Seismic reservoir parameter prediction method and device

Abstract

The present invention provides a method and device for predicting seismic reservoir parameters, the method comprising: inputting seismic data of a target formation into a pre-established reservoir parameter prediction model to obtain multiple seismic reservoir parameters of the target formation; the reservoir parameter prediction model is established according to the following process: obtaining seismic work area data of the target formation to form a training data set; building a multi-task learning reservoir parameter prediction model based on an artificial neural network model, the reservoir parameter prediction model comprising: an input layer, a basic network and multiple sub-networks, each sub-network being used to output a reservoir parameter; and using the training data set to train the reservoir parameter prediction model. Putting multiple seismic reservoir parameter prediction tasks together for learning, by sharing and complementing the learned information, can effectively express the complex nonlinear relationship between seismic attributes and reservoir parameters, and improve the accuracy of seismic reservoir parameter prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of geophysical oil and gas exploration, and in particular to a method and device for predicting seismic reservoir parameters. Background Art

[0002] As oil and gas exploration rapidly extends to deep, ultra-deep, unconventional and other fields, the objects of exploration and development are becoming increasingly complex, and reservoir description is becoming more difficult. The state of underground reservoirs is mainly characterized by parameters such as reservoir morphology, lithology, physical properties and oil and gas content. It is very necessary to improve the accuracy of seismic prediction of these parameters, which can provide an important basis for risk target evaluation and exploration and development decision-making. In addition, the exploration level of major basins in China is currently high, and a large amount of exploration and development data has been accumulated. How to comprehensively utilize various information and realize intelligent seismic reservoir prediction is imminent.

[0003] At present, there are two main methods for predicting seismic reservoir parameters, including rock physics model method and traditional machine learning method. The rock physics model method refers to the rock physics theoretical modeling of the actual reservoir, and then the seismic attributes are converted into reservoir parameters using rock physics templates or rock physics inversion technology. This method is based on the theoretical model and has certain scalability. It can partially eliminate the coupling effect of reservoir parameters, but it is limited by the scope of application of the model and cannot adapt to complex reservoir conditions, resulting in low prediction accuracy. The traditional machine learning method refers to the use of algorithms such as neural networks, logistic regression, and support vector machines to establish a mapping relationship between multiple seismic attributes and a single target reservoir parameter. The machine learning algorithm has a strong nonlinear expression ability and can learn the implicit intrinsic relationship of the data. However, the existing methods can only predict a single reservoir parameter, are interfered by non-target reservoir parameter factors, and have low prediction accuracy. Summary of the invention

[0004] An embodiment of the present invention provides a method for predicting seismic reservoir parameters to improve the accuracy of prediction. The method includes:

[0005] Acquiring seismic data of target formations;

[0006] Inputting the seismic data of the target formation into a pre-established reservoir parameter prediction model to obtain multiple seismic reservoir parameters of the target formation;

[0007] The reservoir parameter prediction model is used to predict multiple seismic reservoir parameters of a formation based on seismic data of the formation, and is established according to the following process:

[0008] Obtain seismic work area data of the target stratum to form a training data set;

[0009] A multi-task learning reservoir parameter prediction model is built based on an artificial neural network model, wherein the reservoir parameter prediction model includes: an input layer, a basic network and multiple sub-networks, each sub-network is used to output a reservoir parameter;

[0010] The reservoir parameter prediction model is trained using the training data set to obtain a trained reservoir parameter prediction model.

[0011] In a specific embodiment, the seismic work area data includes:

[0012] Geological reports, core test data and well logging data of target strata within the seismic work area.

[0013] During the specific implementation process, the seismic work area data of the target stratum is obtained to form a training data set, including:

[0014] Extracting multiple formation parameters of the target formation based on the seismic work area data to form a training data set;

[0015] Wherein, in the training data set, each row represents a sample, and each column represents a data feature, including elastic features, geological features, and reservoir parameters;

[0016] Elastic characteristics include: longitudinal wave velocity, shear wave velocity, density, two-way travel time, longitudinal wave impedance, shear wave impedance, longitudinal to shear wave velocity ratio, Poisson's ratio, bulk modulus, shear modulus, Young's modulus, longitudinal wave modulus, Lame constant, and any combination of the above parameters;

[0017] Geological characteristics include: depth, pressure, temperature, geological age, intervals, and sedimentary facies;

[0018] Reservoir parameters include: lithology, reservoir type, fluid type, porosity, gas saturation, oil saturation, mud content, dolomite content, organic carbon content and permeability.

[0019] In a specific embodiment, the base network is connected to the input layer, and the multiple sub-networks are connected to the base network in parallel;

[0020] Wherein, the input layer has a plurality of nodes, each node corresponds to an elastic feature or a geological feature;

[0021] Each subnetwork corresponds to a subtask, and a subtask is used to determine a reservoir parameter.

[0022] In a specific embodiment of the present invention, the reservoir parameter prediction model is trained using the training data set, including:

[0023] Choose a cost function for each subtask;

[0024] The reservoir parameter prediction model is trained by using a stochastic gradient descent algorithm with the goal of minimizing the value of the cost function of each subtask.

