Multi-reservoir parameter prediction method based on constraint multi-task deep learning model

Abstract

The application relates to a multi-reservoir parameter prediction method based on a constraint-based multi-task deep learning model, which comprises the following steps: obtaining multi-source seismic information, inputting the multi-source seismic information into a pre-trained multi-task deep learning model, and obtaining reservoir parameter prediction results of porosity, gas saturation and lithology; the multi-task deep learning model comprises a shared feature extractor and no less than three task-specific subnetworks; the process of obtaining the reservoir parameter prediction results by the multi-task deep learning model comprises the following steps: inputting the multi-source seismic information into the shared feature extractor, extracting shared high-dimensional geological features suitable for multiple reservoir parameter prediction tasks by the shared feature extractor, and combining one-dimensional and two-dimensional distribution constraints of data; and processing the shared high-dimensional geological features by each task-specific subnetwork to predict reservoir parameters. Compared with the prior art, the application can overcome the limitations of a traditional single-task deep learning model in processing coupled geological parameters, and realize collaborative prediction of multiple key reservoir parameters.

Description

Technical Field

[0001] This invention relates to the field of seismic data interpretation technology, and in particular to a method for predicting multiple reservoir parameters based on a constrained multi-task deep learning model. Background Technology

[0002] Seismic prediction of reservoir heterogeneity based on seismic data has significant application value in oil and gas resource exploration, reservoir development, and geological modeling. This is because the porosity, gas saturation, and lithological characteristics of subsurface rocks are fundamental physical properties describing subsurface reservoirs, crucial for fluid reserve assessment, reservoir quality evaluation, and reservoir unit delineation. However, in practical seismic reservoir parameter prediction, it is usually necessary to find the most reasonable physical model by minimizing the error between the seismic response and the observed data. Such optimization-based seismic inversion algorithms often struggle to achieve rapid convergence, and the error propagation effect ultimately affects the accuracy of reservoir parameter prediction, thus posing challenges to reservoir evaluation and development decisions.

[0003] Deep learning methods, with their powerful nonlinear mapping and feature extraction capabilities, are widely used to uncover the complex relationships between seismic data and reservoir parameters. However, most existing deep learning methods typically treat the prediction of each reservoir parameter as an independent task. For example, parameters such as lithology, porosity, and gas saturation are often trained and predicted separately using different network models. Chinese patent CN111507048B discloses a method, device, equipment, and system for predicting the gas content of tight sandstone reservoirs, used to improve the accuracy of convolutional neural networks in predicting tight sandstone reservoirs based on pre-stack seismic angle gather data. However, in reality, seismic response is not determined solely by a single reservoir parameter, but rather by the combined effect of multiple reservoir parameters and geological features. These parameters and geological features influence and couple with each other, and the prediction results of some parameters may have a chain reaction on the prediction of other parameters.

[0004] Furthermore, an ideal network model should be able to establish a mapping relationship between labels and input attributes based on limited data, exhibit good generalization ability in unlabeled 3D regions, and the prediction results should conform to geological rationality and follow basic physical laws. Even in a multi-task learning framework, a purely data-driven approach often makes the model prone to ignoring reservoir geological and physical constraints. Moreover, due to limitations in the quality of collected data samples, simultaneous prediction by multiple tasks often leads to parameter crosstalk problems. Therefore, in a multi-task deep learning framework, it is necessary to incorporate geological and data information from more sources, such as spatial information and prior data distribution patterns, to ensure that the physical relationships and distribution patterns between the predicted parameters satisfy existing geological prior characteristics. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model, the method comprising:

[0008] Multi-source seismic information is acquired and input into a pre-trained multi-task deep learning model to obtain reservoir parameter prediction results for porosity, gas saturation, and lithology; the multi-task deep learning model includes a shared feature extractor and no fewer than three task-specific subnetworks.

[0009] The process of obtaining reservoir parameter prediction results by the multi-task deep learning model includes: inputting the multi-source seismic information into a shared feature extractor, using the shared feature extractor to extract shared high-dimensional geological features applicable to multiple reservoir parameter prediction tasks; and using task-specific sub-networks to process the shared high-dimensional geological features to predict the porosity, gas saturation, and lithological reservoir parameters.

