Method and system for evaluating quality of a thermal reservoir

Abstract

This disclosure relates to a method and system for evaluating the quality of thermal reservoirs, belonging to the field of thermal reservoir quality evaluation technology. The method includes: obtaining reservoir quality evaluation parameters corresponding to known conditions and thermal reservoir quality evaluations corresponding to the reservoir quality evaluation parameters; training a generator network using the reservoir quality evaluation parameters, random noise under the conditions of the reservoir quality evaluation parameters, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters to obtain a generator; and performing thermal reservoir quality evaluation using the generator based on the reservoir quality evaluation parameters to be evaluated. Embodiments of this disclosure can realize thermal reservoir quality evaluation.

Description

Technical Field

[0001] This disclosure relates to the field of thermal reservoir quality evaluation technology, and in particular to a method and system for thermal reservoir quality evaluation. Background Technology

[0002] Geothermal energy, as a natural, renewable, and clean energy source, has advantages such as being green, low-carbon, and sustainable. It is of great value in achieving the "dual carbon" goal and has become a key new energy source for global development.

[0003] With the increasing depletion of traditional energy sources, many oilfields have entered the medium-to-high water-cut stage. Oilfield areas typically possess high geothermal gradients and rich reservoir data, making the development of geothermal energy in oilfields crucial for the green and low-carbon transformation of older oilfields. This is especially true for identifying geothermal fluids from oil and natural gas, thereby reducing the cost of geothermal water treatment. Based on reservoir identification, further evaluation of reservoir quality plays a vital role in the development of geothermal energy in oilfields. It enables the optimization of development plans based on the heat storage capacity of different reservoirs, ensuring the stable utilization of geothermal resources and ultimately achieving the green and low-carbon transformation of oilfields.

[0004] Reservoir quality evaluation mainly employs parametric comprehensive evaluation methods and artificial intelligence methods. These methods primarily examine reservoir physical properties such as porosity, permeability, flow zone indices reflecting reservoir seepage characteristics, and reservoir thickness reflecting reservoir scale. Multiple approaches are used to determine the parameter limits for different reservoirs. Conventional reservoir quality evaluation methods are relatively simple and do not provide a comprehensive characterization of reservoir quality.

[0005] Existing artificial intelligence methods mostly employ discriminative networks, relying on large-scale data to learn the mapping relationship between features and labels to construct decision boundaries for different categories. However, these methods often underutilize parametric features when the amount of data is small or the samples are imbalanced, and they cannot utilize existing labeled data. Summary of the Invention

[0006] This disclosure presents a technical solution for a method and system for evaluating the quality of thermal reservoirs.

[0007] According to one aspect of this disclosure, a method for evaluating the quality of a thermal reservoir is provided, comprising:

[0008] Obtain the reservoir quality evaluation parameters corresponding to the known conditions and the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters.

[0009] The generator network is trained using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameters, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters to obtain the generator.

[0010] The generator is used to evaluate the quality of the thermal reservoir based on the reservoir quality evaluation parameters to be evaluated.

[0011] Preferably, the step of training the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters to obtain the generator includes: generating a pseudo-evaluation of the thermal reservoir quality based on the generator network using the reservoir quality evaluation parameters and random noise under the reservoir quality evaluation parameter conditions; determining the truth or falsehood of the pseudo-evaluation of the thermal reservoir quality based on the discrimination network using the reservoir quality evaluation parameters, the pseudo-evaluation of the thermal reservoir quality, and the thermal reservoir quality evaluation; calculating a first loss of the pseudo-evaluation of the thermal reservoir quality corresponding to the random noise under the reservoir quality evaluation parameter conditions and a second loss corresponding to determining the truth or falsehood of the pseudo-evaluation of the thermal reservoir quality, and training the generator network to obtain the generator.

[0012] Preferably, the calculation of the first loss for the pseudo-evaluation of thermal reservoir quality corresponding to random noise under the reservoir quality evaluation parameter conditions and the second loss for determining whether the pseudo-evaluation of thermal reservoir quality is true or false includes: constructing a first loss function corresponding to the first loss and a second loss function corresponding to the second loss; calculating the first loss for the pseudo-evaluation of thermal reservoir quality corresponding to random noise under the reservoir quality evaluation parameter conditions based on the first loss function; and calculating the second loss for determining whether the pseudo-evaluation of thermal reservoir quality is true or false based on random noise under the reservoir quality evaluation parameter conditions based on the second loss function.

[0013] Preferably, constructing the first loss function corresponding to the first loss includes: calculating the difference between a set value and a pseudo-evaluation of the thermal reservoir quality corresponding to random noise under the reservoir quality evaluation parameter conditions; taking the logarithm of the difference to obtain the logarithmic value of the difference; calculating the first expected value of the logarithmic value of the difference, and constructing the first loss function corresponding to the first loss.

[0014] Preferably, the set value is configured to be 1.

[0015] Preferably, constructing the first loss function corresponding to the first loss includes: taking the logarithm of the discrimination probability value corresponding to the random noise under the reservoir quality evaluation parameter conditions to determine whether the thermal reservoir quality is false or true, to obtain the discrimination logarithm value; calculating the second expected value of the logarithm value, and constructing the second loss function corresponding to the second loss.

[0016] Preferably, the step of calculating the first loss corresponding to the pseudo-evaluation of the thermal reservoir quality under the conditions of the reservoir quality evaluation parameters and the second loss corresponding to the discrimination of the pseudo-evaluation of the thermal reservoir quality as true or false, and training the generator network to obtain the generator, includes: constructing objective functions corresponding to the generator network and the discrimination network based on the first loss function corresponding to the first loss and the second loss function corresponding to the second loss; maximizing the first expected value of the first loss function corresponding to the first loss in the objective function and minimizing the second expected value of the second loss function corresponding to the second loss in the objective function, and training the generator network to obtain the generator.

[0017] Preferably, the generating network includes: a first shallow feature extraction module, an encoder connected to the shallow feature extraction module, and a decoder connected to the shallow feature extraction module and the encoder respectively; wherein, the shallow feature extraction module is used to extract the reservoir quality evaluation parameters and a first shallow feature of random noise under the conditions of the reservoir quality evaluation parameters; the encoder is used to extract a first deep feature and a second deep feature corresponding to the first shallow feature; the decoder is used to generate a pseudo evaluation of the thermal reservoir quality based on the first shallow feature and its corresponding first deep feature and second deep feature.

[0018] Preferably, the shallow feature extraction module includes: a first convolutional neural unit for extracting the reservoir quality evaluation parameters and the first shallow feature corresponding to random noise under the conditions of the reservoir quality evaluation parameters.

[0019] Preferably, the first convolutional neural unit includes at least one one-dimensional convolutional layer or multiple cascaded one-dimensional convolutional layers.

[0020] Preferably, the encoder includes: a first dense feature extraction module and a second dense feature extraction module connected to the first dense feature extraction module; wherein, the first dense feature extraction module includes: a second convolutional neural unit, a third convolutional neural unit connected to the second convolutional neural unit, the first batch of normalization units connected to the third convolutional neural unit, and a first activation function connected to the first batch of normalization units; wherein, the input of the second convolutional neural unit is configured as the first shallow feature, and the output of the first activation function is configured as the first deep feature; the second dense feature extraction module includes: a fourth convolutional neural unit, The system includes a fifth convolutional neural unit connected to the fourth convolutional neural unit, a second batch normalization unit connected to the fifth convolutional neural unit, and a second activation function connected to the second batch normalization unit. A first skip connection unit is configured between the input of the fourth convolutional neural unit, the input of the second convolutional neural unit, and the output of the third convolutional neural unit. The first skip connection unit is used to concatenate the first shallow feature and the processed feature output by the third convolutional neural unit to obtain a first concatenated feature. The first concatenated feature is configured as the input of the fourth convolutional neural unit, and the output of the second activation function is configured as the second deep feature.

[0021] Preferably, the second, third, fourth, and fifth convolutional neural units each include at least one one-dimensional convolutional layer or multiple cascaded one-dimensional convolutional layers.

[0022] Preferably, a second skip connection unit is configured between the input of the second convolutional neural unit and the input of the third convolutional neural unit; wherein, the second skip connection unit is used to concatenate the first shallow feature and the processed feature output by the second convolutional neural unit to obtain a second concatenated feature; the second concatenated feature is configured as the input of the third convolutional neural unit.

[0023] Preferably, a third skip connection unit is configured between the input of the fourth convolutional neural unit and the input of the fifth convolutional neural unit; wherein, the third skip connection unit is used to splice the spliced ​​feature with the processing feature output by the fourth convolutional neural unit to obtain a third spliced ​​feature; the third spliced ​​feature is configured as the input of the fifth convolutional neural unit.

[0024] Preferably, the decoder includes: a sixth convolutional neural unit, a feature self-attention module connected to the sixth convolutional neural unit, and a linear layer connected to the feature self-attention module; wherein, the sixth convolutional neural unit is used to perform convolution processing on the first shallow features to obtain second shallow features; the feature self-attention module is used to perform self-attention processing on the second shallow features, the first deep features, and the second deep features to obtain self-attention features; and the linear layer is used to perform linear processing on the self-attention features to generate the pseudo-evaluation of the thermal reservoir quality.

[0025] Preferably, the sixth convolutional neural unit includes at least one one-dimensional convolutional layer or multiple cascaded one-dimensional convolutional layers.

[0026] Preferably, the feature self-attention module includes: a channel attention module and a spatial attention module connected to the channel attention module; wherein, the channel attention module is used to perform channel attention processing on the second shallow feature, the first deep feature and the second deep feature to obtain channel attention features; the spatial attention module is used to perform spatial attention processing on the spatial attention features to obtain self-attention features.

[0027] Preferably, before the channel attention module performs channel attention processing on the second shallow feature, the first deep feature, and the second deep feature, the second shallow feature, the first deep feature, and the second deep feature are concatenated to obtain a concatenated feature to be processed; the channel attention module performs channel attention processing on the concatenated feature to be processed to obtain a channel attention feature.

