] Note: This invention only applies to the prediction of combined reservoir productivity for Class I and Class II reservoirs.

[0120] Step 3: Intelligent prediction of single-layer reservoir productivity:

[0121] Figure 4 This is a specific example diagram of a single-layer capacity prediction machine learning method in an embodiment of the present invention, such as... Figure 4 As shown, prediction models are established for different types of reservoirs:

[0122] (1) Construction and prediction of production capacity prediction model for Class I reservoirs:

[0123] ① Feature Selection: Parameters closely related to the productivity of Class I reservoirs are selected as model input features (X_I). Typical features include: porosity ( Permeability (K), gas saturation (Sg), effective thickness (Thickness), and formation pressure (Pr).

[0124] ② Output labels: Use the daily gas production (Qg) and daily water production (Qw) from historical single-layer test or production data as output labels (Y_I).

[0125] ③ Model Selection and Training: Select appropriate machine learning algorithms (such as Random Forest, Gradient Boosting Tree (GBDT, XGBoost, LightGBM), Support Vector Machine (SVM), Backpropagation Neural Network (ANN), etc.). Train the model using historical Class I reservoir sample data (X_I, Y_I) to learn the mapping relationship between features and production capacity.

[0126] ④ Production capacity prediction: Input the corresponding features (X_I) of the sub-layer of the Class I reservoir in the target well into the trained model to predict its theoretical daily gas production (Qg_Pred_I) and daily water production (Qw_Pred_I) per layer.

[0127] (2) Construction and prediction of production capacity prediction model for Class II reservoirs:

[0128] ① Feature Selection: Considering the low porosity and low permeability characteristics of Class II reservoirs and the potentially different controlling factors, suitable feature parameters (X_II) are selected. Except... In addition to basic parameters such as resistivity, K, Sg, and Thickness, it is necessary to pay attention to specific characteristics (such as resistivity response to low saturation, acoustic transit time indicative of pore throat structure, etc.) or introduce new characteristics.

[0129] ② Output Labels: Historical daily gas production (Qg) and daily water production (Qw) of a single reservoir are also used as output labels (Y_II). It must be ensured that the data comes from a Class II reservoir.

[0130] ③ Model Selection and Training: Select a machine learning algorithm suitable for the characteristics of Class II reservoir data (it may be different from or the same as Class I but trained independently). Train an independent model using historical Class II reservoir sample data (X_II, Y_II).

[0131] ④ Production capacity prediction: Input the corresponding features (X_II) of the target well's Class II reservoir sublayer into the trained Class II model to predict its theoretical daily gas production (Qg_Pred_II) and daily water production (Qw_Pred_II) per layer.

[0132] Step 4: Calculation of inter-layer interference correction coefficient:

[0133] Figure 5A This is a schematic diagram illustrating a specific method for calculating inter-layer interference coefficients in an embodiment of the present invention. Figure 5B This is a specific schematic diagram illustrating the composition of an interlayer interference coefficient in an embodiment of the present invention, as shown below. Figure 5A and Figure 5B As shown, for each layer combination in the combined production, the comprehensive coefficient or its component coefficients characterizing the interlayer interference intensity are quantitatively calculated. The following four physical mechanisms are mainly considered: pressure difference interference, phase permeability difference, gravity difference, and gas-water relationship.

[0134] (1) Pressure difference interference coefficient ( ):

[0135] ① Pressure data analysis: Collect static or dynamic pressure data for each sub-layer (take DST and RFT data from well tests).

[0136] ② Data processing: Obtain the magnitude and distribution of pore pressure in each sublayer (Pi) (i=1, 2, ..., n).

[0137] ③ Define the interlayer pressure difference:

[0138]

[0139] in The average pore pressure of adjacent layers i and j; The pressure difference between adjacent layers i and j is denoted as .

[0140] ④ Calculation of crossflow trend: Based on the extended form of Darcy's law, the calculation is performed due to pressure difference ( The tendency and intensity of interlayer fluid crossflow caused by:

[0141]

[0142] in: This represents the crossflow rate from layer i to layer j ( A positive value indicates that the fluid flows from layer i to layer j;

[0143] Indicates the interlayer flow coefficient ( The calculation formula depends on the properties of the rock and fluid:

[0144]

[0145] Interlayer vertical permeability (mD);

[0146] Interlayer contact area ( ), which is related to layer thickness and well pattern;

[0147] Fluid viscosity ( );

[0148] : Effective distance between layers (m), which can be approximated as the interlayer spacing if the layers are directly adjacent;

[0149] ⑤ Define the inter-layer interference coefficient α p There is bidirectional current flow between multiple layers. The interference intensity of layer i on layer j is quantized as follows:

[0150]

[0151] in, , represents the pressure interference coefficient of layer i on layer j, which is dimensionless; when i=j, α=0, there is no self-interference; This represents the production capacity of the i-th layer;

[0152] ⑥ Normalization of interference coefficients: To eliminate the influence of dimensions, the interference coefficients are normalized. Normalize to the interval [0, 1].

[0153]

[0154] in The calculated maximum interference coefficient; This indicates the result of the normalization process.

[0155] (2) Interference coefficient of relative permeability difference between air and water ( ):

[0156] ① Permeability curve acquisition: Obtain or estimate the relative permeability curves of the gas phase (k_rg) and the relative permeability of the water phase (k_rw) of each sublayer through core experiments or empirical models.

[0157] ② Flow capacity analysis: Under combined production conditions, assuming a fixed water saturation level Sw and a fixed production pressure difference ΔP, where Sw is derived from production data. ΔP is the difference between the bottom hole flowing pressure and the formation pressure (assuming the difference is the same across all layers);

[0158] The differences in air and water flow capacity caused by variations in k_rg / k_rw across different sublayers were analyzed.

[0159] a. Extract the relative permeability value under the current Sw: Extract the Sw value for layer i. and ;in, This represents the relative gas phase permeability of layer i; This represents the relative permeability of the aqueous phase in layer i. This indicates the water saturation level of layer i;

[0160] b. Calculate the air-water flow capacity ratio (Mi):

[0161]

[0162] ③ Coefficient Calculation: Calculate the interlayer interference coefficient caused by the difference in relative permeability. The intensity of interlayer interference is directly proportional to the difference in flow capacity.

[0163]

[0164] in, The ratio of air to water flow capacity in layer i; This indicates the ratio of air to water flow capacity in layer j.