[0010] Furthermore, the multi-source seismic information includes pre-stack seismic gathers, post-stack seismic records, P-wave impedance, P-wave / S-wave velocity ratio, seismic sensitivity attributes, and spatiotemporal coordinate information.

[0011] Furthermore, the training process of the multi-task deep learning model includes:

[0012] A training dataset was constructed by using well logging data to collect porosity, gas saturation, and lithology data, combined with multi-source seismic information and preprocessing.

[0013] A multi-task deep learning model is constructed and trained based on a multi-task loss function, a two-dimensional joint distribution consistency constraint, and a physical correlation constraint.

[0014] After training, the multi-task deep learning model is subjected to blind well testing to evaluate its prediction performance at locations where no wells are found, and the multi-task deep learning model is optimized based on the evaluation results.

[0015] Furthermore, the preprocessing includes:

[0016] Based on the established well-seismic relationship, the porosity, gas saturation, and lithological data obtained from well logging, as well as multi-source seismic information, are converted to time-depth and uniformly resampled to millisecond-level time domain sampling intervals.

[0017] Multi-source seismic information and target reservoir parameter labels are synchronously interpolated to a sampling interval of milliseconds;

[0018] The logging curves for porosity and gas saturation are smoothed using a millisecond-level sliding window.

[0019] Centered on each sampling point on the logging trajectory, a corresponding three-dimensional geological slice is extracted in its neighborhood using a sliding window method. A set of sample pairs suitable for the three-dimensional sequence-to-sequence mapping paradigm is constructed. The input features include processed multi-source seismic information, and the output features include processed porosity, gas saturation, and lithological data.

[0020] The min-max normalization method was used to scale the processed porosity, gas saturation, lithology data, and multi-source seismic information to the range of 0-1.

[0021] Furthermore, the expression for the multi-task loss function is:

[0022]

[0023] Where N represents the total number of samples, MSE represents the mean squared error loss function, CrossEntropy represents the cross-entropy loss function, and P pi P si and P li Y represents the porosity, gas saturation, and lithological prediction values ​​of the i-th sample, respectively. pi Y si and Y li W represents the true labels corresponding to the porosity, gas saturation, and lithological prediction values ​​of the i-th sample, respectively. p W s and W l These represent the weights of the loss function for porosity, gas saturation, and lithology prediction tasks, respectively.

[0024] Furthermore, the expression for the two-dimensional joint distribution consistency constraint is:

[0025]

[0026] Where χ represents the set of all predicted porosity and gas saturation values, (p p ,s p ) represents the specific predicted porosity and gas saturation values, and ξ represents the set of all actual porosity and gas saturation values. t ,s t P represents the specific actual porosity and gas saturation value. p and P t They are P pred (p p ,s p ) and P true (p t ,st γ(p) is an abbreviation representing the probability density distribution of porosity and gas saturation between predicted and actual values. p ,s p ;p t ,s t ) is the joint distribution matrix, describing the gas saturation value of the predicted porosity (p) p ,s p ) to the true porosity gas saturation value (p t ,s t The mapping relationship of c(p) p ,s p ;p t ,s t ) is the cost function that measures the square of the Euclidean distance between the predicted and actual values ​​of porosity and gas saturation, and A is a regularization term used to improve computational efficiency and mitigate numerical instability.

[0027]

[0028]

[0029] Where ε is the regularization weight, p p s p p t and s t P represents the predicted porosity, gas saturation, actual porosity, and gas saturation, respectively. pred and P true The two-dimensional probability density functions representing porosity and gas saturation, respectively, are μ and μ'. pp μ sp σ pp and σ sp The mean and standard deviation of predicted porosity and gas saturation are represented in μ. pt μ st σ pt and σ st ρ represents the mean and standard deviation of the corresponding true porosity and gas saturation. p and ρ s These represent the correlation coefficients between predicted porosity and saturation, and between actual porosity and saturation, respectively.

[0030] Furthermore, the expression for the physical correlation constraint is:

[0031] Sat=23.48×In(Por)-0.82×Por+40.06

[0032] Wherein, Por and Sat represent porosity and gas saturation, respectively.