[0028] Preferably, the channel attention module includes: a first pooling unit and a second pooling unit different from the first pooling unit, a shared multilayer perceptron connected to the first pooling unit and the second pooling unit respectively, an adder connected to the shared multilayer perceptron, a third activation function connected to the adder, and a first multiplier; wherein, the first pooling unit and the second pooling unit perform pooling operations on the second shallow feature, the first deep feature, and the concatenated feature to be processed corresponding to the second deep feature respectively to obtain a first pooling feature and a second pooling feature; the shared multilayer perceptron is used to perform weight processing on the first pooling feature and the second pooling feature respectively to obtain a first weighted feature and a second weighted feature; the adder is used to perform addition processing on the first weighted feature and the second weighted feature to obtain an added feature; the third activation function is used to perform nonlinear processing on the added feature to obtain a channel attention weight; and the first multiplier is used to perform multiplication processing on the channel attention weight and the concatenated feature to be processed to obtain a channel attention feature.

[0029] Preferably, the first pooling unit is configured as a maximum pooling layer, and the second pooling unit is configured as an average pooling layer.

[0030] Preferably, the spatial attention module includes: a third pooling unit, a fourth pooling unit connected to the third pooling unit, a convolutional layer connected to the fourth pooling unit, a fourth activation function connected to the convolutional layer, and a second multiplier; wherein, the third pooling unit and the fourth pooling unit are used sequentially to perform pooling processing on the channel attention features to obtain third pooling features; the convolutional layer is used to perform convolution processing on the third pooling features to obtain convolutional features; the fourth activation function is used to perform nonlinear processing on the convolutional features to obtain spatial attention weights; and the second multiplier is used to perform multiplication operations on the channel attention features and the spatial attention weights to obtain self-attention features.

[0031] Preferably, the third pooling unit is configured as a maximum pooling layer, and the fourth pooling unit is configured as an average pooling layer.

[0032] Preferably, the kernel size corresponding to the convolutional layer is configured as 7*7.

[0033] Preferably, the linear layer connected to the feature self-attention module includes: a first linear layer and a second linear layer connected to the first linear layer; wherein, the second linear layer is used to perform linear processing on the linear features after linear processing of the self-attention features by the first linear layer, and then perform linear processing again to generate the pseudo-evaluation of the thermal reservoir quality.

[0034] Preferably, the discriminant network includes: at least one convolutional correlation weight module or multiple cascaded convolutional correlation weight modules; wherein, the convolutional correlation weight module is used to train the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation corresponding to the reservoir quality evaluation parameters.

[0035] Preferably, the number of the multiple cascaded convolutional correlation weight modules is configured to be 3.

[0036] Preferably, each of the convolutional correlation weight modules includes: a seventh convolutional neural unit, a feature self-attention unit connected to the seventh convolutional neural unit, a third batch normalization unit connected to the feature self-attention unit, and a fifth activation function connected to the third batch normalization unit; wherein, the seventh convolutional neural unit is used to perform convolution processing on the reservoir quality evaluation parameters, the reservoir quality evaluation and reservoir quality pseudo-evaluation corresponding to the reservoir quality evaluation parameters, or the correlation weight features output by the previous level convolutional correlation weight module; to obtain the convolutional features to be discriminated; the feature self-attention unit is used to perform self-attention processing on the convolutional features to be discriminated, to obtain the self-attention features to be discriminated; the third batch normalization unit is used to perform batch normalization processing on the self-attention features to be discriminated, to obtain the batch normalized features to be discriminated; the fifth activation function is used to perform nonlinear processing on the batch normalized features to be discriminated, to obtain the discrimination result corresponding to each of the convolutional correlation weight modules.

[0037] Preferably, the seventh convolutional neural unit includes at least one one-dimensional convolutional layer or multiple cascaded one-dimensional convolutional layers.

[0038] Preferably, the discrimination result is configured as the true or false corresponding to the pseudo evaluation of the thermal reservoir quality and / or the evaluation of the thermal reservoir quality.

[0039] Preferably, a skip connection unit is provided between the inputs of each of the convolutional correlation weight modules; wherein, the skip connection unit is used to concatenate the discrimination result corresponding to the output of the previous level convolutional correlation weight module with the input features of the previous level convolutional correlation weight module to obtain the correlation weight concatenation features of the next level convolutional correlation weight module.

[0040] Preferably, determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters includes: calculating the variance explained rate and loading coefficients corresponding to the principal components using principal component analysis; calculating a linear combination coefficient matrix based on the variance explained rate and the loading coefficients; calculating a comprehensive score coefficient matrix corresponding to the reservoir quality evaluation parameters based on the eigenvalues ​​of the linear combination coefficient matrix and the covariance matrix corresponding to the reservoir quality evaluation parameters; normalizing the comprehensive score coefficient matrix to obtain the weight values ​​of each reservoir quality parameter; and determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters based on the reservoir quality evaluation parameters and their corresponding weight values.

[0041] Preferably, before calculating the comprehensive score coefficient matrix corresponding to the reservoir quality evaluation parameters based on the eigenvalues ​​of the linear combination coefficient matrix and the covariance matrix corresponding to the reservoir quality evaluation parameters, a vector feature corresponding to the reservoir quality evaluation parameters is constructed, and the vector feature is standardized to obtain a standardized vector feature; the comprehensive score coefficient matrix corresponding to the reservoir quality evaluation parameters is calculated based on the eigenvalues ​​of the linear combination coefficient matrix and the covariance matrix corresponding to the standardized vector feature.

[0042] Preferably, determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameter based on the reservoir quality evaluation parameter and its corresponding weight value of each reservoir quality parameter includes: multiplying the reservoir quality evaluation parameter by the corresponding weight value of each reservoir quality parameter and summing them to determine the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameter.

[0043] Preferably, the reservoir quality evaluation parameters corresponding to the known conditions include one or more of the following: clay content, fine sand content, silt content, porosity, permeability, and effective thickness of different reservoirs.

[0044] Preferably, before training the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters, the process includes: normalizing and / or standardizing the reservoir quality evaluation parameters to obtain normalized and / or standardized reservoir quality evaluation parameters; and training the generator network using the normalized and / or standardized reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters.

[0045] Preferably, before training the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters, a training set is defined using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation corresponding to the reservoir quality evaluation parameters; the generator network is then trained based on the reservoir quality evaluation parameters corresponding to the training set, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters.

[0046] According to one aspect of this disclosure, a thermal reservoir quality evaluation system is provided, comprising:

[0047] The acquisition unit is used to acquire reservoir quality evaluation parameters corresponding to known conditions and thermal reservoir quality evaluation parameters corresponding to the reservoir quality evaluation parameters.

[0048] The training unit is used to train the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters, so as to obtain the generator.

[0049] An evaluation unit is used to evaluate the quality of a thermal reservoir using the generator based on the quality evaluation parameters of the reservoir to be evaluated.

[0050] According to one aspect of this disclosure, a thermal reservoir quality evaluation system is provided, comprising:

[0051] A processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the above-described thermal reservoir quality evaluation method.

[0052] According to one aspect of this disclosure, a thermal reservoir quality evaluation system is provided, comprising: an electronic device, the electronic device being configured with a processor and a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the aforementioned thermal reservoir quality evaluation method.

[0053] According to one aspect of this disclosure, a thermal reservoir quality evaluation system is provided, comprising: a computer-readable storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement the above-described thermal reservoir quality evaluation method.

[0054] According to one aspect of this disclosure, a thermal reservoir quality evaluation system is provided, comprising: a computer program product including a computer program / instructions that, when executed by a processor, implement the above-described thermal reservoir quality evaluation method.

[0055] In the embodiments of this disclosure, a technical solution for a thermal reservoir quality evaluation method and system is proposed to address the technical problems of current discriminative methods, which suffer from insufficient utilization of the limited data features corresponding to reservoir quality evaluation parameters and inability to utilize reservoir quality information when the amount of data corresponding to the reservoir quality evaluation parameters is small and there is sample bias corresponding to the reservoir quality evaluation parameters. This solution aims to enhance the effective utilization of the limited input data corresponding to the reservoir quality evaluation parameters and improve the accuracy of thermal reservoir quality evaluation.

[0056] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0057] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0058] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the specification, serve to illustrate the technical solutions of this disclosure.

[0059] Figure 1 A flowchart illustrating a method for evaluating the quality of a thermal reservoir according to an embodiment of the present disclosure is shown.

[0060] Figure 2 This diagram illustrates a thermal reservoir quality evaluation zoning map according to an embodiment of the present disclosure.

[0061] Figure 3 The diagram illustrates the discriminant network corresponding to the training thermal reservoir quality evaluation model according to an embodiment of the present disclosure, and the network structure diagram corresponding to the discriminant network.

[0062] Figure 4 The diagram shows the model training results according to an embodiment of the present disclosure; wherein, (a) the changes in generator and discriminator losses; and (b) the changes in accuracy on the training and validation sets;

[0063] Figure 5 This illustrates a confusion matrix for evaluating the quality of thermal reservoirs according to an embodiment of the present disclosure;

[0064] Figure 6 This illustrates different models for identifying confusion matrices according to embodiments of the present disclosure;

[0065] Figure 7 The test accuracy of different generator networks according to embodiments of this disclosure is shown;

[0066] Figure 8 The test confusion matrices of 1DCNN-CGAN and MLP-CGAN according to embodiments of this disclosure are shown;

[0067] Figure 9 This is a block diagram illustrating an electronic device 800 according to an exemplary embodiment;

[0068] Figure 10 This is a block diagram illustrating an electronic device 1900 according to an exemplary embodiment. Detailed Implementation

[0069] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0070] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0071] In this document, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. 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.

[0072] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0073] It is understood that the various method embodiments mentioned above in this disclosure can be combined with each other to form combined embodiments without violating the principle and logic. Due to space limitations, this disclosure will not elaborate further.

[0074] In addition, this disclosure also provides a thermal reservoir quality evaluation system or device, electronic equipment, computer-readable storage medium, and program, all of which can be used to implement any of the thermal reservoir quality evaluation methods provided in this disclosure. The corresponding technical solutions and descriptions are described in the relevant section on thermal reservoir quality evaluation methods and will not be repeated here.