[0165] (3) Gravity effect (depth difference) interference coefficient ( ):

[0166] This coefficient is used to quantify the impact of gravity effects caused by small interlayer depth differences (vertical location) on interlayer crossflow and production dynamics. Gravity effects are particularly important in reservoirs with good vertical connectivity (such as thick layers and fractured zones), and may exacerbate or inhibit crossflow.

[0167] ① Depth difference calculation: Calculate the depth difference between the combined test layers ( ).

[0168]

[0169] in: The depth of a sublayer is usually taken as the depth in the middle of the reservoir, and is the average of the depth of the top boundary and the depth of the bottom boundary of the sublayer.

[0170] ② Pressure difference calculation: using the hydrostatic pressure formula Calculate the pressure difference between layers caused by gravity, where ρ is the interlayer average fluid density, and g is the gravitational acceleration.

[0171] ③ Coefficient conversion: Converting the pressure difference caused by gravity... Quantification coefficients for contribution / suppression of interlayer crossflow or production. Quantify its specific impact on interlayer interference:

[0172]

[0173] in: Average production pressure differential (bottom hole flowing pressure vs. formation pressure differential) is derived from production data.

[0174] when When layer i is above layer j: it promotes fluid flow from j to i.

[0175]

[0176] when When layer i is below layer j: suppress fluid flow from i to j:

[0177]

[0178] (4) Inter-layer air-water relationship interference coefficient ( ):

[0179] This coefficient is used to quantify the risk of interlayer crossflow caused by small interlayer gas-water distribution relationships (such as gas layer above / water layer below, gas and water layers adjacent to each other, etc.), with particular attention to the risk of water intrusion.

[0180] ① Gas-water distribution analysis: Analyze the gas-water distribution relationship between the test layers (such as gas layer above / water layer below, gas and water layers adjacent to each other, water layer above / gas layer with risk of water intrusion, etc.), as shown in Table 2. Table 2 is a table of gas-water distribution relationships between layers.

[0181] Table 2

[0182]

[0183] The specific steps are as follows:

[0184] (a) Geological data: Well logging interpretation of gas saturation (Sg) in each layer:

[0185] Sg>70%: Pure gas layer;

[0186] 30%≤Sg≤70%: Gas and water in the same layer;

[0187] Sg < 30%: Water layer;

[0188] (b) Draw the top and bottom depth profiles of the sublayer and mark the gas-water interface (GWC);

[0189] (c) Output the interlayer gas-water relationship matrix: The distribution pattern encoding of layers i and j is represented (e.g., 0-normal, 1-same layer, 2-inverted, 3-isolated).

[0190] ② Crossflow risk assessment: For modes with crossflow risk (such as water intrusion), crossflow volume or suppression degree is analyzed by numerical simulation based on two-phase flow theory (simplified model) and Buckley-Leverett theory extension, as shown in Table 3. Table 3 is a crossflow risk table for interlayer gas-water distribution.

[0191]

[0192] in: : Water phase flow velocity;

[0193] Indicates the relative permeability of the aqueous phase;

[0194] Indicates the viscosity of water;

[0195] Represents the gravitational potential of water;

[0196] Indicates pore pressure; Indicates the absolute permeability of the rock;

[0197] : Water phase potential energy gradient (including gravity term);

[0198] Cross-current criterion: When Furthermore, when layer j is a water body, water intrusion occurs.

[0199] Table 3

[0200]

[0201] ③ Coefficient determination: Calculate the interference coefficient α caused by the specific gas-water relationship model. f .

[0202]

[0203] Risk value in step ②;

[0204]

[0205] Depth term: Normalized depth difference, used to amplify the risk of neighboring layers.

[0206] (5) Overall inter-layer interference coefficient ( ):

[0207] The four coefficients mentioned above can be combined by weighted superposition into a single coefficient that comprehensively characterizes the intensity of inter-layer interference. The relevant parameters of the inter-layer interference coefficient can be found in Table 4, which shows the calculation method for the relevant parameters of the inter-layer interference coefficient.

[0208]

[0209] Table 4

[0210]

[0211] Step 5: Intelligent prediction of gas well commissioning capacity:

[0212] The following two methods can be used for prediction, and can be mutually verified or the better method can be selected:

[0213] Method 1: Mechanism-driven approach:

[0214] (1) Calculate the theoretical capacity of a single layer without interference: Use the single-layer capacity (Qg_Pred_i, Qw_Pred_i) of each small layer predicted in step 3. This capacity data represents the theoretical capacity under the condition of no inter-layer interference.

[0215] (2) Quantitative deduction of interference loss: Based on the interlayer interference coefficient (e.g., α_total) obtained in step 4, and combined with the seepage mechanics mechanism, a quantitative model of the production capacity loss caused by interference is established. The details are as follows:

[0216] Model construction: Based on the principle of the impact of disturbance on production capacity in seepage mechanics, Darcy's law or production capacity equation is modified and adjusted.

[0217] That is, Qg_effective_i=Qg_Pred_i*f(α_total,...);

[0218] Qw_effective_i=Qw_Pred_i*g(α_total,...).

[0219] Here, f and g are functions reflecting the disturbance loss, and their forms can be determined by fitting experimental data based on actual geological conditions and fluid characteristics.

[0220] Qg_effective_i and Qw_effective_i represent the effective gas production capacity and effective water production capacity of each sublayer after considering the disturbance, respectively. This model quantifies the disturbance factors into specific numerical values ​​of their impact on production capacity.

[0221] (1) Overlapping effective production capacity:

[0222] The effective capacity of each sub-layer (Qg_effective_i, Qw_effective_i) after interference coefficient correction is superimposed to obtain the final combined trial capacity prediction result:

[0223]

[0224]

[0225] Among them, Qg_total_mech and Qw_total_mech are the predicted total gas production capacity and total water production capacity after considering inter-layer interference, respectively. The results comprehensively reflect the impact of inter-layer interference on the production capacity of each sub-layer under actual formation conditions and the final combined test production capacity.

[0226] Method 2: Data-driven approach:

[0227] (1) Constructing a sample set

[0228] ① Collect samples of historical gas well test production data. Each sample must include the following key information:

[0229] 1) Sub-layer feature data: contains all feature data of each sub-layer participating in the joint trial (derived from steps 1 and 2);

[0230] 2) Single-layer predicted capacity: The single-layer predicted capacity of each sub-layer (Qg_Pred_i, Qw_Pred_i) comes from step 3;

[0231] 3) Interference coefficients: The calculated inter-layer interference coefficients include αp, αkr, αg, αf, and αtotal (derived from step 4);

[0232] ② Actual combined test production capacity: The actual combined test production capacity data of this well, namely the historical daily gas production Qg_actual_total and daily water production Qw_actual_total.

