Oilfield oil saturation prediction method fused with attention mechanism and related device

Abstract

The invention discloses an attention mechanism fused oil field oil saturation prediction method and a related device, and belongs to the technical field of oil exploration. Collecting reservoir influence factor characteristics to obtain a data set; establishing a Ctrl-CLSTM model according to the data set, and introducing an attention mechanism; giving different attention weights to reservoir influence factor characteristics and capturing reservoir production system spatio-temporal characteristics; and taking the oil saturation, the well location distribution and the injection-production working system as input of the Ctrl-CLSTM model, and predicting to obtain an oil saturation predicted value. According to the method, the network model fusing multivariate data is constructed, an attention mechanism is introduced to screen the influence degree of external factors on the saturation spatial and temporal distribution, the oil saturation distribution rule of the oil reservoir is accurately predicted, different production system modes are dynamically regulated and controlled, and saturation distribution prediction is more accurate and practical.

Description

Technical Field

[0001] The present invention belongs to the technical field of oil exploration, and relates to an oil saturation prediction method and related device for oil fields that integrates an attention mechanism. Background Art

[0002] In recent years, reservoir numerical simulation technology has developed rapidly. Through fine reservoir history matching, the distribution law of oil saturation can be revealed. However, the computational complexity of the model will affect its running time, and it is limited by the accuracy of reservoir description and model fitting efficiency, and cannot achieve an ideal dynamic prediction effect in actual oil field applications.

[0003] With the rise of deep learning technology and the improvement of computing power, people have begun to continuously try to apply these emerging technologies to prediction and analysis tasks for large-scale spatio-temporal data. Among them, Patent CN104732303A is an oil field production prediction method based on a dynamic radial basis function neural network, which dynamically adjusts the RBF neural network structure using the sensitivity method and corrects the model error using the state transition probability matrix. Although this method avoids the problem of too many or too few hidden layer neurons in the RBF, it is difficult to select a suitable RBF center during the implementation process. Patent CN115169761A is a complex well pattern oil field production prediction method based on GNN and LSTM. This method generates feature vectors for each production well and each injection well, and through trained first and second MLPs, expands and encodes the feature vectors to obtain the encoded features of each production well and each injection well, and finally combines the features and adopts the trained production. However, LSTM also has certain defects: extremely high computational complexity, difficult to interpret, and requires a large amount of data for training. Patent CN114692918A is an oil field single well production prediction method based on the decline curve, which uses a combination of matter element analysis and BP neural network to make predictions, establishes a matter element model, calculates the correlation degree with the typical feature library, accurately judges the decline curve type, and takes the decline curve type into account in the input variables to achieve the predicted production. However, the process of establishing the decline curve class set in this method is relatively complex, and the BP neural network is very sensitive to the initial weights and is extremely likely to converge to a local minimum.

[0004] In summary, the prediction models in the prior art have low prediction accuracy for saturation, high complexity, and are difficult to interpret; therefore, there is an urgent need to develop an oil field oil saturation prediction method with higher prediction accuracy. Summary of the Invention

[0005] The purpose of the present invention is to solve the technical problems in the prior art that the prediction model has low prediction accuracy for saturation, high complexity, and is difficult to interpret, and to provide an oil field oil saturation prediction method and related device that integrates an attention mechanism.

[0006] To achieve the above object, the present invention is implemented by the following technical solutions: In a first aspect, the present invention provides an oil saturation prediction method for oilfields integrating an attention mechanism, comprising the following steps: Collect the characteristics of reservoir influencing factors to obtain a data set; Establish a Ctrl-CLSTM model based on the data set and introduce an attention mechanism; Assign different attention weights to the characteristics of reservoir influencing factors and capture the spatio-temporal characteristics of the reservoir production system; Use the oil saturation, well location distribution, and injection-production working system as the inputs of the Ctrl-CLSTM model to predict the oil saturation prediction value.