[0033] Furthermore, when collecting the porosity, gas saturation, and lithological data using well logging, geological spatiotemporal constraint information is also introduced. The introduction process includes:

[0034] The absolute location information of each sampling point is extracted from the original multi-source seismic data. This absolute location information includes three-dimensional coordinates along the seismic line direction, the transverse seismic line direction, and the time dimension. Relative time information is calculated based on three key seismic horizons interpreted by experts. The temporal or depth-related location information of each sampling point within its corresponding stratum is quantified. The quantification expression is as follows:

[0035]

[0036] Among them, time in,x,t This represents the current sampling point, where {inline, xline, time} is the set of sample locations within the study area, and horizontal. i+1 and Horizon i These represent the current time sampling point, time, respectively. in,x,t The upper and lower seismic strata of the strata in which it is located.

[0037] Furthermore, when training the multi-task deep learning model, the two-dimensional joint distribution consistency constraint and physical correlation constraint are fused with the multi-task loss function to obtain a joint optimization objective function with multiple constraint mechanisms. The expression of the joint optimization objective function is as follows:

[0038] L total =L Multi +W d L W λ d

[0039] Among them, W d The weights λ represent the joint data distribution constraints. d It is the scaling factor, L W Indicates constraints.

[0040] Furthermore, the UNet network is used as a shared feature extractor.

[0041] According to another aspect of the invention, a method is provided

[0042] Compared with the prior art, the beneficial effects of the present invention include:

[0043] 1. This invention, based on a multi-task deep learning model, can fully decouple the complex mapping relationship between seismic response and multiple reservoir parameters, and deeply explore the statistical similarity and physical consistency among parameters, achieving collaborative prediction of key reservoir parameters such as porosity, gas saturation, and lithology. This overcomes the limitations of traditional single-task deep learning models in handling coupled geological parameters. In terms of model architecture design, this invention uses a UNet network as a shared feature extractor in multi-task learning to mine shared features among different reservoir parameters (such as porosity, gas saturation, and lithology). Subsequently, the extracted shared features are input into... Each task-specific subnetwork enables collaborative seismic prediction of multiple reservoir parameters. Seismic response is essentially a comprehensive reflection of various reservoir physical parameters and geological characteristics, and these parameters have complex physical coupling relationships. Multi-task learning, by sharing the underlying feature space and utilizing the correlation between tasks, can effectively improve the model's decoupling of reservoir parameters under complex geological conditions. This invention not only improves the prediction accuracy of single reservoir parameters, but also optimizes the prediction results from multiple dimensions such as statistical consistency, physical self-consistency, and spatial continuity, providing efficient and reliable technical support for reservoir evaluation, geological modeling, and development decision-making in oil and gas exploration.

[0044] 2. This invention uses pre-stack seismic records, post-stack seismic data, seismic inversion results, and sensitive seismic attributes as multi-source input features to construct an input data volume with rich geological information.

[0045] 3. In the preprocessing stage, this invention takes each sampling point on the logging trajectory as the center and extracts the corresponding three-dimensional geological slices in its neighborhood using a sliding window method. It constructs a set of sample pairs suitable for the three-dimensional sequence-to-sequence mapping paradigm, and innovatively constructs a multi-dimensional spatiotemporal dataset that integrates geological time, stratigraphy, and spatial dimensions. This more realistically simulates the spatial distribution characteristics of actual geological structures, enhances the model's ability to understand and generalize the spatial distribution characteristics of multiple reservoir parameters under complex geological conditions, and provides a new modeling idea and technical path for reservoir parameter prediction tasks.

[0046] 4. This invention innovatively introduces two-dimensional joint distribution consistency constraints and physical correlation constraints. The two-dimensional joint distribution consistency constraints ensure the statistical consistency of the prediction results in the joint distribution space, while the physical correlation constraints construct deterministic correlations between parameters. This composite constraint strategy, through the organic combination of mathematical statistical constraints and physical law constraints, not only improves the prediction accuracy of multiple reservoir parameters, but also systematically improves the physical self-consistency between multiple parameter predictions.

[0047] 5. The model of this invention can output the three-dimensional distribution results of porosity, gas saturation and lithology in one go. Compared with the traditional parameter-by-parameter inversion method, the calculation efficiency is improved, providing fast and comprehensive data support for reservoir development scheme formulation. Attached Figure Description

[0048] Figure 1 This is a flowchart of the method of the present invention;

[0049] Figure 2 This is a flowchart illustrating the training and application process of the multi-task deep learning model in this invention.