[0075] Figure 1 A flowchart illustrating a method for evaluating the quality of a thermal reservoir according to an embodiment of this disclosure is shown. Figure 1 As shown, the thermal reservoir quality evaluation method includes: Step S101: Obtaining reservoir quality evaluation parameters corresponding to known conditions and thermal reservoir quality evaluations corresponding to the reservoir quality evaluation parameters; Step S102: Training a generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discriminant network corresponding to the reservoir quality evaluation parameters to obtain a generator; Step S103: Performing thermal reservoir quality evaluation using the generator based on the reservoir quality evaluation parameters to be evaluated. This method addresses the technical problems of current discriminative methods, such as insufficient utilization of limited data features corresponding to reservoir quality evaluation parameters and inability to utilize reservoir quality information when the amount of data corresponding to the reservoir quality evaluation parameters is small and sample bias exists. It aims to enhance the effective utilization of limited input data corresponding to reservoir quality evaluation parameters and improve the accuracy of thermal reservoir quality evaluation.

[0076] Step S101: Obtain the reservoir quality evaluation parameters corresponding to the known conditions and the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters.

[0077] In the embodiments of this disclosure, the reservoir quality evaluation parameters corresponding to the known conditions include one or more of the following: clay content, fine sand content, silt content, porosity, permeability, and effective thickness of different reservoirs.

[0078] In the embodiments of this disclosure and other possible embodiments, the acquired thermal reservoir quality evaluation data are standardized to eliminate the dimensional influence between different reservoir parameters; wherein, the thermal reservoir quality evaluation data (reservoir quality evaluation parameters corresponding to known conditions) includes: clay content, fine sand content, silt content, porosity, permeability, and effective thickness of different reservoirs.

[0079] In embodiments of this disclosure, determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters includes: calculating the variance explained rate and loading coefficients corresponding to the principal components using principal component analysis; calculating a linear combination coefficient matrix based on the variance explained rate and the loading coefficients; calculating a comprehensive score coefficient matrix corresponding to the reservoir quality evaluation parameters based on the eigenvalues ​​of the linear combination coefficient matrix and the covariance matrix corresponding to the reservoir quality evaluation parameters; normalizing the comprehensive score coefficient matrix to obtain the weight values ​​of each reservoir quality parameter; and determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters based on the reservoir quality evaluation parameters and their corresponding weight values.

[0080] In the embodiments of this disclosure, before calculating the comprehensive score coefficient matrix corresponding to the reservoir quality evaluation parameters based on the eigenvalues ​​of the linear combination coefficient matrix and the covariance matrix corresponding to the reservoir quality evaluation parameters, a vector feature corresponding to the reservoir quality evaluation parameters is constructed, and the vector feature is standardized to obtain a standardized vector feature; the comprehensive score coefficient matrix corresponding to the reservoir quality evaluation parameters is calculated based on the eigenvalues ​​of the linear combination coefficient matrix and the covariance matrix corresponding to the standardized vector feature.

[0081] In embodiments of this disclosure, determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters based on the reservoir quality evaluation parameters and their corresponding weight values ​​includes: multiplying the reservoir quality evaluation parameters by their corresponding weight values ​​and summing the results to determine the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters. The weight values ​​of each reservoir quality parameter can be configured to be negative.

[0082] In the embodiments of this disclosure and other possible embodiments, different thermal reservoir quality grades (thermal reservoir quality evaluation) are classified according to thermal reservoir quality parameters such as clay content, fine sand content, silt content, porosity, permeability and effective thickness.

[0083] In the embodiments and other possible embodiments disclosed herein, different reservoir parameters (reservoir quality evaluation parameters or reservoir quality assessment parameters for different reservoirs) constitute a vector feature characterizing reservoir quality, and the vector feature characterizing reservoir quality composed of different reservoir parameters corresponds to different thermal reservoir quality grades; the classification criteria (thermal reservoir quality evaluation) for the three types of thermal reservoirs corresponding to different thermal reservoir quality grades include: Thermal reservoir quality evaluation type I: TRQ > 1.13; Thermal reservoir quality evaluation type II: 0.25 < TRQ < 1.13; Thermal reservoir quality evaluation type III: TRQ < 0.25. Wherein, TRQ represents the thermal reservoir quality evaluation.

[0084] In the embodiments and other possible embodiments disclosed herein, the thermal reservoir quality TRQ = 0.42 * effective thickness + 0.50 * porosity + 0.46 * permeability - 0.38 * clay content; wherein, thickness, porosity, permeability, and clay content are all standardized parameters.

[0085] In the embodiments and other possible embodiments disclosed herein, the thermal reservoir quality TRQ = 0.42 * thickness + 0.46 * porosity + 0.45 * permeability + 0.21 * fine sand - 0.16 silt - 0.38 * clay content; wherein, effective thickness, porosity, permeability, and clay content are all standardized parameters.

[0086] In the embodiments of this disclosure and other possible embodiments, vector features characterizing reservoir quality are composed of different reservoir clay content, fine sand content, silt content, porosity, permeability, and effective thickness. Based on data normalization and standardization, the variance explanation rate (cumulative contribution rate) and loading coefficients corresponding to the principal components are calculated using principal component analysis to select an appropriate number of principal components. Furthermore, based on the variance explanation rate (cumulative contribution rate) corresponding to the principal components and the coefficients of each index (clay content, fine sand content, silt content, porosity, permeability, and effective thickness) (reservoir quality parameters) in the linear combination of the principal components, the weights of each reservoir quality parameter are calculated. The comprehensive score for reservoir quality evaluation (the classification standard for different thermal reservoir quality grades) is obtained by summing the weights of the thermal reservoir quality parameters, thus realizing the classification of thermal reservoir quality (thermal reservoir quality evaluation).

[0087] In the embodiments of this disclosure and other possible embodiments, the acquired thermal reservoir quality evaluation data (reservoir quality evaluation parameters corresponding to known conditions) are standardized. The present invention selects maximum-minimum normalization processing, which can make the data distributed between [0,1], eliminate the influence of the dimensions between parameters, enhance the training effect of the model (generator or thermal reservoir quality evaluation model) and prevent gradient explosion.

[0088] The maximum-minimum normalization process is shown in the formula:

[0089]

[0090] In the embodiments of this disclosure and other possible embodiments, a vector feature X(X1,X2,...,X6) characterizing reservoir quality is composed of clay content X1, fine sand content X2, silt content X3, porosity X4, permeability X5, and effective thickness X6, resulting in a standardized vector feature X. std Principal component analysis was used to obtain the variance explained rate Λ and the corresponding loading matrix L of the principal components. The weight values ​​of each reservoir quality parameter were calculated using the following process:

[0091] pass Calculate the linear combination coefficient matrix A;

[0092] pass Calculate the comprehensive score coefficient matrix W; where λ i Configured as standardized vector feature X std The eigenvalues ​​corresponding to the covariance matrix;

[0093] Weight normalization involves summing and normalizing the comprehensive score coefficients of the comprehensive score coefficient matrix W to obtain the weight values ​​of each reservoir quality parameter.

[0094] Subsequently, the comprehensive score for reservoir quality evaluation was obtained by weighting and summing the quality parameters of the thermal reservoir. Combined with the porosity classification standard and the trend line between porosity and reservoir quality, the classification standard for the three types of water layers was determined.

[0095] Figure 2 A geothermal reservoir quality assessment and grading diagram according to an embodiment of this disclosure is shown. Figure 2 As shown, the thermal reservoir quality evaluation criteria are as follows: Class I: TRQ > 1.13; Class II: 0.25 < TRQ < 1.13; Class III: TRQ < 0.25.

[0096] Step S102: Using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters, train the generator network to obtain the generator (thermal reservoir quality evaluation model).

[0097] In the embodiments and other possible embodiments disclosed herein, the quality assessment of thermal reservoirs can determine the quality of identified thermal reservoirs, providing a basis for proposing suitable development schemes and utilization paths for thermal reservoirs of different qualities. In previous reservoir quality assessment methods, in addition to conventional assessment methods, many experts have also used machine learning and deep learning algorithms to achieve reservoir quality assessment. However, these previous algorithms mainly use discriminative networks, relying on a large data base, and are insufficient for extracting limited data features when the data volume is small or sample bias exists. Furthermore, discriminative methods always obtain a suitable decision boundary by learning the mapping relationship between input features and output categories, and cannot effectively utilize existing reservoir quality category information. Conditional Generative Adversarial Networks (CGANs) can simultaneously consider input features and class label information in the generation and discrimination processes, making the generated data more realistic. Given that CGANs simulate data generation through conditional probability, they are rarely directly used in classification tasks. Therefore, in order to better improve the model's learning ability on limited data and make full use of sample features in the case of small samples, this invention proposes a CGAN thermal reservoir quality evaluation method (thermal reservoir quality evaluation method) based on dense feature extraction, which can effectively utilize reservoir evaluation parameters and reservoir quality information.

[0098] In embodiments of this disclosure, training a generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and a reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters includes: generating a pseudo-evaluation of thermal reservoir quality based on the generator network using the reservoir quality evaluation parameters and random noise z (random noise space) under the reservoir quality evaluation parameter conditions; determining the truth or falsehood of the pseudo-evaluation of thermal reservoir quality based on the discrimination network using the reservoir quality evaluation parameters, the pseudo-evaluation of thermal reservoir quality, and the thermal reservoir quality evaluation; and calculating a first loss for the pseudo-evaluation of thermal reservoir quality corresponding to the random noise under the reservoir quality evaluation parameter conditions and a second loss for determining the truth or falsehood of the pseudo-evaluation of thermal reservoir quality, and training the generator network to obtain the generator.

[0099] In embodiments of this disclosure, calculating the first loss corresponding to the false evaluation of thermal reservoir quality under the conditions of the reservoir quality evaluation parameters and the second loss corresponding to the determination of whether the false evaluation of thermal reservoir quality is true or false includes: constructing a first loss function corresponding to the first loss. and the second loss function corresponding to the second loss. Based on the first loss function Calculate the first loss of the pseudo-evaluation of thermal reservoir quality corresponding to random noise under the conditions of the reservoir quality evaluation parameters; based on the second loss function Calculate the second loss corresponding to the random noise under the reservoir quality evaluation parameter conditions to determine whether the thermal reservoir quality false evaluation is true or false.