[0233] (2) Model building and training

[0234] ① Model selection: Build machine learning models, such as multilayer perceptrons (MLP), convolutional neural networks (CNN) for processing sequence features, long short-term memory networks (LSTM) for processing temporal dependencies, or other advanced regression models;

[0235] ② Determine the input and output:

[0236] Input (X_total): The first four data items in the sample, namely, the small-layer features, the single-layer predicted capacity, and the interference coefficient;

[0237] Output (Y_total): Combined trial capacity, including Qg_total_Pred and Qw_total_Pred;

[0238] ③ Model training: Use the collected historical sample data to train the constructed model and optimize the model parameters to improve prediction accuracy.

[0239] (3) Production capacity prediction of combined gas wells

[0240] For the target well, follow these steps to obtain the combined test production capacity prediction results:

[0241] ① Data Preparation: Complete steps 1 to 4 to obtain the characteristics of each sub-layer of the target well test section, the predicted single-layer productivity of each sub-layer (Qg_Pred_i, Qw_Pred_i), and the calculated inter-layer interference coefficient (α). p α kr α g α f α total );

[0242] ② Data Input: The above data is integrated into the input (X_total_target) and input into the pre-trained intelligent prediction model for combined trial production capacity;

[0243] Results output: The model directly outputs the predicted daily gas production (Qg_total_Pred_AI) and daily water production (Qw_total_Pred_AI) of the well.

[0244] Second specific embodiment:

[0245] This specific embodiment illustrates the application of this method in predicting the production capacity of multi-stage combined well tests in a gas field:

[0246] Taking Well A in a tight sandstone gas field in western China as an example, this well contains three Class I sub-layers (I-1, I-2, I-3) and two Class II sub-layers (II-1, II-2). The method of this invention is used to predict the combined test production capacity. The specific steps are as follows:

[0247] Step 1: Well logging layer division and parameter extraction:

[0248] ① Data acquisition: Collect well logging data, including natural gamma (GR), resistivity (RT), sonic transit time (AC), neutron porosity (CNL), density (DEN), etc.

[0249] ② Sub-layer division: Based on the regional geological stratification standard (the marker layer is a stably distributed mudstone interlayer), sub-layers were divided according to the morphological characteristics of well logging curves. The top and bottom depths and thicknesses of each sub-layer were determined. The results are as follows:

[0250] Layer I-1: Top depth 3200m, bottom depth 3215m, thickness 15m;

[0251] Layer I-2: Top depth 3230m, bottom depth 3242m, thickness 12m;

[0252] Layer I-3: Top depth 3255m, bottom depth 3268m, thickness 13m;

[0253] Layer II-1: Top depth 3280m, bottom depth 3290m, thickness 10m;

[0254] Layer II-2: Top depth 3305m, bottom depth 3318m, thickness 13m;

[0255] ③ Parameter Calculation: Key parameters for each sub-layer were calculated using the well logging interpretation model. Some results are shown in Table 5 below:

[0256] Table 5

[0257]

[0258] Step 2: Reservoir Identification and Classification

[0259] Based on the SY / T6832-2020 standard and the classification rules in Table 1, and combined with the above parameters, the following determination is made:

[0260] Layers I-1, I-2, and I-3 have a porosity >12%, permeability >1mD, and gas saturation >70%, classifying them as Class I reservoirs.

[0261] Layers II-1 and II-2 have a porosity of 8-12%, a permeability of 0.3-1 mD, and a gas saturation of 60-70%, classifying them as Class II reservoirs.

[0262] Step 3: Intelligent Prediction of Single-Layer Capacity:

[0263] ① Type I reservoir model: Random forest algorithm is selected, with input features as follows: Using historical production data (120 samples in total) from 30 wells in this gas field for Class I reservoirs, based on K, Sg, thickness, and Pr, a model was trained to predict the production capacity of the Class I sub-layer in well A.

[0264]

[0265] ② Type II reservoir model: The XGBoost algorithm is used, and in addition to the basic parameters, the resistivity to acoustic transit time ratio (RT / AC) is added to the input features. The model is trained using historical data (95 samples) of Type II reservoirs from 25 wells. Prediction results:

[0266]

[0267] Step 4: Calculation of inter-layer interference coefficient:

[0268] ① Pressure difference interference coefficient (αp):

[0269] Calculate the interlayer pressure difference ΔP_ij, such as ΔP_I-1, I-2 = |35.6-34.8| = 0.8 MPa;

[0270] Based on the extended Darcy's law, the crossflow rate q_ij is calculated. Combined with parameters such as interlayer vertical permeability (taken as 0.5mD) and contact area (calculated as 1.5m² based on a well diameter of 0.1m), the αp matrix is ​​obtained. After normalization, the αp* range is 0.12-0.35.

[0271] ② Relative permeability difference coefficient (αkr):

[0272] The k_rg and k_rw values ​​for each layer were obtained through core experiments, and the flow capacity ratio Mi was calculated. For example, Mi = 5.2 for layer I-1 and Mi = 1.8 for layer II-1. The result αkr_ij = ln(Mi / Mj) was calculated, with a range of -0.95 to 1.02.

[0273] ③ Gravitational coefficient (αg):

[0274] Calculate the interlayer depth difference ΔD_ij, such as ΔD_I-1,II-1=|3207.5-3285|=77.5m;

[0275] According to the formula αg=±(ρ_avg×g×ΔD) / ΔP_prod (ΔP_prod is taken as 5MPa), the range of αg is -0.21-0.18.

[0276] ④ Gas-water relationship coefficient (αf):

[0277] Well logging interpretation of Sg indicates that Class I is a pure gas layer (Sg > 70%), and Class II is a gas-water co-layer (Sg = 65-68%).

[0278] The interlayer relationship matrix shows that Type I and Type II are gas-water co-layered, with a crossflow risk value of I_ij=0.3. Combined with the depth difference, the calculated αf range is 0.08-0.15.

[0279] ⑤ Overall interference coefficient:

[0280] Using weighted aggregation and regression analysis based on historical data, we obtained β1=0.3, β2=0.25, β3=0.2, and β4=0.25, and calculated α_total to be in the range of 0.15-0.42.

[0281] Step 5: Trial Capacity Forecast:

[0282] ① Mechanism-driven method:

[0283] Establish the interference loss function f(α_total) = 1 - 0.8 × α_total, and correct the single-layer effective capacity superposition to obtain:

[0284] .

[0285] ② Data-driven approach:

[0286] Historical data (including layer characteristics, interference coefficients, and actual production capacity) from 20 wells tested in this gas field were collected. The data from well A was used as input to train an MLP model. The model output was:

[0287] .