[0007] In a second aspect, the present invention provides an oil saturation prediction system for oilfields integrating an attention mechanism, comprising: A feature collection module for collecting the characteristics of reservoir influencing factors to obtain a data set; A modeling module for establishing a Ctrl-CLSTM model based on the data set and introducing an attention mechanism; A weight assignment module for assigning different attention weights to the characteristics of reservoir influencing factors and capturing the spatio-temporal characteristics of the reservoir production system; A prediction module that uses the oil saturation, well location distribution, and injection-production working system as the inputs of the Ctrl-CLSTM model to predict the oil saturation prediction value.

[0008] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the above method are implemented.

[0009] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the steps of the above method are implemented.

[0010] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses an oil saturation prediction method and related device for oil fields integrating an attention mechanism. By constructing a network model integrating multiple data, various influencing factors such as injection-production working parameters, permeability, and well location distribution are comprehensively considered, and combined with historical saturation distribution image data, accurate prediction of the oil saturation distribution in various oil field systems is achieved. The model uses the oil saturation sequence as a prerequisite to capture the influence of spatial information such as oil saturation distribution and well location distribution on the oil field production; an attention mechanism is introduced to divide the importance degree of different reservoir factors, enhance the factors with large attention weights, and weaken the factors with small attention weights, realizing the importance analysis of external factors, providing a solution for the decision-making of injection-production working systems, accurately predicting the oil saturation distribution law of the oil reservoir, dynamically regulating different production system modes, and making the saturation distribution prediction more accurate and practical. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0012] Figure 1 is a flowchart of the method of the present invention; Figure 2 is a schematic diagram of the system of the present invention; Figure 3 is an architecture diagram of a controllable spatio-temporal memory network according to an embodiment of the present invention; Figure 4 is a schematic diagram of an attention weight extraction model according to an embodiment of the present invention; Figure 5 is a diagram of an input sequence and a numerical simulation sequence according to an embodiment of the present invention; Figure 6 is a diagram comparing the prediction results and errors of each model according to an embodiment of the present invention; Figure 7 is an oil saturation sequence diagram of reservoir A according to an embodiment of the present invention; Figure 8 is an input sequence diagram of reservoir B according to an embodiment of the present invention Figure 9 is a diagram of the prediction result and error of reservoir B according to an embodiment of the present invention; Figure 10 is a schematic diagram of the structure of a computer device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0013] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. Components of the embodiments of the present invention generally described and illustrated in the drawings herein can be arranged and designed in a variety of different configurations.

[0014] Therefore, the detailed description of the embodiments of the present invention provided in the drawings below is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0015] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not require further definition and explanation in subsequent drawings.

[0016] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper", "lower", "horizontal", "inner", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the inventive product is customarily placed during use, it is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention. In addition, terms such as "first", "second", etc. are only used for descriptive distinction and should not be construed as indicating or implying relative importance.

[0017] In addition, if the term "horizontal" appears, it does not mean that the component is required to be absolutely horizontal, but it can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and does not mean that the structure must be completely horizontal, but it can be slightly inclined.

[0018] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly specified and defined, if terms such as "set", "installed", "connected", "coupled" are to be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0019] The present invention will be further described in detail below with reference to the accompanying drawings: See Figure 1, an embodiment of the present invention discloses an oil saturation prediction method for oilfields integrating an attention mechanism, including the following steps: S1, collect the characteristics of reservoir influencing factors to obtain a data set; S2, establish a Ctrl-CLSTM model according to the data set and introduce an attention mechanism; S3, assign different attention weights to the characteristics of reservoir influencing factors and capture the spatio-temporal characteristics of the reservoir production system; S4, take the oil saturation, well location distribution, and injection-production working system as the input of the Ctrl-CLSTM model, and predict the oil saturation prediction value.

[0020] In a feasible embodiment of the present invention, the collecting the characteristics of reservoir influencing factors to obtain a data set specifically includes: collecting the characteristics of reservoir influencing factors and performing normalization processing, integrating the characteristics of reservoir influencing factors according to the time series, and then dividing them according to the time ratio to obtain a data set. The characteristics of reservoir influencing factors include oil saturation, permeability, and porosity.