[0050] Figure 3 This is a comparison chart of the cross-validation prediction results of a multi-task deep learning model and other single-task learning models on data from five blind wells in one embodiment of the present invention.

[0051] Figure 4 This is a schematic diagram of porosity prediction results based on a multi-task deep learning model.

[0052] Figure 5 This is a schematic diagram of the porosity prediction results based on a single-task learning model.

[0053] Figure 6 This is a schematic diagram of the gas saturation prediction results based on a multi-task deep learning model.

[0054] Figure 7 This is a schematic diagram of the gas saturation prediction results based on a single-task learning model.

[0055] Figure 8 This is a schematic diagram of lithology prediction results based on a multi-task deep learning model.

[0056] Figure 9 This is a schematic diagram of lithology prediction results based on a single-task learning model. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0058] Example 1

[0059] This embodiment discloses a method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model, the method being as follows: Figure 1 As shown, it includes:

[0060] Step S1: Obtain multi-source seismic information;

[0061] Step S2: Input multi-source seismic information into the shared feature extractor of the multi-task deep learning model, and use the shared feature extractor to extract shared high-dimensional geological features suitable for multiple reservoir parameter prediction tasks.

[0062] Step S3: Use the task-specific sub-networks of the multi-task deep learning model to process shared high-dimensional geological features and predict porosity, gas saturation, and lithology prediction results.

[0063] The multi-task deep learning model includes a shared feature extractor and at least three task-specific sub-networks. The UNet network is used as the shared feature extractor for the multi-task deep learning model. The task-specific sub-networks of the multi-task deep learning model include a porosity sub-network, a gas saturation sub-network, and a lithology sub-network. Each task-specific sub-network contains an input layer, a nonlinear transformation layer, a classification layer, and an output layer. The shared high-dimensional geological features are input into each task-specific sub-network and then processed sequentially through the input layer, nonlinear transformation layer, classification layer, and output layer to output porosity, gas saturation, or lithological reservoir parameters.

[0064] In this embodiment, multi-source seismic information is obtained through seismic exploration techniques (data acquisition, inversion, attribute extraction). Multi-source seismic information includes pre-stack seismic gathers, post-stack seismic records, P-wave impedance, P-wave / S-wave velocity ratio, seismic sensitivity attributes, and spatiotemporal coordinate information.

[0065] Well logging technology involves placing sensors (logging tools) into the borehole to directly measure the physical response of the formation around the wellbore and obtain logging data. The logging data includes high-precision reservoir parameters such as porosity, gas saturation, and lithology, which serve as label data for model training.

[0066] In this embodiment, the multi-task deep learning model is a multi-task learning multi-reservoir parameter prediction model based on joint data distribution and physical constraints. Its training and application process is as follows: Figure 2 As shown, the process includes: seismic data preprocessing, 3D sample construction, multi-task network construction, introduction of joint data distribution and physical constraint mechanisms, and final prediction of the spatial distribution of 3D reservoir parameters. The specific steps are described below:

[0067] 1. Three-dimensional input of multi-source seismic information and construction of training labels

[0068] Porosity, gas saturation, and lithological data were collected using well logging technology and preprocessed in conjunction with multi-source seismic information from the entire 3D work area to construct a training dataset. The multi-source seismic information included pre-stack seismic gathers, post-stack seismic records, seismic attributes, seismic inversion results, and seismic spatiotemporal information (Xline, Inline, time, and relative time).

[0069] Preprocessing includes:

[0070] Based on the established well-seismic relationship, time-depth conversion was performed on multi-source seismic information, porosity, gas saturation and lithological data obtained from well logging, and uniform resampling was performed to a 2-millisecond time domain sampling interval to ensure that the porosity, gas saturation and lithological data are consistent with the multi-source seismic information in the time dimension.

[0071] To increase the amount of sample data for deep learning training, multi-source seismic information and target reservoir parameter labels are synchronously interpolated to a 1-millisecond sampling interval.

[0072] Considering the scale difference between well logging reservoir parameters and seismic dominant frequency, a 20-millisecond sliding window is used to smooth the well logging curves of porosity and gas saturation to reduce high-frequency noise interference and achieve matching with the seismic response frequency band.