[0288] The results are verified as follows:

[0289] The actual production data of this well during the combined testing was 25.8 × 10⁻⁶ gas per day. 4 With a daily water production of 90 m³ / d, the prediction errors of the two methods were 0.8% and 1.2%, respectively, which are significantly higher than those of the traditional method (error > 15%), thus verifying the accuracy of the present invention.

[0290] Third specific embodiment:

[0291] This embodiment presents a smart prediction scheme for natural gas production capacity from well logging tests of small layers, including the following:

[0292] Step 1: Obtain the logging curve data of the target gas well, and in conjunction with the regional geological stratification standard, perform fine sub-layer division of the vertical strata of the gas well, determine the top and bottom depths and thicknesses of each sub-layer, apply the logging interpretation model to calculate the key reservoir parameters of each sub-layer, and organize the basic data of each sub-layer.

[0293] Step 2: Based on the reservoir parameters obtained in Step 1 and combined with core analysis data, set the lower limit standards for reservoir properties and gas content, identify and exclude non-reservoir strata, determine effective reservoirs, and formulate reservoir classification standards according to industry standards for the identified effective reservoirs, classifying the reservoirs into Class I, Class II, Class IIIA, Class IIIB, and Class IV. Only Class I and Class II reservoirs are used for combined production capacity prediction.

[0294] Step 3: Establish prediction models for Class I and Class II reservoirs respectively, and perform intelligent prediction of single-layer productivity for different types of reservoirs;

[0295] Step 4: For each combination of sub-layers in the joint trial production, quantitatively calculate the comprehensive coefficient or its component coefficients that characterize the intensity of interlayer interference. The component coefficients include the pressure difference interference coefficient, the gas-water relative permeability difference interference coefficient, the gravity interference coefficient, and the interlayer gas-water relationship interference coefficient.

[0296] Step 5: Intelligent prediction of gas well combined test production capacity is carried out using both mechanism-driven and data-driven methods.

[0297] In step 1, the logging curve data includes natural gamma, resistivity, sonic transit time, neutron porosity, and density; the key reservoir parameters include porosity, permeability, gas saturation, water saturation, clay content, and mineral composition; the basic data includes depth, thickness, stratigraphic position, and various physical properties such as porosity, permeability, gas saturation, water saturation, and clay content.

[0298] In step 2, the lower limit standards for reservoir physical properties and gas content include the lower limit of porosity, the lower limit of permeability, and the lower limit of gas saturation. The reservoir classification standard is formulated according to industry standards, referring to the SY / T6832-2020 oil and gas reservoir evaluation method, which mainly divides reservoirs into Class I, Class II, Class IIIA, Class IIIB, and Class IV based on electrical properties, physical properties, and gas content.

[0299] Step 3, for the construction and prediction of the production capacity prediction model for Class I reservoirs, includes: selecting parameters closely related to the production capacity of Class I reservoirs as input features of the model, using the daily gas production and daily water production from historical single-layer test or production data as output labels, selecting a machine learning algorithm, training the model using historical Class I reservoir sample data, inputting the corresponding features of the target well's Class I reservoir sub-layer into the trained model, and predicting its theoretical daily gas production and daily water production per layer; for the construction and prediction of the production capacity prediction model for Class II reservoirs, includes: selecting suitable feature parameters as input features of the model, using the historical daily gas production and daily water production from single layers as output labels, selecting a machine learning algorithm suitable for the characteristics of Class II reservoir data, training an independent model using historical Class II reservoir sample data, inputting the corresponding features of the target well's Class II reservoir sub-layer into the trained Class II model, and predicting its theoretical daily gas production and daily water production per layer.

[0300] The parameters closely related to the productivity of Class I reservoirs include porosity, permeability, gas saturation, effective thickness, and formation pressure; the characteristic parameters adapted to Class II reservoirs include porosity, permeability, gas saturation, effective thickness, resistivity response to low saturation, acoustic transit time indicative of pore throat structure, or other new features; the machine learning algorithms include random forest, gradient boosting tree, support vector machine, and BP neural network.

[0301] In step 4, calculating the pressure difference interference coefficient includes: collecting static or dynamic pressure data for each sublayer, obtaining the magnitude and distribution of pore pressure in each sublayer, defining the interlayer pressure difference, calculating the interlayer fluid flow trend and intensity caused by the pressure difference based on the extended form of Darcy's law, defining the interlayer interference coefficient, and normalizing it to the [0, 1] interval; calculating the gas-water relative permeability difference interference coefficient includes: obtaining or estimating the gas phase relative permeability and water phase relative permeability curves of each sublayer through core experiments or empirical models, and analyzing the differences in gas phase relative permeability and water phase relative permeability in different sublayers under combined production conditions. The differences in permeability ratios lead to differences in air and water flow capacity. The interlayer interference coefficient caused by these relative permeability differences is calculated. The gravity interference coefficient is calculated by: calculating the depth difference between the combined test layers, using the hydrostatic pressure formula to calculate the pressure difference caused by gravity between layers, and converting the pressure difference caused by gravity into a quantitative coefficient affecting interlayer crossflow and production dynamics. The interlayer air-water relationship interference coefficient is calculated by: analyzing the air-water distribution relationship of the combined test layers, assessing crossflow risk, and calculating the interference coefficient caused by specific air-water relationship patterns. The comprehensive interlayer interference coefficient is a weighted sum of the above four coefficients.

[0302] In step 5, the mechanism-driven method includes: using the single-layer production capacity of each sub-layer predicted in step 3, which represents the theoretical production capacity under the condition of no inter-layer interference, and based on the inter-layer interference coefficient obtained in step 4, combined with the seepage mechanics mechanism, establishing a quantitative model of production capacity loss caused by interference, and superimposing the effective production capacity of each sub-layer after the interference coefficient correction to obtain the final combined test production capacity prediction result; the data-driven method includes: collecting combined test production data samples of historical gas wells, constructing a machine learning model, determining the input and output of the model, using the collected historical sample data to train the constructed model, and for the target well, completing steps 1 to 4, integrating the acquired data as input, inputting it into the trained combined test production capacity intelligent prediction model, and the model directly outputting the predicted daily gas production and daily water production of the well.