[0100] Where G represents the generator network, D represents the discriminator network, z represents random noise (random noise space), D(G(z|y)) is the judgment probability of the generated data (pseudo-evaluation of thermal reservoir quality); E represents the expectation; the reservoir quality evaluation parameters X_input corresponding to the known condition y (one or more of the following parameters of different reservoirs are normalized and / or standardized to obtain normalized and / or standardized reservoir quality evaluation parameters or their corresponding vector features), and x represents the real thermal reservoir quality evaluation corresponding to the pseudo-evaluation of thermal reservoir quality.

[0101] In embodiments of this disclosure, constructing the first loss function corresponding to the first loss includes: calculating the difference 1-D(G(z|y)) between a set value and a pseudo-evaluation of the thermal reservoir quality corresponding to random noise under the reservoir quality evaluation parameter conditions; taking the logarithm of the difference to obtain the logarithmic difference value log(1-D(G(z|y))); and calculating the first expected value of the logarithmic difference value log(1-D(G(z|y))). Construct a first loss function corresponding to the first loss; wherein the set value is configured to be 1.

[0102] In embodiments of this disclosure, constructing the first loss function corresponding to the first loss includes: taking the logarithm of the discrimination probability value D(x|y) corresponding to the random noise under the reservoir quality evaluation parameter conditions to determine whether the thermal reservoir quality false evaluation is true or false, to obtain the discrimination logarithm value logD(x|y); and calculating the second expected value of the logarithm value. Construct a second loss function corresponding to the second loss.

[0103] In embodiments of this disclosure, the step of calculating a first loss corresponding to a false evaluation of the thermal reservoir quality under the reservoir quality evaluation parameter conditions and a second loss corresponding to determining whether the false evaluation of the thermal reservoir quality is true or false, and training the generator network to obtain a generator, includes: constructing objective functions corresponding to the generator network and the discrimination network based on the first loss function corresponding to the first loss and the second loss function corresponding to the second loss. Maximize the first expected value of the first loss function corresponding to the first loss in the objective function and minimize the second expected value of the second loss function corresponding to the second loss in the objective function, i.e. The generator network is trained to obtain the generator.

[0104] In the embodiments of this disclosure and other possible embodiments, during the data generation process of a GAN, the generator's input is only a randomly generated noise vector z, resulting in a lack of effective guiding information in the data generation process. Since the only usable information is the discrimination result returned by the discriminator, this lack of effective information causes the GAN to fail to generate high-quality data. To address this limitation, the Conditional Generative Adversarial Network (CGAN) was proposed. CGAN adds some prior conditional information, such as class labels, text descriptions, and data from other modalities, to the original GAN, making the generation process of CGAN more controllable. Specifically, CGAN adds the same conditional constraint (known condition) y to the inputs of both the generator and the discriminator. For the generator, the conditional constraint (known condition) y and the noise vector z together serve as input, guiding the generation process in a specific direction; for the discriminator, the conditional constraint (known condition) y, together with the input data (whether real data or generated data), serves as the basis for judgment, ensuring that the discrimination process is also constrained by the same conditions.

[0105] In the embodiments of this disclosure and other possible embodiments, the generation and discrimination of data (pseudo-evaluations of thermal reservoir quality) are both based on the objective function of CGAN, given the reservoir quality evaluation parameter X_input corresponding to the known condition y, as shown in the formula:

[0106]

[0107] Where G represents the generator network, D represents the discriminator network, z represents random noise (random noise space), D(G(z|y)) is the judgment probability of the generated data (pseudo-evaluation of thermal reservoir quality); E represents the expectation; the reservoir quality evaluation parameters X_input corresponding to the known condition y (one or more of the following parameters of different reservoirs are normalized and / or standardized to obtain normalized and / or standardized reservoir quality evaluation parameters or their corresponding vector features), and x represents the real thermal reservoir quality evaluation corresponding to the pseudo-evaluation of thermal reservoir quality.

[0108] In embodiments of this disclosure, before training the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters, the process includes: normalizing and / or standardizing the reservoir quality evaluation parameters to obtain normalized and / or standardized reservoir quality evaluation parameters; and training the generator network using the normalized and / or standardized reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters.

[0109] In the embodiments of this disclosure, before training the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters, a training set is defined using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation corresponding to the reservoir quality evaluation parameters; the generator network is then trained based on the reservoir quality evaluation parameters corresponding to the training set, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters.

[0110] In the embodiments of this disclosure and other possible embodiments, the data (reservoir quality evaluation parameters) for realizing the classification of thermal reservoir quality (thermal reservoir quality evaluation) are stratified and sampled to divide different thermal reservoir quality types (thermal reservoir quality evaluation) in a ratio of 8:1:1, thereby obtaining the training set, validation set and test set used in the experiment.

[0111] In the embodiments of this disclosure and other possible embodiments, the present invention uses a vector feature X(X1,X2,...,X6) corresponding to the total amount of data, with a dimension of 668 (668×6). Through stratified sampling, different quality types (thermal reservoir quality evaluation) are divided in a set ratio of 8:1:1 to obtain the training set, validation set, and test set used in the experiment. The training set contains 32, 282, and 257 reservoir quality evaluations of types I, II, and III, respectively; the validation set and test set contain 7, 32, and 29 reservoir quality evaluations of types I, II, and III, respectively.

[0112] Figure 3 This diagram illustrates the discriminant network corresponding to the training thermal reservoir quality evaluation model according to an embodiment of this disclosure, and the network structure diagram corresponding to the discriminant network. Figure 3 As shown in the embodiments of this disclosure, the generating network includes: a first shallow feature extraction module, an encoder connected to the shallow feature extraction module, and a decoder connected to the shallow feature extraction module and the encoder respectively; wherein, the shallow feature extraction module is used to extract the reservoir quality evaluation parameters and a first shallow feature of random noise under the reservoir quality evaluation parameter conditions; the encoder is used to extract a first deep feature and a second deep feature corresponding to the first shallow feature; the decoder is used to generate a pseudo evaluation of the thermal reservoir quality based on the first shallow feature and its corresponding first deep feature and second deep feature.

[0113] like Figure 3As shown in the embodiments of this disclosure, the shallow feature extraction module includes: a first convolutional neural unit for extracting the reservoir quality evaluation parameters and the first shallow feature corresponding to the random noise under the reservoir quality evaluation parameter conditions; and / or, wherein the first convolutional neural unit includes at least one one-dimensional convolutional layer (1DCNN) or multiple cascaded one-dimensional convolutional layers (1DCNN).

[0114] like Figure 3 As shown in the embodiments of this disclosure, the encoder includes: a first dense feature extraction module and a second dense feature extraction module connected to the first dense feature extraction module; wherein, the first dense feature extraction module includes: a second convolutional neural unit, a third convolutional neural unit connected to the second convolutional neural unit, the first batch of normalization units connected to the third convolutional neural unit, and a first activation function connected to the first batch of normalization units; wherein, the input of the second convolutional neural unit is configured as the first shallow feature, and the output of the first activation function is configured as the first deep feature; the second dense feature extraction module includes: a fourth convolutional neural unit, a fifth convolutional neural unit connected to the fourth convolutional neural unit, and a first activation function connected to the fifth convolutional neural unit. The convolutional neural unit is connected to a second batch of normalization units and a second activation function connected to the second batch of normalization units; a first skip connection unit is configured between the input of the fourth convolutional neural unit and the input of the second convolutional neural unit and the output of the third convolutional neural unit; the first skip connection unit is used to concatenate the first shallow features and the processed features output by the third convolutional neural unit to obtain a first concatenated feature; the first concatenated feature is configured as the input of the fourth convolutional neural unit, and the output of the second activation function is configured as the second deep feature; wherein, the second convolutional neural unit, the third convolutional neural unit, the fourth convolutional neural unit, and the fifth convolutional neural unit include at least one one-dimensional convolutional layer or multiple cascaded one-dimensional convolutional layers.

[0115] like Figure 3 As shown in the embodiments of this disclosure, a second skip connection unit is configured between the input of the second convolutional neural unit and the input of the third convolutional neural unit; wherein, the second skip connection unit is used to concatenate the first shallow feature and the processed feature output by the second convolutional neural unit to obtain a second concatenated feature; the second concatenated feature is configured as the input of the third convolutional neural unit.

[0116] like Figure 3As shown, in an embodiment of this disclosure, a third skip connection unit is configured between the input of the fourth convolutional neural unit and the input of the fifth convolutional neural unit; wherein, the third skip connection unit is used to splice the spliced ​​feature with the processing feature output by the fourth convolutional neural unit to obtain a third spliced ​​feature; the third spliced ​​feature is configured as the input of the fifth convolutional neural unit.

[0117] like Figure 3 As shown in the embodiments of this disclosure, the decoder includes: a sixth convolutional neural unit, a feature self-attention module connected to the sixth convolutional neural unit, and a linear layer connected to the feature self-attention module; wherein, the sixth convolutional neural unit is used to perform convolution processing on the first shallow features to obtain second shallow features; the feature self-attention module is used to perform self-attention processing on the second shallow features, the first deep features, and the second deep features to obtain self-attention features; the linear layer is used to perform linear processing on the self-attention features to generate the pseudo-evaluation of the thermal reservoir quality; wherein, the sixth convolutional neural unit includes at least one one-dimensional convolutional layer or multiple cascaded one-dimensional convolutional layers.

[0118] like Figure 3 As shown in the embodiments of this disclosure, the feature self-attention module includes: a channel attention module and a spatial attention module connected to the channel attention module; wherein, the channel attention module is used to perform channel attention processing on the second shallow feature, the first deep feature and the second deep feature to obtain channel attention features; the spatial attention module is used to perform spatial attention processing on the spatial attention features to obtain self-attention features.

[0119] like Figure 3 As shown in the embodiments of this disclosure, before the channel attention module performs channel attention processing on the second shallow feature, the first deep feature, and the second deep feature, the second shallow feature, the first deep feature, and the second deep feature are spliced ​​together to obtain a spliced ​​feature to be processed; the channel attention module performs channel attention processing on the spliced ​​feature to be processed to obtain a channel attention feature.