[0073] Considering that seismic response is essentially a spatial combination of underground three-dimensional geological features, a three-dimensional sequence-to-sequence mapping paradigm is adopted to construct each training sample into a small three-dimensional volume with spatial context information. Specifically, a three-dimensional seismic volume with a size of 3×3×19 is extracted from the neighborhood of each sampling point on the well logging trajectory: 3 traces are taken in the Inline and Xline directions, and 19 sampling points are taken in the time direction to form an input sample containing seismic gathers and their spatial location information.

[0074] To accelerate network convergence, the min-max normalization method is used to scale the processed porosity, gas saturation, lithology data, and multi-source seismic information to the range of 0-1.

[0075] To enhance the model's ability to model the temporal evolution characteristics and spatial distribution patterns of seismic data, geological spatiotemporal constraint information is introduced during the preprocessing stage. The specific process is as follows:

[0076] The absolute location information of each sampling point is extracted from the original seismic data volume, including three-dimensional coordinates along the seismic line direction (Inline), the transverse seismic line direction (Xline), and the time dimension. Simultaneously, relative time information is calculated to characterize the relative sedimentary sequence and geological age relationship of the strata in the time domain. Based on three key seismic horizons interpreted by experts, relative time information is calculated to quantify the temporal or depth-related location information of each sampling point within its corresponding stratum. The quantification expression is as follows:

[0077]

[0078] Among them, time in,x,t This represents the current sampling point, where {inline, xline, time} is the set of sample locations within the study area, and horizontal. i+1 and Horizon iThese represent the current time sampling point, time, respectively. in,x,t The upper and lower seismic strata of the strata in which it is located.

[0079] 2. Construction of Multi-Task Learning Network Architecture

[0080] A multi-task deep learning model is constructed, comprising a shared feature extractor and three task-specific sub-networks. The UNet network is used as the shared feature extractor in multi-task learning to mine shared features among different reservoir parameters (such as porosity, gas saturation, and lithology). Seismic response is essentially a comprehensive reflection of multiple reservoir physical parameters and geological characteristics, and these parameters have complex physical coupling relationships. Multi-task learning, by sharing the underlying feature space and utilizing the correlation between tasks, can effectively improve the model's decoupling of reservoir parameters under complex geological conditions.

[0081] In this embodiment, the multi-task deep learning model is actually a shared-dedicated hybrid neural network architecture based on UNet built within a multi-task deep learning framework, used to achieve collaborative prediction of multiple reservoir parameters (such as porosity, gas saturation, and lithology). During the model training phase, the network calculates the error loss between the predicted output of each task and the corresponding true label, and uses a weighted loss function to balance and adjust the training contribution of each task.

[0082] The expression for the multi-task loss function is:

[0083]

[0084] Where N represents the total number of samples, MSE represents the mean squared error loss function, and CrossEntropy represents the cross-entropy loss function. In the formula, for parameter prediction tasks (such as porosity and gas saturation), the mean squared error (MSE) loss function is used; while for lithology identification, a classification task, the cross-entropy loss function is used to evaluate the error. pi P si and P li Y represents the porosity, gas saturation, and lithological prediction values ​​of the i-th sample, respectively. pi Y si and Y li W represents the true labels corresponding to the porosity, gas saturation, and lithological prediction values ​​of the i-th sample, respectively. p W s and W l denoted by , respectively, the weights of the loss function for porosity, gas saturation, and lithology prediction tasks. To accelerate the parameter optimization process and narrow the search space, we performed a grid search on the weights of each task within the range {0.3, 0.5, 0.7, 0.9}, and the final determined weights are as follows: Porosity W p=0.3, gas saturation W s =0.3, Lithology W l =0.7. In addition, a scaling factor λ is introduced. p =0.01, λ s =1 and λ l =0.005 is used to adjust the loss functions for porosity, gas saturation, and lithology tasks, respectively, to ensure that these three losses are on the same order of magnitude, thereby avoiding the loss of a certain task from dominating the overall optimization process.