[0021] In a feasible embodiment of the present invention, the establishing a Ctrl-CLSTM model according to the data set and introducing an attention mechanism specifically includes: establishing a Ctrl-CLSTM model according to the data set, introducing an attention mechanism on the basis of Ctrl-CLSTM, and extracting the importance weights of the relative spatial states of multiple reservoirs on the oilfield production; then feedback the extracted importance degree to the data characteristics, enhance the characteristics with high importance degree, and suppress the characteristics with low importance degree. The relative spatial states of the multiple reservoirs include well location distribution, oil saturation, permeability, and porosity.

[0022] In a feasible embodiment of the present invention, the assigning different attention weights to the characteristics of reservoir influencing factors and capturing the spatio-temporal characteristics of the reservoir production system specifically includes: performing preliminary feature extraction on the sequences in the data set through a convolutional neural network, then dividing the characteristics of reservoir influencing factors according to the importance degree, sorting the weights from large to small, taking a weight as a threshold, summing the first i attention weights until it is greater than the threshold, and dividing the reservoir factor characteristics corresponding to the first i attention weights into main influencing factors, and the rest as non-main influencing factors; finally, applying the obtained attention weight values to the corresponding reservoir factors.

[0023] In a feasible embodiment of the present invention, the predicting the oil saturation prediction value specifically includes: S401, during the prediction process, update the states of each time memory unit of the front and back time series according to the time memory state transition path equation of Ctrl-CLSTM; S402. Update the states of the spatial memory units in the front and back time series according to the spatial memory state transition path equation of Ctrl-CLSTM.

[0024] See Figure 2 , an oil saturation prediction system for oilfields integrating an attention mechanism is disclosed in an embodiment of the present invention, including: A feature collection module for collecting reservoir influencing factor features to obtain a data set; A modeling module for establishing a Ctrl-CLSTM model according to the data set and introducing an attention mechanism; A weight assignment module for assigning different attention weights to the reservoir influencing factor features and capturing the spatio-temporal features of the reservoir production system; A prediction module, taking the oil saturation, well location distribution, and injection-production working system as the input of the Ctrl-CLSTM model, and predicting the oil saturation prediction value.

[0025] See Figure 10 , a computer device is disclosed in an embodiment of the present invention, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method are implemented.

[0026] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the method are implemented.

[0027] Currently, the major main oilfields have generally entered a development stage with high water cut, low production, and serious ineffective water cycle. The overall utilization degree is relatively low, and it is urgent to change the development mode to improve the recovery rate. The study of oil saturation is an important task in reservoir management in the middle and late stages of oilfield development and an important working link to implement the "water control and oil stabilization" development strategy. Numerous oilfield development practices have shown that the recovery rate of conventional waterflooding reservoirs is only about 30%, and there is still a large amount of remaining oil remaining underground. Therefore, studying the distribution law of oil saturation, clarifying the influencing factors of oil saturation distribution and the existence form of remaining oil is of great significance for oilfields to increase recoverable reserves and improve the recovery rate. To analyze the continuous changes of oil saturation in time and space and summarize the laws therein, it is necessary to consider the influence of control factors such as artificial gas injection and water injection on the change of oil saturation.

[0028] Based on the existing research, the present invention combines the dynamic and static data of reservoir development, applies the convolutional neural network algorithm and the long short-term memory network algorithm, integrates the attention mechanism, and establishes as Figure 3The shown Controllable Spatiotemporal Memory Network (Ctrl-CLSTM) comprehensively considers various influencing factors such as injection-production working parameters, permeability, and well location distribution, and combines historical saturation distribution image data to achieve accurate prediction of the oil saturation distribution in various oilfield systems. The model uses the oil saturation sequence as a prerequisite to capture the influence of spatial information such as oil saturation distribution and well location distribution on oilfield production; introduces an attention mechanism to divide the importance of different reservoir factors, enhance the factors with large attention weights, and weaken the factors with small attention weights, realizing the importance analysis of external factors, and proposing a plan for the decision-making of injection-production working systems. The attention weight extraction model is as shown in Figure 4 as shown.