[0120] like Figure 3As shown in the embodiments of this disclosure, the channel attention module includes: a first pooling unit and a second pooling unit different from the first pooling unit, a shared multilayer perceptron connected to the first pooling unit and the second pooling unit respectively, an adder connected to the shared multilayer perceptron, a third activation function connected to the adder, and a first multiplier; wherein, the first pooling unit and the second pooling unit perform pooling operations on the second shallow feature, the first deep feature, and the second deep feature corresponding to the features to be processed and spliced, respectively, to obtain the first pooled feature and the second pooled feature; the shared multilayer perceptron... The system comprises: a pooling unit for weighting the first pooling feature and the second pooling feature to obtain a first weighted feature and a second weighted feature; an adder for adding the first weighted feature and the second weighted feature to obtain an additive feature; a third activation function for nonlinearly processing the additive feature to obtain a channel attention weight; and a first multiplier for multiplying the channel attention weight and the concatenated feature to be processed to obtain a channel attention feature. The first pooling unit is configured as a max pooling layer, and the second pooling unit is configured as a mean pooling layer.

[0121] like Figure 3 As shown in the embodiments of this disclosure, the spatial attention module includes: a third pooling unit, a fourth pooling unit connected to the third pooling unit, a convolutional layer connected to the fourth pooling unit, a fourth activation function connected to the convolutional layer, and a second multiplier; wherein, the third pooling unit and the fourth pooling unit are used sequentially to perform pooling processing on the channel attention features to obtain third pooling features; the convolutional layer is used to perform convolution processing on the third pooling features to obtain convolutional features; the fourth activation function is used to perform nonlinear processing on the convolutional features to obtain spatial attention weights; the second multiplier is used to perform multiplication operations on the channel attention features and the spatial attention weights to obtain self-attention features; wherein, the third pooling unit is configured as a max pooling layer, and the fourth pooling unit is configured as a mean pooling layer; wherein, the convolutional kernel size corresponding to the convolutional layer is configured as 7*7.

[0122] like Figure 3 As shown in the embodiments of this disclosure, the linear layer connected to the feature self-attention module FSAtt includes: a first linear layer and a second linear layer connected to the first linear layer; wherein, the second linear layer is used to perform linear processing on the linear features after linear processing of the self-attention features by the first linear layer, and then perform linear processing again to generate the pseudo-evaluation of the thermal reservoir quality.

[0123] like Figure 3As shown in the embodiments of this disclosure, the discriminant network includes: at least one convolutional correlation weight module or multiple cascaded convolutional correlation weight modules; wherein, the convolutional correlation weight module is used to train the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation corresponding to the reservoir quality evaluation parameters; wherein, the number of the multiple cascaded convolutional correlation weight modules can be configured to be 3.

[0124] like Figure 3 As shown in the embodiments of this disclosure, each of the convolutional correlation weight modules includes: a seventh convolutional neural unit, a feature self-attention unit connected to the seventh convolutional neural unit, a third batch normalization unit connected to the feature self-attention unit, and a fifth activation function connected to the third batch normalization unit; wherein, the seventh convolutional neural unit is used to perform convolution processing on the reservoir quality evaluation parameters, the reservoir quality evaluation and reservoir quality pseudo-evaluation corresponding to the reservoir quality evaluation parameters, or the correlation weight features output by the previous level convolutional correlation weight module; to obtain the convolutional features to be discriminated; the feature self-attention unit... The unit is used to perform self-attention processing on the convolutional features to be discriminated, to obtain self-attention features to be discriminated; the third batch normalization unit is used to perform batch normalization processing on the self-attention features to be discriminated, to obtain batch normalized features to be discriminated; the fifth activation function is used to perform nonlinear processing on the batch normalized features to be discriminated, to obtain the discrimination result corresponding to each convolutional correlation weight module; wherein, the seventh convolutional neural unit includes at least one one-dimensional convolutional layer or multiple cascaded one-dimensional convolutional layers; wherein, the discrimination result is configured as the true or false corresponding to the thermal reservoir quality pseudo evaluation and / or the thermal reservoir quality evaluation.

[0125] like Figure 3 As shown in the embodiments of this disclosure, a skip connection unit is provided between the inputs of each of the convolutional correlation weight modules; wherein, the skip connection unit is used to concatenate the discrimination result corresponding to the output of the previous level convolutional correlation weight module with the input features of the previous level convolutional correlation weight module to obtain the correlation weight concatenation features of the next level convolutional correlation weight module.

[0126] like Figure 3As shown, in the embodiments and other possible embodiments of this disclosure, a CGAN (DE-CGAN) thermal reservoir quality evaluation method based on dense feature extraction is introduced. First, to better achieve feature extraction of the input information (the reservoir quality evaluation parameters and random noise under the conditions of the reservoir quality evaluation parameters) and enhance the model's learning ability, the generator network model utilizes the encoder and decoder concepts. Specifically, a 1DCNN (One-Dimensional Convolutional Neural Network) is first used to obtain shallow features of the input data. These shallow features are then input into a dense feature extraction module to enhance the feature learning ability on limited data, outputting deeper features and thus encoding the input features. Then, the shallow and deep features are fused through a hybrid attention module. The fused features are then used to predict the generation of the thermal reservoir through two linear network layers, thus decoding the input features. This completes the generator structure design. Second, using the residual concept, a discriminator network model is designed. The primary approach involves first employing a 3-layer 1D CNN and a feature self-attention mechanism to learn features from the input data. Then, the original data and the extracted features are concatenated. Next, two linear layers are used to distinguish whether the input data originates from the generator or is real data. Finally, through continuous adversarial interaction between the generator and discriminator network models, the generator's ability to generate data is improved, enabling the generator model to be used for quality assessment of unknown thermal reservoirs. The discriminator network (discriminator network) aims to classify real data (real thermal reservoir quality assessments) as "1" (true) and generated data (pseudo-assessments of thermal reservoir quality relative to the real assessments) as "0" (false).

[0127] In the embodiments of this disclosure and other possible embodiments, during training, the generator and discriminator need to be optimized simultaneously, promoting and improving each other through adversarial game theory. Specifically, the generator aims to continuously optimize network parameters and improve the quality of generated data so that the discriminator cannot distinguish between true and false data; that is, to maximize the probability that the discriminator classifies the generated data as real. The discriminator, on the other hand, needs to continuously minimize the classification error rate of the input data, i.e., maximize the difference between its probability of classifying real data and its probability of classifying generated data.

[0128] In the embodiments of this disclosure and other possible embodiments, by jointly optimizing the objective functions of the generator and the discriminator, the training objective is to find a Nash equilibrium point between the generator and the discriminator, so that the data generated by the generator is realistic enough that the discriminator cannot distinguish between real data and generated data.

[0129] In the embodiments of this disclosure and other possible embodiments, the generator and discriminator are updated alternately during training. First, the discriminator is trained using real data and generated data, and its parameters are updated to enable it to better distinguish between real and generated data. Second, the discriminator parameters are fixed, while the generator parameters are updated to make the generator produce more "realistic" data and increase the probability that the discriminator classifies the generated data as "real". Finally, through continuous alternating updates and optimizations, the discriminator's error rate is increased, the generator can estimate the distribution of sample data, and the generated samples are more realistic.

[0130] In the embodiments of this disclosure and other possible embodiments, CGAN can effectively utilize given conditional constraints to generate the data people need. As can be seen from its objective function, mathematically, CGAN transforms ordinary probability into conditional probability. In the implementation of CGAN, the condition (reservoir quality evaluation parameter X_input) in the conditional probability represents the input, and the output is the thermal reservoir quality evaluation.

[0131] In the embodiments of this disclosure and other possible embodiments, to better apply CGAN directly to the process of evaluating the quality of thermal reservoirs, that is, to enable the model to provide a quality evaluation of the thermal reservoir under given thermal reservoir quality parameters, the data generation process is equivalent to the discrimination process of a conventional discriminant network.

[0132] In the embodiments of this disclosure, the generator network (generator network) utilizes the encoder and decoder concept, dividing the generator into two parts: an encoder and a decoder, to fully utilize the input data and complete the dense feature extraction. The dense feature extraction generator model takes the latent noise space Z and the reservoir quality evaluation parameters as constraints as input. The input data (the reservoir quality evaluation parameters and random noise under the conditions of the reservoir quality evaluation parameters) is first processed by a 1DCNN network (the first convolutional neural unit) to obtain the unencoded first shallow feature 1. Then, the first shallow feature 1 is input into the designed dense feature extraction module to obtain the encoded first deep feature 1 and second deep feature 2, thus encoding the input features. Afterwards, the unencoded first shallow feature 1 is processed again by a 1DCNN feature extraction layer and fused with the first deep feature 1 and second deep feature 2 through the attention module CBAM. The fused features are then passed through two cascaded linear network layers (the first linear layer and the second linear layer) to predict the generation of the thermal reservoir, thus decoding the input features and obtaining the pseudo-evaluation of the thermal reservoir quality.

[0133] like Figure 3As shown in the embodiments of this disclosure and other possible embodiments, the entire encoder mainly consists of two cascaded dense feature modules. Each dense feature module (DE) includes: two layers of 1DCNN network (the second convolutional neural unit and the third convolutional neural unit, or the fourth convolutional neural unit and the fifth convolutional neural unit), a batch normalization layer (BN) (the first batch normalization unit or the second batch normalization unit), and an activation function layer (the first activation function or the second activation function) ReLU. To effectively achieve dense extraction between data and enhance the feature extraction effect, the module establishes a residual link (the second skip connection unit or the third skip connection unit) between the two layers of 1DCNN, concatenating the input features and the output features of the first layer of 1DCNN (the second convolutional neural unit or the fourth convolutional neural unit) in the channel dimension, which together serve as the input of the dense module to the second layer of 1DCNN network (the third convolutional neural unit or the fifth convolutional neural unit). After processing by the 1DCNN network (either the third or fifth convolutional neural unit), the features sequentially pass through a Batch Normalization (BN) layer (either the first or second batch of normalization units) and a ReLU layer (either the first or second activation function). The BN layer (either the first or second batch of normalization units) standardizes the feature distribution, maintaining zero mean and unit variance, thus increasing the model's stability. Finally, the output of a single dense feature module is shown in the formula:

[0134] The first deep feature 1out1 = ReLU(BN(Conv2(Concat(x,Conv1(x)))).