[0085] 3. Construction of Consistency Constraints for Two-Dimensional Joint Data Distribution

[0086] To enhance the generalization ability of multi-task deep learning models under complex geological conditions, a joint data distribution consistency constraint mechanism is introduced. The two-dimensional Wasserstein distance between the predicted output and the label in the porosity-gas saturation range is calculated as a distribution matching metric and incorporated into the loss function to guide the network to learn prediction results that better conform to geostatistical laws. Specifically, all samples are extracted from the three-parameter prediction results output by the network, and only the center point of each three-dimensional sample is selected as the predicted value and label value. Assuming the data follows a Gaussian distribution, the two-dimensional probability density functions of the predicted value and the true label's porosity and gas saturation are calculated using the following formulas:

[0087]

[0088] Where, p p s p p t and s t P represents the predicted porosity, gas saturation, actual porosity, and gas saturation, respectively. pred and P true The two-dimensional probability density functions representing porosity and gas saturation, respectively, are μ and μ'. pp μ sp σ pp and σ sp The mean and standard deviation of predicted porosity and gas saturation are represented in μ. pt μ st σ pt and σ st ρ represents the mean and standard deviation of the corresponding true porosity and gas saturation. p and ρ s These represent the correlation coefficients between predicted porosity and saturation, and between actual porosity and saturation, respectively. These parameters are calculated from the predicted values ​​and well labels in each batch of data.

[0089] The Wasserstein distance is then applied to quantify the difference between the predicted results and the true labels in the two-dimensional joint distribution space of porosity and gas saturation. The core idea of ​​this metric is to measure the minimum "transportation cost" required to transform one distribution into another, thus providing an efficient and intuitive indicator for assessing the similarity between two distributions. The expression for the consistency constraint of the two-dimensional joint distribution is:

[0090]

[0091] Where χ represents the set of all predicted porosity and gas saturation values, (P p ,s p ) represents the specific predicted porosity and gas saturation values, and ξ represents the set of all actual porosity and gas saturation values, (P t ,S t ) represents the specific actual porosity and gas saturation value, p p and p t They are P pred (P p ,S p ) and p true (p t ,s t γ(p) is an abbreviation representing the probability density distribution of porosity and gas saturation between predicted and actual values. p ,s p ;p t ,s t ) is the joint distribution matrix, describing the gas saturation value of the predicted porosity (p) p ,s p ) to the true porosity gas saturation value (p t ,s t The mapping relationship of c(p) p ,s p ;p t ,s t ) is the cost function that measures the square of the Euclidean distance between the predicted and actual values ​​of porosity and gas saturation, and A is a regularization term used to improve computational efficiency and mitigate numerical instability.

[0092]

[0093] Where ε is the regularization weight, with a value of 0.01.

[0094] 4. Construction of Physical Association Constraints

[0095] To enhance the physical consistency of multi-task deep learning models in predicting multiple reservoir parameters, this invention introduces a physical constraint mechanism based on the empirical relationship between porosity and gas saturation. The first-order empirical relationship between porosity and gas saturation data in the tags is obtained through cross-plot analysis; the expression for the physical correlation constraint is as follows:

[0096] Sat=23.48×In(Por)-0.82×Por+40.06

[0097] Wherein, Por and Sat represent porosity and gas saturation, respectively.

[0098] This first-order empirical relationship is then used as a fitting operator. During network training, the gas saturation predicted by the multi-task network is converted into the corresponding pseudo-porosity through this fitting operator. Subsequently, this pseudo-porosity is used as an enhancement feature and connected to the input of the porosity prediction sub-network. In this way, the model not only relies on the original input seismic information when predicting porosity, but also integrates the physical correlation features transformed from the gas saturation prediction results, thereby enhancing the coupling modeling ability between parameter predictions and improving the accuracy and physical rationality of porosity prediction.

[0099] 5. Joint optimization and application of multi-task learning models that integrate data distribution with physical constraints

[0100] When training a multi-task deep learning model, the two-dimensional joint distribution consistency constraint and physical correlation constraint are fused with the multi-task loss function to obtain a joint optimization objective function with multiple constraint mechanisms. The expression of the joint optimization objective function for training the multi-task deep learning model is as follows:

[0101] L total =L Multi +W d L W λ d

[0102] Among them, W d The weights λ represent the joint data distribution constraints. d It is the scaling factor, L W Indicates constraints.