[0029] The embodiments of the present invention analyze and compare the prediction results of different deep learning networks. The input of each model is the previous historical oil saturation sequence (as shown in Figure 5 -a) and the injection-production working parameters and well location distribution at the corresponding stage. The prediction result of the numerical simulation method (as shown in Figure 5 -b) is used as the comparison standard , and is compared with well-performing spatiotemporal sequence models, including the traditional long short-term memory network FC-LSTM, TrajGRU with an encoding-prediction network structure that can learn sub-network parameters and improve the overall network state, and ST-LSTM with the introduction of spatial memory flow. The experimental results of each model on the test set are as shown in Figure 5 -b. Figure 6-a shows the oil saturation sequences predicted by Ctrl-CLSTM (the present invention), ST-LSTM, Traj-GRU, and FC-LSTM respectively . For each sequence, 5 oil saturation distributions are generated, which are the predicted values on the 360th day, 720th day, 1080th day, 1440th day, and 1800th day respectively. Figure 6-b shows the errors between the prediction results of each model and the true values ( - ).

[0030] Secondly, the embodiments of the present invention calculate the average quantization results of different models on each evaluation index, as shown in Table 1. It can be seen that Ctrl-CLSTM has the best performance on each evaluation index, and the SSIM (structural similarity, used to measure the similarity between two images) value exceeds 98%, and the MSE (mean square error, used to evaluate the pixel-level absolute error between two images) is reduced from 36.61 to 21.37. The lower the value of LPIPS (structural similarity index, an index used to quantify the structural similarity between two images), the more similar the two images are, and Ctrl-CLSTM has the smallest LPIPS value, which indicates that it has excellent ability to generate high-fidelity images. Using the proposed control gate structure, the influence of injection-production working parameters on oil saturation can be more accurately grasped.

[0031] Table 1 Quantification results of each index

[0032] The controllable spatio-temporal memory network model proposed by the present invention has good transfer learning ability. In many deep learning algorithms, an important prerequisite for obtaining good training results is to have sufficient training samples. The richer the sample set, the more feature information the neural network learns. However, in actual situations, this prerequisite cannot be met many times, and there are often problems such as few sample numbers or few label numbers. Transfer learning can apply the knowledge learned from the source domain samples to the target domain, and to a certain extent, solve the problem of insufficient sample numbers. To verify that Ctrl-CLSTM has a certain transfer learning ability, we pre-train on the reservoir A dataset, apply the obtained pre-trained model to reservoir B with insufficient sample numbers for further training, and test the prediction effect of the Ctrl-CLSTM model on reservoir B.

[0033] As Figure 7 shown, it shows the change of oil saturation distribution during the entire production cycle of reservoir A. The model trained on this reservoir is used to predict the oil saturation distribution of reservoir B. Figure 8 is the input oil saturation sequence of reservoir B. In Figure 9 , the first column is the numerical simulation result of reservoir B, the second column is the prediction result of the Ctrl-CLSTM model, and the third column is the visualization effect of the error between the two. It can be seen that Ctrl-CLSTM can well predict the change of oil saturation in reservoir B, and the error in most areas is less than 0.1, indicating that the Ctrl-CLSTM model has excellent transfer learning ability between different reservoirs.

[0034] Table 2 below shows the quantitative performance of the Ctrl-CLSTM model on Reservoir B. It can be seen that after feature migration, although the values of each index have decreased, they still remain at a relatively high prediction level. Among them, the SSIM value exceeds 90%, proving that the original data and the predicted images have a high structural similarity.

[0035] Quantification results of each index in Table 2

[0036] The present invention constructs a network model integrating multi-source data, introduces an attention mechanism to screen the influence degree of external factors on the spatio-temporal distribution of saturation, accurately predicts the distribution law of oil saturation in the reservoir, and dynamically adjusts different production regime modes, making the prediction of saturation distribution more accurate and practical.