[0303] The quantitative model for capacity loss caused by disturbance in the mechanism-driven method is as follows:

[0304] Qg_effective_i=Qg_Pred_i*f(α_total);

[0305] Qw_effective_i=Qw_Pred_i*g(α_total);

[0306] Where Qg_effective_i and Qw_effective_i represent the effective gas production capacity and effective water production capacity of each sublayer after considering interference, respectively; Qg_Pred_i and Qw_Pred_i represent the theoretical single-layer gas production capacity and water production capacity of each sublayer, respectively; f and g are functions reflecting the interference loss; the machine learning model constructed in the data-driven method includes multilayer perceptron, convolutional neural network, long short-term memory network or other advanced regression model.

[0307] Of course, it is understood that there may be other variations of the above detailed process, and all such variations should fall within the protection scope of this invention.

[0308] In this embodiment of the invention, based on the acquired logging curve data of the target gas well, multiple target sub-layers of the target gas well reservoir are divided, and multiple reservoir sub-layer parameters for each target sub-layer are determined. For different types of target sub-layers, corresponding production prediction machine learning models are constructed. These production prediction machine learning models are used to predict the daily gas production and daily water production of a single layer, using the reservoir sub-layer parameters as input features. The input features of the production prediction machine learning models are preset types of reservoir sub-layer parameters corresponding to the reservoir type. For multiple target sub-layers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer... The gas-water distribution relationship is used to determine the interference coefficient, which characterizes the inter-layer interference of multiple target sub-layers. The interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sub-layers. The natural gas production capacity prediction of the combined test of the target sub-layers is performed as follows: based on the interference coefficient, the daily gas production and daily water production of each target sub-layer, the combined test production capacity prediction result of the logging sub-layer after deducting interference losses is obtained; and / or, a deep learning algorithm is used to establish a combined test production capacity prediction model of the logging sub-layer based on various reservoir sub-layer parameters, daily gas production, daily water production and the interference coefficient of the target sub-layer; the combined test production capacity prediction result of the corresponding target sub-layer is output through the logging sub-layer production capacity prediction model. This invention addresses the reservoir description distortion problem caused by neglecting the heterogeneity of reservoir layers in traditional methods by utilizing multiple reservoir layer parameters for each target layer. It also solves the problem of existing single-layer prediction models being incompatible with reservoir quality differences by constructing corresponding production prediction machine learning models for different types of target layers, thus improving the accuracy of single-layer prediction for gas well reservoir layers. Furthermore, it determines the interference coefficient characterizing interlayer interference among multiple target layers based on the pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship, thus solving the problem of neglecting key interference factors such as interlayer crossflow and water intrusion risk in existing technologies. A mechanism-driven method is used to establish an interference loss correction function, subtracting interference loss from the theoretical production capacity of a single layer and then superimposing the effective production capacity. Finally, a data-driven method is used to construct an end-to-end model based on deep learning algorithms, integrating layer characteristic parameters, single-layer production capacity prediction values, and interference coefficients to directly output the combined testing results. This achieves high-precision intelligent prediction of combined logging layer production capacity, thereby improving the efficiency and accuracy of natural gas combined testing well production capacity prediction.

[0309] As described above, this invention belongs to the field of natural gas exploration and development technology, specifically relating to an intelligent prediction method for natural gas production capacity levels in multi-layer well testing. This method is used to accurately predict reservoir production capacity during multi-layer well testing, providing a scientific basis for the formulation of multi-layer natural gas production plans. This invention provides an intelligent prediction method for natural gas production capacity in multi-layer well testing, addressing the problems of poor accuracy and low efficiency in predicting reservoir production capacity in multi-layer well testing. The method first performs fine-grained sub-layer division based on well logging curve data and extracts key reservoir parameters. Then, it classifies the reservoirs according to industry standards and uses machine learning models to predict the daily gas and water production of individual layers. It considers the influence of pore pressure difference, gas-water permeability difference, depth difference, and gas-water relationship on inter-layer interference, and establishes a quantitative calculation model to obtain a comprehensive interference coefficient. Finally, it achieves high-precision intelligent prediction of multi-layer production capacity through two methods: mechanism-driven (based on seepage theory to deduct interference losses) and data-driven (direct prediction by machine learning models). Compared with traditional methods, this invention significantly improves prediction accuracy and computational efficiency, and can be widely applied to the optimization of multi-layer natural gas production plans and production decisions.

[0310] The embodiments of the present invention have the following beneficial effects:

[0311] (1) Refinement and accuracy: Through fine division of small layers, classification of reservoirs and targeted modeling, the accuracy of single-layer production capacity prediction has been significantly improved, especially the prediction capability of Class I and Class II reservoirs in complex gas reservoirs.

[0312] (2) Quantification of interlayer interference: The complex interlayer interference is decomposed into quantifiable physical mechanisms (pressure, phase permeability, gravity, gas-water relationship), which provides a physical basis for understanding the nature of interference and accurately correcting production capacity, and overcomes the blindness of empirical coefficients.

[0313] (3) High accuracy and efficiency in combined test prediction: The mechanism-driven method has clear physical meaning and the results can be interpreted; the data-driven method can capture complex nonlinear relationships and has high prediction efficiency. The dual-track prediction complements each other and greatly improves the accuracy, reliability and practicality of combined test capacity prediction.

[0314] (4) Guiding development decisions: Accurate combined test production capacity prediction results and quantitative inter-layer interference information provide key decision-making basis for optimizing gas well production system (such as production pressure difference and production allocation), adjusting layer combination, and tapping potential measures (such as layered fracturing and water shut-off), which helps to improve the development efficiency of gas fields.

[0315] (5) Automation and intelligence: The entire process is easy to automate and can be efficiently applied to the evaluation of a large number of gas wells.

[0316] This invention also provides an intelligent prediction device for natural gas production capacity from well logging small-layer combined testing, as described in the following embodiments. Since the principle behind this device is similar to the intelligent prediction method for natural gas production capacity from well logging small-layer combined testing, its implementation can refer to the implementation of the intelligent prediction method for natural gas production capacity from well logging small-layer combined testing; repeated details will not be elaborated further.

[0317] Figure 6 This is a structural example diagram of a natural gas production capacity intelligent prediction device for well logging small-layer combined testing according to an embodiment of the present invention. The embodiment of the present invention also provides a natural gas production capacity intelligent prediction device for well logging small-layer combined testing to improve the efficiency and accuracy of natural gas combined testing well production capacity prediction, such as... Figure 6 As shown, the device includes:

[0318] The target sub-layer division module 601 is used to divide the target gas well reservoir into multiple target sub-layers based on the acquired multiple logging curve data of the target gas well, and to determine multiple reservoir sub-layer parameters for each target sub-layer.