[0103] In this embodiment, λ is obtained after hyperparameter tuning. d =0.001, W d =0.3. After training, the trained multi-task deep learning model with multiple constraints was used to conduct blind well tests to evaluate its predictive performance at locations where no wells have been found. Furthermore, it was applied to 3D seismic data volumes to achieve simultaneous prediction of three reservoir parameters—porosity, gas saturation, and lithology—in three-dimensional space.

[0104] Figure 3 This is a comparison of the cross-validation prediction results of the multi-task deep learning model based on the method in this embodiment and other single-task learning models on data from five blind wells. From left to right in each well, the results are: single-task and multi-task porosity labels and prediction results, single-task and multi-task gas saturation labels and prediction results, lithology labels, single-task lithology prediction results, and multi-task lithology prediction results. Black arrows indicate areas where multi-task prediction is better, gray arrows indicate areas where multi-task performance is slightly inferior, and black boxes indicate areas where the multiple parameters predicted by multi-task have better self-consistency. Analysis of the statistical and prediction results of single-task and multi-task methods demonstrates the effectiveness of the proposed constraint-based multi-task strategy. Specifically, the multi-task learning multi-reservoir parameter prediction strategy based on joint data distribution and physical constraints exhibits superior performance in terms of prediction accuracy, spatial continuity, and physical consistency for the three reservoir parameters.

[0105] Figures 4-9 The following is a schematic diagram of the three-parameter seismic profile prediction results based on a single-task learning model and a multi-task deep learning model: Figure 4 For porosity prediction results (multi-task), Figure 5 For porosity prediction results (single task), Figure 6 For gas saturation prediction results (multi-task), Figure 7 The results are for gas saturation prediction (single task). Figure 8 For lithological prediction results (multi-task), Figure 9 This represents lithological prediction results (single mission), where wells W29 and W27 indicate blind wells. Gray circles indicate that the three parameters predicted by the multi-mission are more accurate and more consistent with the well labels. The gray irregular closed shape on the porosity prediction profile indicates stronger geological continuity of the multi-mission prediction results in this area. Black circles indicate that, compared to single missions, the porosity and gas saturation predictions of the multi-missions show better self-consistency; that is, low porosity corresponds to low gas saturation, while high porosity in non-single mission predictions corresponds to low gas saturation.

[0106] Example 2

[0107] Based on Embodiment 1, this embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, the one or more programs including instructions for executing the multi-reservoir parameter prediction method of the aforementioned constraint-based multi-task deep learning model.

[0108] At the hardware level, the electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the aforementioned multi-layer parameter prediction method for a constraint-based multi-task deep learning model. Of course, besides software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0109] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0110] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0111] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model, characterized in that, The method includes: Multi-source seismic information is acquired and input into a pre-trained multi-task deep learning model to obtain reservoir parameter prediction results for porosity, gas saturation, and lithology; the multi-task deep learning model includes a shared feature extractor and no fewer than three task-specific subnetworks. The process of obtaining reservoir parameter prediction results by the multi-task deep learning model includes: inputting the multi-source seismic information into a shared feature extractor, using the shared feature extractor to extract shared high-dimensional geological features applicable to multiple reservoir parameter prediction tasks; and using task-specific sub-networks to process the shared high-dimensional geological features to predict the porosity, gas saturation, and lithological reservoir parameters. The training process of the multi-task deep learning model includes: A training dataset was constructed by using well logging data to collect porosity, gas saturation, and lithology data, combined with multi-source seismic information and preprocessing. A multi-task deep learning model is constructed and trained based on a multi-task loss function, a two-dimensional joint distribution consistency constraint, and a physical correlation constraint. After training, the multi-task deep learning model is subjected to blind well testing to evaluate its prediction performance at locations where no wells are found, and the multi-task deep learning model is optimized based on the evaluation results. The expression for the two-dimensional joint distribution consistency constraint is: in, This represents the set of all predicted porosity and gas saturation values. This represents the specific predicted porosity and gas saturation values. This represents the set of all true porosity and gas saturation values. This represents the specific actual porosity and gas saturation values. and They are and The abbreviation represents the probability density distribution of porosity and gas saturation between predicted and actual values. It is a joint distribution matrix that describes the gas saturation values ​​predicted from the porosity. To the true porosity gas saturation value The mapping relationship, The cost function is used to measure the squared Euclidean distance between predicted and actual values ​​of porosity and gas saturation. It is a regularization term used to improve computational efficiency and mitigate numerical instability; in, It is the regularization weight. , , and These represent the predicted porosity, gas saturation, actual porosity, and gas saturation, respectively. and Two-dimensional probability density functions representing porosity and gas saturation, respectively, for predicted and actual values. , , and This represents the mean and standard deviation of the predicted porosity and gas saturation. , , and This represents the mean and standard deviation of the corresponding true porosity and gas saturation. and These represent the correlation coefficients between predicted porosity and saturation, and between actual porosity and saturation, respectively.