[0037] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0038] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0039] These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0040] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions for implementing the steps for realizing the functions specified in one process or a plurality of processes and / or boxes Figure 1 one process or a plurality of processes and / or boxes Figure 1 steps for realizing the functions specified in one box or a plurality of boxes.

[0041] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific implementation manners of the present invention, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.

Claims

1. An oil saturation prediction method for oilfields integrating an attention mechanism, characterized in that It includes the following steps: Collect the characteristics of reservoir influencing factors to obtain a data set; Establish a Ctrl-CLSTM model based on the data set and introduce an attention mechanism; Assign different attention weights to the characteristics of reservoir influencing factors and capture the spatio-temporal characteristics of the reservoir production system; Use the oil saturation, well location distribution, and injection-production working system as the input of the Ctrl-CLSTM model to predict the predicted value of oil saturation.

2. The oil saturation prediction method for oilfields integrating an attention mechanism according to claim 1, wherein The step of collecting the characteristics of reservoir influencing factors to obtain a data set specifically includes: collecting the characteristics of reservoir influencing factors and performing normalization processing, integrating the characteristics of reservoir influencing factors according to the time series, and then dividing them according to the time ratio to obtain a data set.

3. The oil saturation prediction method for oilfields integrating an attention mechanism according to claim 2, wherein The characteristics of reservoir influencing factors include oil saturation, permeability, and porosity.

4. A method for predicting oil saturation in oilfields integrating an attention mechanism according to claim 1, characterized in that The step of establishing a Ctrl-CLSTM model based on the data set and introducing an attention mechanism specifically includes: establishing a Ctrl-CLSTM model according to the data set, introducing an attention mechanism on the basis of Ctrl-CLSTM, and extracting the importance weights of the relative spatial states of various reservoirs on the oilfield production; then feedback the extracted importance degree to the data characteristics, enhance the characteristics with high importance degree, and suppress the characteristics with low importance degree.

5. The oil saturation prediction method for oilfields integrating an attention mechanism according to claim 4, characterized in that, The relative spatial states of various reservoirs include well location distribution, oil saturation, permeability, and porosity.

6. A method for predicting oil saturation in oilfields integrating an attention mechanism, as claimed in claim 1, wherein The step of assigning different attention weights to the characteristics of reservoir influencing factors and capturing the spatio-temporal characteristics of the reservoir production system specifically includes: performing preliminary feature extraction on the sequences in the data set through a convolutional neural network, then dividing the characteristics of reservoir influencing factors according to the importance degree, sorting the weights from large to small, taking a weight as a threshold, summing the first i attention weights until it is greater than the threshold, and dividing the reservoir factor characteristics corresponding to the first i attention weights into main influencing factors, and the rest as non-main influencing factors; finally, apply the obtained attention weight values to the corresponding reservoir factors.

7. A method for predicting oil saturation in oilfields integrating an attention mechanism according to claim 1, characterized in that, The step of predicting the predicted value of oil saturation specifically includes: During the prediction process, update the states of each time memory unit of the front and back time series according to the time memory state transition path equation of Ctrl-CLSTM; Update the states of each spatial memory unit of the front and back time series according to the spatial memory state transition path equation of Ctrl-CLSTM.

8. An oil saturation prediction system for oilfields integrating an attention mechanism, characterized in that, It includes: A feature collection module for collecting the characteristics of reservoir influencing factors to obtain a data set; A modeling module for establishing a Ctrl-CLSTM model based on the data set and introducing an attention mechanism; A weight assignment module for assigning different attention weights to the characteristics of reservoir influencing factors and capturing the spatio-temporal characteristics of the reservoir production system; A prediction module that uses the oil saturation, well location distribution, and injection-production working system as the input of the Ctrl-CLSTM model to predict the predicted value of oil saturation.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method described in any one of claims 1-7.

Images