[0135] Wherein, Conv1 represents the first 1DCNN convolution (1DCNN network, i.e., the second convolutional neural unit or the fourth convolutional neural unit), and Conv2 represents the second 1DCNN convolution (1DCNN network, i.e., the third convolutional neural unit or the fifth convolutional neural unit).

[0136] like Figure 3As shown, in the embodiments and other possible embodiments of this disclosure, in order to further achieve dense feature extraction, data is transferred in the same way in both dense modules. The output features of the previous dense feature module (DE) are concatenated with the input data and used as the input of the next dense feature module (DE) (a second skip connection unit is configured between the input of the second convolutional neural unit and the input of the third convolutional neural unit; wherein, the second skip connection unit is used to concatenate the first shallow features with the processed features output by the second convolutional neural unit to obtain a second concatenated feature; the second concatenated feature is configured as the input of the third convolutional neural unit). This cascaded feature transfer allows subsequent modules to directly utilize the feature information extracted by the previous modules, significantly improving the feature utilization rate, and also enabling the fusion of multi-level features, allowing the network to simultaneously utilize shallow local features and deep global features. Finally, the outputs of the two dense feature modules (DE) are concatenated together to provide a portion of the input for the subsequent decoder (a first skip connection unit is configured between the input of the fourth convolutional neural unit, the input of the second convolutional neural unit, and the output of the third convolutional neural unit; the first skip connection unit is used to concatenate the first shallow feature and the processed feature output by the third convolutional neural unit to obtain a first concatenated feature; the first concatenated feature is configured as the input of the fourth convolutional neural unit).

[0137] like Figure 3 As shown in the embodiments and other possible embodiments of this disclosure, for the decoder, the first shallow feature 1 extracted by the 1DCNN network (the first convolutional neural unit) is input into the decoder in addition to being input into the encoder. The input first shallow feature 1 is then processed again by a 1DCNN network (the sixth convolutional neural unit) to extract features, resulting in the second shallow feature 2. Subsequently, the extracted second shallow feature 2, along with the first deep feature 1 and the second deep feature 2 from the encoder, are fused together by the convolutional attention mechanism module CBAM to improve the model's ability to focus on features at different levels, achieving effective feature fusion. Finally, the fused features are sequentially input into two linear network layers (the first linear layer and the second linear layer) to generate and predict the thermal reservoir quality data, obtaining the pseudo-evaluation of the thermal reservoir quality.

[0138] In the embodiments of this disclosure and other possible embodiments, the Convolutional Attention Mechanism (CBAM) module performs attention mechanism operations in the channel and spatial dimensions, making the network pay more attention to important channel features and spatial location features, suppressing unwanted channel and spatial location features, and increasing the network's ability to capture input features.

[0139] In embodiments of this disclosure and other possible embodiments, Channel Attention CAM is a method that compresses spatial dimensions through pooling layers, focusing on the importance of each feature map channel, aiming to enhance the feature representation ability between different channels. First, Channel Attention Weight CAM calculates the maximum feature value on each channel through max pooling (MaxPool) (the first pooling unit) and average pooling (AvgPool) (the second pooling unit) operations, respectively. (First pooling feature) and average feature (Second pooling feature). Max pooling extracts the most significant feature values ​​from each channel, highlighting key information in the feature map; while average pooling reflects the overall distribution of features in each channel, providing global statistical information. Next, the obtained feature vector is processed through a shared-parameter multilayer perceptron (MLP) to learn the weight parameters (first and second weight features) for each channel. Finally, the channel attention weight Mc(F) is obtained by normalization using a third activation function σ (such as Sigmoid), and multiplied by the original feature map F (the concatenated feature obtained by concatenating the second shallow feature, the first deep feature, and the second deep feature) to adjust the weights for different channels. The calculation process of the channel attention weight CAM is shown in the formula:

[0140]

[0141] Where W is configured as the weight matrix of a multilayer perceptron (MLP);

[0142] In the embodiments of this disclosure and other possible embodiments, Spatial Attention (SAM) is an important method for enhancing the network's representation ability in the spatial dimension. It primarily identifies important features at each spatial location, overcoming the spatial limitations of channel attention. First, the input feature map (the channel attention feature) is compressed using average pooling (the fourth pooling unit) and max pooling (the third pooling unit) to obtain two 1*H*W feature maps (the third pooling feature and the fourth pooling feature), extracting global information in the spatial dimension. Then, the two feature maps (the third pooling feature and the fourth pooling feature) obtained from average pooling and max pooling are concatenated using a concat operation (the third pooling feature), or the third pooling unit and the fourth pooling unit are used sequentially to pool the channel attention feature to obtain the third pooling feature. Channel information is then fused using the 7*7 convolutional kernel corresponding to convolutional layer f to obtain convolutional features. The 7×7 convolutional kernel corresponding to convolutional layer f has a large receptive field, enabling it to better capture long-distance dependencies in the feature map in space. Finally, the convolutional features are normalized using an activation function σ (such as Sigmoid) to obtain the spatial attention weights M. s (F). Spatial attention weights M s The calculation process for (F) is shown in the formula:

[0143]

[0144] In the embodiments of this disclosure and other possible embodiments, in the CBAM module, the input (input feature F, i.e., the concatenated feature obtained by concatenating the two high-level features (first deep feature 1 and second deep feature 2) from the encoder and the feature extracted from the decoder (second shallow feature 2)) is used as the channel concatenation result. First, channel attention is used to calculate the weights of different channels, resulting in a channel attention vector Mc (channel attention weight), which is then multiplied by the input feature to complete the CAM output F` (channel attention feature). Then, using this as input, the weights of different spatial locations are calculated to obtain the SAM vector Ms (spatial attention weight), which is then multiplied by the input feature (the concatenated feature to be processed) to complete the final output result F`` (self-attention feature). Using CBAM can effectively fuse the second shallow feature 2, the first deep feature 1, and the second deep feature 2, complete the extraction of the importance of features at different levels, and extract information from different spatial locations, enhance the overall representation ability of the generator network, and improve the model to generate high-quality thermal reservoir data (pseudo-evaluation of thermal reservoir quality) given thermal reservoir quality parameters, thereby achieving accurate evaluation of thermal reservoir quality.

[0145] like Figure 3 As shown, in embodiments of this disclosure and other possible embodiments, the discriminator network (discriminator network) includes: a 1DCNN network (configured within the convolutional correlation weight module) and two linear connection layers, i.e., at least one linear layer is provided after at least one convolutional correlation weight module or multiple cascaded convolutional correlation weight modules. For example, two linear layers are provided after at least one convolutional correlation weight module or multiple cascaded convolutional correlation weight modules, respectively configured as a third linear layer and a fourth linear layer connected to the third linear layer. Within the convolutional correlation weight module, after each 1DCNN network layer (the seventh convolutional neuron), a feature self-attention module (FSAtt) (the feature self-attention unit), a batch normalization (BN) layer (the third batch normalization unit), and a ReLU activation function (the fifth activation function) are sequentially added. The feature self-attention module FSAt enhances the expressive power of important features by calculating the correlation weights between features; the BN layer is used to stabilize the feature distribution and accelerate model convergence; and the ReLU activation function introduces nonlinear transformation capability. To optimize feature transfer efficiency, layers are connected through residuals, using the input and output of the current layer as the input of the next layer, thereby improving the ability to represent input features and enhancing feature reuse.

[0146] In the embodiments of this disclosure and other possible embodiments, the discriminator network adopts a dual-input design. The first input is configured as a first composite feature obtained by concatenating the real thermal reservoir quality evaluation with the parameter X_input (the reservoir quality evaluation parameter corresponding to the known conditions and the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameter). The other input is a second composite feature obtained by concatenating the thermal reservoir quality evaluation (pseudo-evaluation of thermal reservoir quality) generated by the generator with the parameter X_input (the reservoir quality evaluation parameter corresponding to the known conditions and the pseudo-evaluation of thermal reservoir quality corresponding to the reservoir quality evaluation parameter). During training, after the first composite feature passes through the discriminator network, the probability of it being judged as a real sample (the discriminant probability value) is calculated using the binary cross-entropy loss function; simultaneously, the second composite feature is also processed by the same discriminator network, and the probability of it being identified as a generated sample (the discriminant probability value) is calculated using the cross-entropy loss function. This adversarial training mechanism forces the discriminator network to continuously improve its discrimination ability, while also driving the generator network to improve its generation quality.

[0147] Step S103: Based on the reservoir quality evaluation parameters to be evaluated, perform thermal reservoir quality evaluation using the generator.

[0148] In the embodiments of this disclosure and other possible embodiments, the reservoir quality evaluation parameters to be evaluated are consistent with the reservoir quality evaluation parameters corresponding to the training of the generator, including one or more of the following: clay content, fine sand content, silt content, porosity, permeability, and effective thickness of different reservoirs.

[0149] The entity executing the thermal reservoir quality assessment method can be a thermal reservoir quality assessment system or device. For example, the method can be executed by a terminal device, server, or other processing device. The terminal device can be user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, vehicle-mounted device, wearable device, etc. In some possible implementations, the thermal reservoir quality assessment method can be implemented by a processor calling computer-readable instructions stored in memory.

[0150] Those skilled in the art will understand that in the above-described method for evaluating the quality of thermal reservoirs in specific embodiments, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0151] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the embodiments of the thermal reservoir quality evaluation method above. The specific implementation can be referred to the description of the method embodiments above, and for the sake of brevity, it will not be repeated here.

[0152] This disclosure also proposes a thermal reservoir quality evaluation system, comprising: an acquisition unit for acquiring reservoir quality evaluation parameters corresponding to known conditions and thermal reservoir quality evaluations corresponding to the reservoir quality evaluation parameters; a training unit for training a generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters to obtain a generator; and an evaluation unit for performing thermal reservoir quality evaluation using the generator based on the reservoir quality evaluation parameters to be evaluated.

[0153] This disclosure also proposes a thermal reservoir quality evaluation system, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the aforementioned thermal reservoir quality evaluation method.

[0154] This disclosure also proposes a thermal reservoir quality evaluation system, including: an electronic device, the electronic device being configured with a processor and a memory for storing processor-executable instructions; wherein, the processor is configured to invoke the instructions stored in the memory to execute the above-described thermal reservoir quality evaluation method.