[0319] The production capacity prediction machine learning model construction module 602 is used to construct corresponding production capacity prediction machine learning models for different types of target sub-layers; the production capacity prediction machine learning model is used to predict the daily gas production and daily water production of a single layer using the reservoir sub-layer parameters of the target sub-layer as input features; the input features of the production capacity prediction machine learning model are preset types of reservoir sub-layer parameters corresponding to the reservoir type.

[0320] The interference coefficient determination module 603 is used to determine the interference coefficient, which characterizes the interlayer interference of multiple target sublayers, based on the pore pressure difference, the difference in relative permeability of gas and water, the gravity effect caused by the depth difference, and the interlayer gas and water distribution relationship between the multiple target sublayers. The interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sublayers.

[0321] The capacity forecasting module 604 is used to forecast the natural gas production capacity of the target small-layer combined test as follows:

[0322] Based on the interference coefficient, the daily gas production and daily water production of each target sublayer, the combined testing capacity prediction result of the logging sublayer after deducting interference losses is obtained; and / or, using a deep learning algorithm, a combined testing capacity prediction model of the logging sublayer is established based on various reservoir sublayer parameters, daily gas production, daily water production and the interference coefficient of the target sublayer; through the logging sublayer capacity prediction model, the combined testing capacity prediction result of the corresponding target sublayer is output.

[0323] In one embodiment, the various logging curve data include natural gamma, resistivity, sonic transit time, neutron porosity, and density curves; the various reservoir layer parameters include porosity, permeability, gas saturation, water saturation, clay content, and mineral composition obtained through logging interpretation models.

[0324] The capacity forecasting machine learning model building module is specifically used for:

[0325] Based on the distribution characteristics of marker layers, sedimentary cycles, and lithological abrupt change interfaces in the regional geological stratification standard, vertical stratigraphy is divided according to the changes in well logging curve morphology, and the top and bottom depths, thicknesses, and various reservoir sub-layer parameters of each target sub-layer are determined.

[0326] In one embodiment, it also includes:

[0327] Based on the set lower limits of porosity, permeability, and gas saturation, the effective target sublayers are obtained by excluding the dense layers and mudstone layers in the target sublayers.

[0328] Based on electrical resistivity, physical properties porosity and permeability, and gas-bearing properties and gas saturation, the effective target layers are divided into Class I reservoirs and Class II reservoirs.

[0329] In one embodiment, the input features of the production capacity prediction machine learning model constructed for the first type of reservoir are porosity, permeability, gas saturation, thickness, and formation pressure; the input features of the production capacity prediction machine learning model constructed for the second type of reservoir are porosity, permeability, gas saturation, thickness, resistivity, and sonic transit time ratio.

[0330] In one embodiment, corresponding capacity prediction machine learning models are constructed for different types of target sub-layers, including:

[0331] By employing random forest, gradient boosting tree, or XGBoost algorithms, a corresponding production capacity prediction machine learning model is constructed by establishing a mapping relationship between various reservoir sublayer parameter features and production capacity of the target sublayer.

[0332] In one embodiment, the interference coefficient determination module is specifically used for:

[0333] Based on the pore pressure difference, the interlayer fluid channeling trend is calculated by an extended form of Darcy's law, and a pressure difference interference coefficient is generated.

[0334] Based on the difference in relative permeability between air and water, the differences in flow capacity of different sub-layers under a fixed water saturation are analyzed by obtaining the air-water relative permeability curves of each sub-layer, and the interference coefficient of the difference in relative permeability between air and water is calculated.

[0335] The depth difference interference coefficient is calculated based on the gravitational potential energy difference determined by the gravitational effect caused by the depth difference.

[0336] Based on the interlayer air-water distribution relationship, the interference coefficient of the interlayer air-water relationship is calculated using a crossflow risk assessment model.

[0337] The interference coefficients characterizing the interlayer interference of multiple target sublayers are obtained by weighting the interference coefficients of pressure difference, gas-water relative permeability difference, depth difference, and interlayer gas-water relationship.

[0338] In one embodiment, the capacity forecasting module is specifically used for:

[0339] An interference loss correction function is established based on the interference coefficient; the inter-layer interference loss is deducted from the predicted values ​​of daily gas production and daily water production of each target sub-layer according to the interference loss correction function, and the effective daily gas production and effective daily water production of each sub-layer are obtained.

[0340] The effective daily gas production and effective daily water production of each sub-layer are superimposed to generate the combined test production capacity prediction result of the well logging sub-layer.

[0341] In one embodiment, the capacity forecasting module is specifically used for:

[0342] Using the reservoir layer parameters, daily gas production, daily water production, and interference coefficient of historical gas wells as input features, and the actual combined test daily gas production and actual combined test daily water production as output labels, a training sample set is constructed.

[0343] A multilayer perceptron, convolutional neural network, or long short-term memory network is used to train a sample set to obtain a well logging small-layer combined test production capacity prediction model.

[0344] This invention provides an embodiment of a computer device for implementing all or part of the above-mentioned intelligent prediction method for natural gas production capacity in well logging small-layer combined testing. The computer device specifically includes the following components:

[0345] The computer device comprises a processor, memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between related devices; the computer device can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the computer device can be implemented with reference to the embodiments of the intelligent prediction method for natural gas production capacity of well logging small-layer combined testing and the embodiments of the intelligent prediction device for natural gas production capacity of well logging small-layer combined testing, the contents of which are incorporated herein, and repeated parts will not be described again.

[0346] Figure 7 This is a schematic diagram of a computer device provided in an embodiment of the present invention, which discloses a schematic block diagram of the system configuration of a computer device 1000 according to an embodiment of this application. Figure 7 As shown, the computer device 1000 may include a central processing unit 1001 and a memory 1002; the memory 1002 is coupled to the central processing unit 1001. It is worth noting that... Figure 7 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0347] In one embodiment, the intelligent prediction function of natural gas production capacity from well logging layer fusion testing can be integrated into the central processing unit 1001. The central processing unit 1001 can be configured to perform the following control:

[0348] Based on the various logging curve data of the target gas well, the reservoir of the target gas well is divided into multiple target sub-layers, and various reservoir sub-layer parameters of each target sub-layer are determined.

[0349] For different types of target sublayers, corresponding production capacity prediction machine learning models are constructed respectively; the production capacity prediction machine learning models are used to predict the daily gas production and daily water production of a single layer using the reservoir sublayer parameters of the target sublayer as input features; the input features of the production capacity prediction machine learning models are preset types of reservoir sublayer parameters corresponding to the reservoir type.

[0350] For multiple target sublayers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship between the multiple target sublayers, an interference coefficient characterizing the interlayer interference of multiple target sublayers is determined; the interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sublayers.