2. The method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model according to claim 1, characterized in that, The multi-source seismic information includes pre-stack seismic gathers, post-stack seismic records, P-wave impedance, P-wave / S-wave velocity ratio, seismic sensitivity attributes, and spatiotemporal coordinate information.

3. The method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model according to claim 1, characterized in that, The preprocessing includes: Based on the established well-seismic relationship, the porosity, gas saturation, and lithological data obtained from well logging, as well as multi-source seismic information, are converted to time-depth and uniformly resampled to millisecond-level time domain sampling intervals. Multi-source seismic information and target reservoir parameter labels are synchronously interpolated to a sampling interval of milliseconds; The logging curves for porosity and gas saturation are smoothed using a millisecond-level sliding window. Centered on each sampling point on the logging trajectory, a corresponding three-dimensional geological slice is extracted in its neighborhood using a sliding window method. A set of sample pairs suitable for the three-dimensional sequence-to-sequence mapping paradigm is constructed. The input features include processed multi-source seismic information, and the output features include processed porosity, gas saturation, and lithological data. The min-max normalization method was used to scale the processed porosity, gas saturation, lithology data, and multi-source seismic information to the range of 0-1.

4. The method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model according to claim 1, characterized in that, The expression for the multi-task loss function is: in, Represents the total number of samples. This represents the mean squared error loss function. Represents the cross-entropy loss function. , and They represent the first i Porosity, gas saturation, and lithological prediction values ​​for each sample. , and They represent the first i The true labels corresponding to the porosity, gas saturation, and lithological prediction values ​​of each sample. , and These represent the weights of the loss function for porosity, gas saturation, and lithology prediction tasks, respectively.

5. The method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model according to claim 1, characterized in that, The expression for the physical correlation constraint is: in, Por and Sat These represent porosity and gas saturation, respectively.

6. The method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model according to claim 1, characterized in that, When collecting porosity, gas saturation, and lithological data using well logging, geological spatiotemporal constraint information is also introduced. The introduction process includes: The absolute location information of each sampling point is extracted from the original multi-source seismic data. This absolute location information includes three-dimensional coordinates along the seismic line direction, the transverse seismic line direction, and the time dimension. Relative time information is calculated based on three key seismic horizons interpreted by experts. The temporal or depth-related location information of each sampling point within its corresponding stratum is quantified. The quantification expression is as follows: in, Indicates the current sampling point. The set of sample locations within the study area. and These represent the current time sampling points. The upper and lower seismic strata of the strata in which it is located.

7. The method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model according to claim 1, characterized in that, When training the multi-task deep learning model, the two-dimensional joint distribution consistency constraint and physical correlation constraint are fused with the multi-task loss function to obtain a joint optimization objective function with multiple constraint mechanisms. The expression of the joint optimization objective function is as follows: in, The weights represent the joint data distribution constraints. It is the scaling factor. Indicates constraints.

8. The method for predicting multiple reservoir parameters based on a constraint-based multi-task deep learning model according to claim 1, characterized in that, The UNet network is used as the shared feature extractor of the multi-task deep learning model. The task-specific sub-networks of the multi-task deep learning model include a porosity sub-network, a gas saturation sub-network, and a lithology sub-network. Each task-specific sub-network contains an input layer, a nonlinear transformation layer, a classification layer, and an output layer. After the shared high-dimensional geological features are input into each task-specific sub-network, they are processed sequentially through the input layer, the nonlinear transformation layer, the classification layer, and the output layer to output porosity, gas saturation, or lithological reservoir parameters.

Images