[0155] This disclosure also proposes a thermal reservoir quality evaluation system, comprising: a computer-readable storage medium storing computer program instructions thereon, wherein the computer program instructions, when executed by a processor, implement the above-described thermal reservoir quality evaluation method.

[0156] This disclosure also proposes a thermal reservoir quality evaluation system, including: a computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the above-described thermal reservoir quality evaluation method.

[0157] This disclosure also proposes a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the aforementioned thermal reservoir quality evaluation method. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0158] This disclosure also proposes an electronic device, including: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured for the aforementioned thermal reservoir quality evaluation method. The electronic device may be provided as a terminal, a server, or other form of device.

[0159] To effectively evaluate the quality of thermal reservoirs, this invention introduces fine sand content and silt content in addition to the parameters used for classifying reservoir quality. In thermal reservoir quality evaluation, fine sand content and silt content are important geophysical parameters that significantly affect the reservoir's physical properties, permeability, thermal storage capacity, and thermal conductivity. Fine sand, with its smaller particle size, typically provides a certain degree of porosity and permeability channels. However, due to its small particle size, its effect on reservoir permeability is complex. An appropriate amount of fine sand can maintain good permeability while preserving relatively ideal thermal conductivity. Silt, with its even smaller particle size, usually leads to lower reservoir porosity and poor permeability, thereby inhibiting fluid flow and heat conduction, reducing the reservoir's effective thermal storage capacity. Silt layers in thermal reservoirs often exhibit lower thermal conductivity, resulting in decreased thermal energy storage and extraction efficiency. Therefore, the content of fine sand and silt is also a crucial factor that cannot be ignored when evaluating thermal reservoirs.

[0160] After identifying the water layer and calculating the effective thickness, the total amount of data obtained in this invention is 668. Stratified sampling is still used in data partitioning to ensure that the data has a similar distribution across the training and validation sets. The data distribution in different datasets is shown in Table 1.

[0161] Table 1. Data distribution of training set, validation set, and test set

[0162]

[0163] Figure 4 The diagram illustrates the model training results according to an embodiment of this disclosure; wherein, (a) the changes in generator and discriminator losses; and (b) the changes in accuracy on the training and validation sets. Figure 4 As shown, a constructed conditional generative adversarial network based on dense feature extraction is used to train and validate the training and validation sets. During training, the model parameters with the highest recognition accuracy in the validation set are saved to improve the model's recognition accuracy. Figure (a) shows that as the number of epochs increases, the generator's loss function value continuously decreases, while the discriminator's loss function value continuously increases. This indicates that the generator and discriminator are constantly engaged in an adversarial game, and that the probability of the discriminator classifying the data generated by the generator as true is continuously increasing. Figure (b) shows that as the generator is continuously optimized, the quality of the generated data improves, resulting in better evaluation of the thermal reservoir quality.

[0164] Secondly Figure 5 The confusion matrix for evaluating the quality of a thermal reservoir according to an embodiment of this disclosure is shown. Using a saved optimal model, the model is tested on a test set, and the resulting confusion matrix is ​​shown below. Figure 5 As shown, the model demonstrates good accuracy in identifying different geothermal reservoir quality types. It correctly identifies all Class I and Class III geothermal reservoirs, but misidentifies one Class II reservoir as a Class I reservoir. The overall identification rate reaches 98.51%, exhibiting good accuracy. Furthermore, the model shows good balance, effectively learning the quality evaluation parameters of different geothermal reservoirs even in the presence of uneven distribution.

[0165] To verify the effectiveness of the method of this invention in the evaluation of thermal reservoir quality, it was compared with commonly used existing reservoir evaluation methods. Table 2 shows the overall accuracy of different models in evaluating thermal reservoir quality and their performance in identifying thermal reservoirs of different qualities. From the data in the table, DE-CGAN has the highest overall accuracy at 98.51%, indicating that it performs best in the thermal reservoir quality classification task. 1DCNN and XGBoost perform second best in thermal reservoir quality evaluation, with an overall recognition rate of 94.03%. The K-Means clustering algorithm has the worst overall recognition performance at 88.06%, possibly due to its unsupervised learning nature, making it difficult to capture complex class boundaries. Random Forest (RF) and GDBT show improvements over K-Means, at 92.54% and 91.04%, respectively.

[0166] DE-CGAN performed best in identifying Type I thermal reservoirs, achieving 100% precision and a high recall (87.5%), with an F1 score of 93.3%. 1DCNN and RF performed second best in identifying Type I thermal reservoirs, both achieving 71.43% precision, but 1DCNN had better recall and F1 score than RF. XGBoost and GDBT were next, while K-Means performed the worst, with low precision, recall, and F1 score of 28.57%, 40%, and 33.33%, respectively. DE-CGAN achieved 100% recall and the highest F1 score (98.4%) for identifying Type II thermal reservoirs. XGBoost achieved 100% precision for Type II thermal reservoirs, but its recall was slightly lower (88.57%). RF, GDBT, and 1DCNN also performed well in identifying Type II thermal reservoirs, while K-Means performed relatively poorly. DE-CGAN achieved 100% precision, recall, and F1 score in Class III, demonstrating the best performance. Other algorithms also performed well in Class III, but DE-CGAN's performance was more outstanding.

[0167] Table 2 Model Evaluation Score Table

[0168]

[0169] Figure 6 The diagram illustrates confusion matrices for different models according to embodiments of this disclosure. The confusion matrices for thermal reservoir quality evaluation of different algorithm models are shown below. Figure 6 As shown in the figure, the K-Means algorithm not only identifies Class II thermal reservoirs as Class I reservoirs, but also identifies Class III reservoirs as Class I reservoirs, resulting in the worst quality Class I reservoir identification. RF and 1DCNN are models with good identification performance for Class I thermal reservoirs, except for DE-CGAN. However, 1DCNN does not identify other thermal reservoirs as Class I reservoirs, and the identified Class I reservoirs have better purity. GDBT and XGBoost have improved the identification ability of Class I thermal reservoirs compared to K-Means, and the purity of the identified Class I thermal reservoirs is good, but still lower than RF, 1DCNN, and DE-CGAN. The DE-CGAN proposed in this invention has the best effect on identifying Class I reservoirs compared to other models, but relatively speaking, there is a risk of misidentifying Class II reservoirs as Class I reservoirs.

[0170] In terms of overall recognition performance, DE-CGAN demonstrates excellent ability to evaluate the quality of thermal reservoirs. Moreover, DE-CGAN effectively addresses the issue of data imbalance between thermal reservoirs of different qualities compared to other algorithms, exhibiting better data feature utilization and data balancing capabilities, and resulting in more stable model performance.

[0171] Figure 7 The test accuracy of different generative networks according to embodiments of this disclosure is shown. To verify the effectiveness of the dense feature extraction module, the present invention directly uses 1DCNN and MLP layers to construct conditional generative adversarial networks, respectively, retaining the optimal model parameters obtained on the validation set, and then using the optimal model for testing. During training, the overall accuracy of the validation set changes with epochs, as shown in the figure. Figure 7 As shown, DE-CGAN has higher recognition accuracy and stronger stability compared to 1DCNN-CGAN and MLP-CGAN.

[0172] After testing the models, the recognition performance of 1DCNN-CGAN and MLP-CGAN on the test set was obtained. The overall accuracy of the models is shown in Table 3, which are 95.52% and 92.54%, respectively.

[0173] Figure 8 The test confusion matrices for 1DCNN-CGAN and MLP-CGAN according to embodiments of this disclosure are shown. The confusion matrix results are as follows: Figure 8 As shown, compared with DE-CGAN, both models also perform slightly worse in identifying Class I reservoirs, which in turn affects the model's recognition accuracy.

[0174] Table 3 Comparison of Validation Results for Dense Feature Modules

[0175]

[0176] The above experiments demonstrate that the DE-CGAN proposed in this invention has a good effect on the quality evaluation of thermal reservoirs. Moreover, it can make full use of the input data information when the data sample is small, reduce the impact of data imbalance, and identify Class I thermal reservoirs with good quality, which can provide a basis for subsequent delineation of favorable areas using sedimentary facies and temperature.

[0177] Thermal reservoir quality assessment can determine the quality of identified thermal reservoirs. This invention addresses the problems of discriminative methods, such as insufficient utilization of limited data features and inability to utilize reservoir quality information when data volume is limited and sample bias exists. It proposes a CGAN-based method for thermal reservoir quality assessment using dense feature extraction, which can effectively utilize reservoir assessment parameters and reservoir quality information. First, a generator is constructed using an encoder and encoder concept. A 1DCNN network is used to obtain shallow features of the input data. These shallow features are then input into a designed dense feature extraction module to obtain deeper features, achieving dense extraction of input features and encoding the information. Next, the shallow and deep features are fused through a hybrid attention module. The fused features are then passed through two linear network layers to predict the generation of thermal reservoirs, thus decoding the input features. A discriminator network model is designed using the residual concept. It first uses a 3-layer 1DCNN and a feature self-attention mechanism to learn features from the input data. Then, the original data and the extracted features are concatenated, and finally, two linear layers are used to distinguish whether the input data comes from the generator or real data. Secondly, the constructed model was trained, validated, and tested, and compared with other commonly used methods for evaluating the quality of thermal reservoirs. The results show that the model proposed in this invention has better overall recognition performance and the best recognition accuracy in Class I thermal reservoir identification. It can reduce the impact of data imbalance when the amount of data is small, and achieve effective utilization of small sample data. Furthermore, the effectiveness of the dense feature extraction module was verified, which can make full use of data information, enhance the model's learning ability, and improve the model's ability to evaluate the quality of thermal reservoirs.

[0178] Figure 9 This is a block diagram illustrating an electronic device 800 according to an exemplary embodiment. For example, the electronic device 800 may be a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, or other terminal.

[0179] Reference Figure 9 The electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input / output (I / O) interface 812, a sensor component 814, and a communication component 816.

[0180] Processing component 802 typically controls the overall operation of electronic device 800, such as operations associated with display, telephone calls, data communication, camera operation, and recording operations. Processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the methods described above. Furthermore, processing component 802 may include one or more modules to facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.