[0351] The natural gas production capacity forecast for the target small-scale strata test is performed as follows:

[0352] Based on the interference coefficient, the daily gas production and daily water production of each target sublayer, the combined testing capacity prediction result of the logging sublayer after deducting interference losses is obtained; and / or, using a deep learning algorithm, a combined testing capacity prediction model of the logging sublayer is established based on various reservoir sublayer parameters, daily gas production, daily water production and the interference coefficient of the target sublayer; through the logging sublayer capacity prediction model, the combined testing capacity prediction result of the corresponding target sublayer is output.

[0353] In another embodiment, the intelligent prediction device for natural gas production capacity of well logging layer combined test can be configured separately from the central processing unit 1001. For example, the intelligent prediction device for natural gas production capacity of well logging layer combined test can be configured as a chip connected to the central processing unit 1001, and the intelligent prediction function for natural gas production capacity of well logging layer combined test can be realized through the control of the central processing unit.

[0354] like Figure 7 As shown, the computer device 1000 may further include: a communication module 1003, an input unit 1004, an audio processor 1005, a display 1006, and a power supply 1007. It is worth noting that the computer device 1000 does not necessarily need to include... Figure 7 All components shown; in addition, the computer device 1000 may also include Figure 7 For components not shown, please refer to existing technologies.

[0355] like Figure 7 As shown, the central processing unit 1001, sometimes also referred to as a controller or operation control, may include a microprocessor or other processor device and / or logic device. The central processing unit 1001 receives input and controls the operation of various components of the computer device 1000.

[0356] The memory 1002 may be, for example, one or more of a cache, flash memory, hard drive, removable medium, volatile memory, non-volatile memory, or other suitable device. It can store the aforementioned device-related information, and may also store programs for executing that information. The central processing unit 1001 can execute the program stored in the memory 1002 to perform information storage or processing, etc.

[0357] Input unit 1004 provides input to central processing unit 1001. This input unit 1004 may be, for example, a keypad or touch input device. Power supply 1007 provides power to computer device 1000. Display 1006 displays images, text, and other display objects. This display may be, for example, an LCD display, but is not limited to this.

[0358] The memory 1002 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 1002 can also be some other type of device. The memory 1002 includes a buffer memory 1021 (sometimes referred to as a buffer). The memory 1002 may include an application / function storage unit 1022 for storing application programs and function programs or processes for executing operations of the computer device 1000 via the central processing unit 1001.

[0359] The memory 1002 may also include a data storage unit 1023 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the computer device. The driver storage unit 1024 of the memory 1002 may include various drivers for the computer device for communication functions and / or for performing other functions of the computer device (such as messaging applications, address book applications, etc.).

[0360] The communication module 1003 is a transmitter / receiver that transmits and receives signals via the antenna 1008. The communication module (transmitter / receiver) 1003 is coupled to the central processing unit 1001 to provide input signals and receive output signals, which is the same as in a conventional mobile communication terminal.

[0361] Based on different communication technologies, multiple communication modules 1003 can be configured in the same computer device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module (transmitter / receiver) 1003 is also coupled to a speaker 1009 and a microphone 1010 via an audio processor 1005 to provide audio output via the speaker 1009 and receive audio input from the microphone 1010, thereby realizing typical telecommunications functions. The audio processor 1005 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 1005 is also coupled to a central processing unit 1001, enabling on-device recording via the microphone 1010 and on-device playback of stored sound via the speaker 1009.

[0362] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent prediction method for natural gas production capacity from well logging small-layer combined testing.

[0363] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-mentioned intelligent prediction method for natural gas production capacity from well logging small-layer combined testing.

[0364] In this embodiment of the invention, based on the acquired logging curve data of the target gas well, multiple target sub-layers of the target gas well reservoir are divided, and multiple reservoir sub-layer parameters for each target sub-layer are determined. For different types of target sub-layers, corresponding production prediction machine learning models are constructed. These production prediction machine learning models are used to predict the daily gas production and daily water production of a single layer, using the reservoir sub-layer parameters as input features. The input features of the production prediction machine learning models are preset types of reservoir sub-layer parameters corresponding to the reservoir type. For multiple target sub-layers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer... The gas-water distribution relationship is used to determine the interference coefficient, which characterizes the inter-layer interference of multiple target sub-layers. The interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sub-layers. The natural gas production capacity prediction of the combined test of the target sub-layers is performed as follows: based on the interference coefficient, the daily gas production and daily water production of each target sub-layer, the combined test production capacity prediction result of the logging sub-layer after deducting interference losses is obtained; and / or, a deep learning algorithm is used to establish a combined test production capacity prediction model of the logging sub-layer based on various reservoir sub-layer parameters, daily gas production, daily water production and the interference coefficient of the target sub-layer; the combined test production capacity prediction result of the corresponding target sub-layer is output through the logging sub-layer production capacity prediction model. This invention addresses the reservoir description distortion problem caused by neglecting the heterogeneity of reservoir layers in traditional methods by utilizing multiple reservoir layer parameters for each target layer. It also solves the problem of existing single-layer prediction models being incompatible with reservoir quality differences by constructing corresponding production prediction machine learning models for different types of target layers, thus improving the accuracy of single-layer prediction for gas well reservoir layers. Furthermore, it determines the interference coefficient characterizing interlayer interference among multiple target layers based on the pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship, thus solving the problem of neglecting key interference factors such as interlayer crossflow and water intrusion risk in existing technologies. A mechanism-driven method is used to establish an interference loss correction function, subtracting interference loss from the theoretical production capacity of a single layer and then superimposing the effective production capacity. Finally, a data-driven method is used to construct an end-to-end model based on deep learning algorithms, integrating layer characteristic parameters, single-layer production capacity prediction values, and interference coefficients to directly output the combined testing results. This achieves high-precision intelligent prediction of combined logging layer production capacity, thereby improving the efficiency and accuracy of natural gas combined testing well production capacity prediction.