[0181] Memory 804 is configured to store various types of data to support the operation of electronic device 800. Examples of this data include instructions for any application or method operating on electronic device 800, contact data, phonebook data, messages, pictures, videos, etc. Memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0182] Power supply component 806 provides power to various components of electronic device 800. Power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 800.

[0183] Multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 808 includes a front-facing camera and / or a rear-facing camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0184] Audio component 810 is configured to output and / or input audio signals. For example, audio component 810 includes a microphone (MIC) configured to receive external audio signals when electronic device 800 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 804 or transmitted via communication component 816. In some embodiments, audio component 810 also includes a speaker for outputting audio signals.

[0185] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0186] Sensor assembly 814 includes one or more sensors for providing state assessments of various aspects of electronic device 800. For example, sensor assembly 814 can detect the on / off state of electronic device 800, the relative positioning of components such as the display and keypad of electronic device 800, changes in position of electronic device 800 or a component of electronic device 800, the presence or absence of user contact with electronic device 800, orientation or acceleration / deceleration of electronic device 800, and temperature changes of electronic device 800. Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 814 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0187] Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices. Electronic device 800 can access wireless networks based on communication standards, such as WiFi, 2G, or 3G, or combinations thereof. In one exemplary embodiment, communication component 816 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 816 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0188] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0189] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 804 including computer program instructions that can be executed by a processor 820 of an electronic device 800 to perform the above-described method.

[0190] Figure 10 This is a block diagram illustrating an electronic device 1900 according to an exemplary embodiment. For example, the electronic device 1900 may be provided as a server. (Refer to...) Figure 10 The electronic device 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by memory 1932 for storing instructions, such as application programs, that can be executed by the processing component 1922. The application programs stored in memory 1932 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 1922 is configured to execute instructions to perform the methods described above.

[0191] Electronic device 1900 may also include a power supply component 1926 configured to perform power management of electronic device 1900, a wired or wireless network interface 1950 configured to connect electronic device 1900 to a network, and an input / output (I / O) interface 1958. Electronic device 1900 can operate on an operating system stored in memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0192] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions that can be executed by a processing component 1922 of an electronic device 1900 to perform the above-described method.

[0193] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0194] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0195] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0196] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0197] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should 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-readable program instructions.

[0198] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0199] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0200] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0201] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method of evaluating the quality of a thermal reservoir, characterized in that, include: Obtain the reservoir quality evaluation parameters corresponding to the known conditions and the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters. A generator is trained using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and a reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters. The generator includes: generating a pseudo-evaluation of the thermal reservoir quality based on the generator network, using the reservoir quality evaluation parameters and random noise under the reservoir quality evaluation parameter conditions; determining the authenticity of the pseudo-evaluation of the thermal reservoir quality based on the discrimination network, using the reservoir quality evaluation parameters, the pseudo-evaluation of the thermal reservoir quality, and the actual thermal reservoir quality; calculating a first loss for the pseudo-evaluation of the thermal reservoir quality corresponding to the random noise under the reservoir quality evaluation parameter conditions, and determining the thermal reservoir quality... The second loss corresponding to the true or false nature of the pseudo-evaluation is used to train the generator network to obtain a generator; the generator network includes: a first shallow feature extraction module, an encoder connected to the shallow feature extraction module, and a decoder connected to the shallow feature extraction module and the encoder respectively; wherein, the shallow feature extraction module is used to extract the reservoir quality evaluation parameters and the first shallow features of random noise under the conditions of the reservoir quality evaluation parameters; the encoder is used to extract the first deep features and the second deep features corresponding to the first shallow features; the decoder is used to generate a pseudo-evaluation of the thermal reservoir quality based on the first shallow features and their corresponding first deep features and the second deep features. The generator is used to evaluate the quality of the thermal reservoir based on the reservoir quality evaluation parameters to be evaluated.

2. The method for evaluating the quality of a thermal reservoir according to claim 1, characterized by, The calculation of the first loss for the pseudo-evaluation of thermal reservoir quality corresponding to random noise under the reservoir quality evaluation parameters and the second loss for determining whether the pseudo-evaluation of thermal reservoir quality is true or false includes: constructing a first loss function corresponding to the first loss and a second loss function corresponding to the second loss; calculating the first loss for the pseudo-evaluation of thermal reservoir quality corresponding to random noise under the reservoir quality evaluation parameters based on the first loss function; and calculating the second loss for determining whether the pseudo-evaluation of thermal reservoir quality is true or false based on random noise under the reservoir quality evaluation parameters based on the second loss function.

3. The method for evaluating the quality of thermal reservoirs according to claim 2, characterized in that, Constructing the first loss function corresponding to the first loss includes: calculating the difference between a set value and a pseudo-evaluation of the thermal reservoir quality corresponding to random noise under the reservoir quality evaluation parameter conditions; taking the logarithm of the difference to obtain the logarithmic value of the difference; calculating the first expected value of the logarithmic value of the difference, and constructing the first loss function corresponding to the first loss.

4. The method for evaluating the quality of thermal reservoirs according to claim 3, characterized in that, The set value is configured to be 1.

5. The method for evaluating the quality of thermal reservoirs according to any one of claims 2-4, characterized in that, Constructing the second loss function corresponding to the second loss includes: taking the logarithm of the discrimination probability value corresponding to the random noise under the reservoir quality evaluation parameter conditions to determine whether the thermal reservoir quality is false or not, to obtain the discrimination logarithm value; calculating the second expected value of the logarithm value, and constructing the second loss function corresponding to the second loss.

6. The method for evaluating the quality of thermal reservoirs according to any one of claims 1-5, characterized in that, The step of calculating a first loss corresponding to a pseudo-evaluation of the thermal reservoir quality under the conditions of the reservoir quality evaluation parameters and a second loss corresponding to the determination of whether the pseudo-evaluation of the thermal reservoir quality is true or false, and training the generator network to obtain a generator, includes: constructing objective functions corresponding to the generator network and the discrimination network based on the first loss function corresponding to the first loss and the second loss function corresponding to the second loss; maximizing the first expected value of the first loss function corresponding to the first loss in the objective function and minimizing the second expected value of the second loss function corresponding to the second loss in the objective function, and training the generator network to obtain a generator.

7. The method for evaluating the quality of thermal reservoirs according to any one of claims 1-6, characterized in that, The discriminant network includes: at least one convolutional correlation weight module or multiple cascaded convolutional correlation weight modules; wherein, the convolutional correlation weight module is used to train the generator network using the reservoir quality evaluation parameters, random noise under the conditions of the reservoir quality evaluation parameters, and the reservoir quality evaluation corresponding to the reservoir quality evaluation parameters.

8. The method for evaluating the quality of thermal reservoirs according to any one of claims 1-7, characterized in that, Determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters includes: calculating the variance explained rate and loading coefficients corresponding to the principal components using principal component analysis; calculating the linear combination coefficient matrix based on the variance explained rate and the loading coefficients; calculating the comprehensive score coefficient matrix corresponding to the reservoir quality evaluation parameters based on the eigenvalues ​​of the linear combination coefficient matrix and the covariance matrix corresponding to the reservoir quality evaluation parameters; normalizing the comprehensive score coefficient matrix to obtain the weight values ​​of each reservoir quality parameter; and determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters based on the reservoir quality evaluation parameters and their corresponding weight values.

9. The method for evaluating the quality of thermal reservoirs according to claim 8, characterized in that, The step of determining the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters based on the reservoir quality evaluation parameters and their corresponding weight values ​​of each reservoir quality parameter includes: multiplying the reservoir quality evaluation parameters by their corresponding weight values ​​of each reservoir quality parameter and summing them to determine the thermal reservoir quality evaluation corresponding to the reservoir quality evaluation parameters.

10. The method for evaluating the quality of thermal reservoirs according to any one of claims 1-9, characterized in that, The reservoir quality evaluation parameters corresponding to the known conditions include one or more of the following: clay content, fine sand content, silt content, porosity, permeability, and effective thickness of different reservoirs.

11. A thermal reservoir quality evaluation system, characterized in that, include: The acquisition unit is used to acquire reservoir quality evaluation parameters corresponding to known conditions and thermal reservoir quality evaluation parameters corresponding to the reservoir quality evaluation parameters. The training unit is used to train the generator network using the reservoir quality evaluation parameters, random noise under the reservoir quality evaluation parameter conditions, and the reservoir quality evaluation and discrimination network corresponding to the reservoir quality evaluation parameters, to obtain a generator. The training includes: generating a pseudo-evaluation of the thermal reservoir quality based on the generator network, using the reservoir quality evaluation parameters and random noise under the reservoir quality evaluation parameter conditions; determining the authenticity of the pseudo-evaluation of the thermal reservoir quality based on the discrimination network, using the reservoir quality evaluation parameters, the pseudo-evaluation of the thermal reservoir quality, and the actual thermal reservoir quality evaluation; calculating a first loss for the pseudo-evaluation of the thermal reservoir quality corresponding to the random noise under the reservoir quality evaluation parameter conditions, and determining the authenticity of the pseudo-evaluation of the thermal reservoir quality. The generator network is trained using a second loss corresponding to the true or false nature of the reservoir quality pseudo-evaluation to obtain a generator. The generator network includes: a first shallow feature extraction module, an encoder connected to the shallow feature extraction module, and a decoder connected to both the shallow feature extraction module and the encoder. The shallow feature extraction module extracts the reservoir quality evaluation parameters and first shallow features of random noise under the reservoir quality evaluation parameter conditions. The encoder extracts first deep features and second deep features corresponding to the first shallow features. The decoder generates a pseudo-evaluation of the thermal reservoir quality based on the first shallow features and their corresponding first deep features and second deep features. An evaluation unit is used to evaluate the quality of a thermal reservoir using the generator based on the quality evaluation parameters of the reservoir to be evaluated.

12. A thermal reservoir quality evaluation system, characterized in that, include: processor; A memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the thermal reservoir quality evaluation method according to any one of claims 1 to 10.

13. A thermal reservoir quality evaluation system, characterized in that, include: A computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the thermal reservoir quality evaluation method according to any one of claims 1 to 10.

14. A thermal reservoir quality evaluation system, characterized in that, include: A computer program product, comprising a computer program / instruction that, when executed by a processor, implements the thermal reservoir quality evaluation method according to any one of claims 1 to 10.

Images