[0365] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0366] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0367] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0368] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0369] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent prediction of natural gas production capacity from well logging small-layer combined testing, characterized in that, include: Based on the various logging curve data of the target gas well, the reservoir of the target gas well is divided into multiple target sub-layers, and various reservoir sub-layer parameters of each target sub-layer are determined. For different types of target sublayers, corresponding production capacity prediction machine learning models are constructed respectively; the production capacity prediction machine learning models are used to predict the daily gas production and daily water production of a single layer using the reservoir sublayer parameters of the target sublayer as input features; the input features of the production capacity prediction machine learning models are preset types of reservoir sublayer parameters corresponding to the reservoir type. For multiple target sublayers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship between the multiple target sublayers, an interference coefficient characterizing the interlayer interference of multiple target sublayers is determined; the interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sublayers. The natural gas production capacity forecast for the target small-scale strata test is performed as follows: Based on the interference coefficient, the daily gas production and daily water production of each target sublayer, the combined testing capacity prediction result of the logging sublayer after deducting interference losses is obtained; and / or, using a deep learning algorithm, a combined testing capacity prediction model of the logging sublayer is established based on various reservoir sublayer parameters, daily gas production, daily water production and the interference coefficient of the target sublayer; through the logging sublayer capacity prediction model, the combined testing capacity prediction result of the corresponding target sublayer is output.

2. The method as described in claim 1, characterized in that, The various logging curve data include natural gamma, resistivity, sonic transit time, neutron porosity, and density curves; the various reservoir layer parameters include porosity, permeability, gas saturation, water saturation, clay content, and mineral composition obtained through logging interpretation models. Based on the acquired logging data of various target gas wells, the reservoir of the target gas well is divided into multiple target sub-layers, and various reservoir sub-layer parameters for each target sub-layer are determined, including: Based on the distribution characteristics of marker layers, sedimentary cycles, and lithological abrupt change interfaces in the regional geological stratification standard, vertical stratigraphy is divided according to the changes in well logging curve morphology, and the top and bottom depths, thicknesses, and various reservoir sub-layer parameters of each target sub-layer are determined.

3. The method as described in claim 1, characterized in that, Also includes: Based on the set lower limits of porosity, permeability, and gas saturation, the effective target sublayers are obtained by excluding the dense layers and mudstone layers in the target sublayers. Based on electrical resistivity, physical properties porosity and permeability, and gas-bearing properties and gas saturation, the effective target layers are divided into Class I reservoirs and Class II reservoirs.

4. The method as described in claim 1, characterized in that, The input features for the production capacity prediction machine learning model for the first type of reservoir are porosity, permeability, gas saturation, thickness, and formation pressure; the input features for the production capacity prediction machine learning model for the second type of reservoir are porosity, permeability, gas saturation, thickness, resistivity, and sonic transit time ratio.

5. The method as described in claim 1, characterized in that, For different types of target sub-layers, corresponding capacity prediction machine learning models are constructed, including: By employing random forest, gradient boosting tree, or XGBoost algorithms, a corresponding production capacity prediction machine learning model is constructed by establishing a mapping relationship between various reservoir sublayer parameter features and production capacity of the target sublayer.

6. The method as described in claim 1, characterized in that, For multiple target sublayers, based on the determined pore pressure difference, gas-water relative permeability difference, gravity effect caused by depth difference, and interlayer gas-water distribution relationship between the multiple target sublayers, the interference coefficient characterizing the interlayer interference of the multiple target sublayers is determined, including: Based on the pore pressure difference, the interlayer fluid channeling trend is calculated by an extended form of Darcy's law, and a pressure difference interference coefficient is generated. Based on the difference in relative permeability between air and water, the differences in flow capacity of different sub-layers under a fixed water saturation are analyzed by obtaining the air-water relative permeability curves of each sub-layer, and the interference coefficient of the difference in relative permeability between air and water is calculated. The depth difference interference coefficient is calculated based on the gravitational potential energy difference determined by the gravitational effect caused by the depth difference. Based on the interlayer air-water distribution relationship, the interference coefficient of the interlayer air-water relationship is calculated using a crossflow risk assessment model. The interference coefficients characterizing the interlayer interference of multiple target sublayers are obtained by weighting the interference coefficients of pressure difference, gas-water relative permeability difference, depth difference, and interlayer gas-water relationship.

7. The method as described in claim 1, characterized in that, Based on the interference coefficient, the daily gas production and daily water production of each target sub-layer, the combined test production capacity prediction results of the logging sub-layer after deducting interference losses are obtained, including: An interference loss correction function is established based on the interference coefficient; the inter-layer interference loss is deducted from the predicted values ​​of daily gas production and daily water production of each target sub-layer according to the interference loss correction function, and the effective daily gas production and effective daily water production of each sub-layer are obtained. The effective daily gas production and effective daily water production of each sub-layer are superimposed to generate the combined test production capacity prediction result of the well logging sub-layer.

8. The method as described in claim 1, characterized in that, Using a deep learning algorithm, a well logging test production prediction model is established based on various reservoir parameters of the target sub-layer, daily gas production of a single layer, daily water production of a single layer, and the aforementioned interference coefficient. This model includes: Using the reservoir layer parameters, daily gas production, daily water production, and interference coefficient of historical gas wells as input features, and the actual combined test daily gas production and actual combined test daily water production as output labels, a training sample set is constructed. A multilayer perceptron, convolutional neural network, or long short-term memory network is used to train a sample set to obtain a well logging small-layer combined test production capacity prediction model.

9. A smart predictive device for natural gas production capacity from well logging small-layer combined testing, characterized in that, include: The target sub-layer division module is used to divide the target gas well reservoir into multiple target sub-layers based on the acquired multiple logging curve data of the target gas well, and to determine multiple reservoir sub-layer parameters for each target sub-layer. The production capacity prediction machine learning model construction module is used to construct corresponding production capacity prediction machine learning models for different types of target sub-layers. The production capacity prediction machine learning model is used to predict the daily gas production and daily water production of a single layer using the reservoir sub-layer parameters of the target sub-layer as input features. The input features of the production capacity prediction machine learning model are preset types of reservoir sub-layer parameters corresponding to the reservoir type. The interference coefficient determination module is used to determine the interference coefficient, which characterizes the interlayer interference of multiple target sublayers, based on the pore pressure difference, the difference in relative permeability of gas and water, the gravity effect caused by the depth difference, and the interlayer gas and water distribution relationship between the multiple target sublayers. The interference coefficient is a quantitative parameter that quantitatively characterizes the mutual interference intensity between the combined test sublayers. The capacity forecasting module is used to forecast the natural gas production capacity of the target small-layer combined test as follows: Based on the interference coefficient, the daily gas production and daily water production of each target sublayer, the combined testing capacity prediction result of the logging sublayer after deducting interference losses is obtained; and / or, using a deep learning algorithm, a combined testing capacity prediction model of the logging sublayer is established based on various reservoir sublayer parameters, daily gas production, daily water production and the interference coefficient of the target sublayer; through the logging sublayer capacity prediction model, the combined testing capacity prediction result of the corresponding target sublayer is output.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 8.

12. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 8.

Images