[0025] In the specific implementation process, the process of establishing the reservoir parameter prediction model also includes:

[0026] Amplifying the training data set using a rock physics model to form an amplified training data set;

[0027] The reservoir parameter prediction model is trained using the training data set to obtain a trained reservoir parameter prediction model, including:

[0028] The reservoir parameter prediction model is trained using the augmented training data set to obtain a trained reservoir parameter prediction model.

[0029] The method of amplifying the training data set by using a rock physics model to form an amplified training data set includes:

[0030] According to the target stratum, a rock physics model corresponding to the target stratum is determined;

[0031] According to the geological characteristics of the target formation, the simulated reservoir parameters are randomly generated using the Monte Carlo simulation method;

[0032] Inputting simulated reservoir parameters into the rock physics model to obtain simulated elastic characteristics;

[0033] An augmented training data subset is formed according to the simulated elastic characteristics, the geological characteristics of the target formation and the simulated reservoir parameters, and the augmented training data subset is appended to the training data set to obtain the augmented training data set.

[0034] The embodiment of the present invention further provides a seismic reservoir parameter prediction device for improving the accuracy of prediction, the device comprising:

[0035] A data acquisition module, used to acquire seismic data of a target formation;

[0036] A multiple seismic reservoir parameter determination module is used to input the seismic data of the target formation into a pre-established reservoir parameter prediction model to obtain multiple seismic reservoir parameters of the target formation;

[0037] Wherein, the reservoir parameter prediction model is used to predict multiple seismic reservoir parameters of the formation based on the seismic data of the formation;

[0038] The multi-seismic reservoir parameter determination module includes: a model pre-establishment unit, which is used to:

[0039] Obtain seismic work area data of the target stratum to form a training data set;

[0040] A multi-task learning reservoir parameter prediction model is built based on an artificial neural network model, wherein the reservoir parameter prediction model includes: an input layer, a basic network and multiple sub-networks, each sub-network is used to output a reservoir parameter;

[0041] The reservoir parameter prediction model is trained using the training data set to obtain a trained reservoir parameter prediction model.

[0042] An embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned seismic reservoir parameter prediction method when executing the computer program.

[0043] An embodiment of the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for executing the above-mentioned seismic reservoir parameter prediction method.

[0044] In an embodiment of the present invention, seismic data of a target formation is obtained; the seismic data of the target formation is input into a pre-established reservoir parameter prediction model to obtain a plurality of seismic reservoir parameters of the target formation; wherein the reservoir parameter prediction model is used to predict a plurality of seismic reservoir parameters of a formation based on the seismic data of the formation, and is established according to the following process: seismic work area data of the target formation is obtained to form a training data set; a multi-task learning reservoir parameter prediction model is established based on an artificial neural network model, the reservoir parameter prediction model comprising: an input layer, a basic network and a plurality of sub-networks, each sub-network being used to output a reservoir parameter; and the reservoir parameter prediction model is trained using the training data set to obtain a trained reservoir parameter prediction model. By establishing a multi-task learning reservoir parameter prediction model that can simultaneously predict multiple seismic reservoir parameters, multiple seismic reservoir parameter prediction tasks are put together for learning. By sharing and complementing the learned information, the many-to-many complex nonlinear relationship between seismic attributes and reservoir parameters can be effectively expressed. Compared with the traditional machine learning method that can only predict a single reservoir parameter, it not only reduces the interference of non-target reservoir parameter factors and improves the accuracy of seismic reservoir parameter prediction; and compared with the rock physics model method, it is not limited by the scope of model application and can predict complex reservoirs, thereby improving the accuracy of seismic reservoir parameter prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0046] Figure 1 Schematic diagram of a method for predicting seismic reservoir parameters in an embodiment of the present invention.

[0047] Figure 2 Schematic diagram of a method for establishing a reservoir parameter prediction model in an embodiment of the present invention.

[0048] Figure 3 It is a schematic diagram of a specific implementation method of step 203 in a specific embodiment of the present invention.

[0049] Figure 4 It is a schematic diagram of a method for establishing a reservoir parameter prediction model in a specific embodiment of the present invention.

[0050] Figure 5 It is a schematic diagram of a specific implementation method of step 401 in a specific embodiment of the present invention.

[0051] Figure 6 This is a flow chart of the scheme of Example 1 in the specific implementation of the present invention.

[0052] Figure 7 This is a flow chart of the scheme of Example 2 in the specific implementation of the present invention.

[0053] Figure 8 This is a schematic diagram of the training data set amplification results in Example 2 in the specific implementation of the present invention.

[0054] Fig. 9 It is a structural schematic diagram of the multi-task learning reservoir parameter prediction model in the specific implementation of the present invention.

[0055] Fig.10 It is a schematic diagram of the seismic reservoir parameter prediction results obtained in Example 2 in the specific implementation of the present invention.

[0056] Fig.11 Schematic diagram of a seismic reservoir parameter prediction device in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] The embodiment of the present invention provides a method for predicting seismic reservoir parameters to improve the accuracy of prediction. Figure 1 As shown, the method includes:

[0059] Step 101: Acquire seismic data of a target stratum;

[0060] Step 102: input the seismic data of the target formation into a pre-established reservoir parameter prediction model to obtain a plurality of seismic reservoir parameters of the target formation.

[0061] The reservoir parameter prediction model is used to predict multiple seismic reservoir parameters of the formation according to the seismic data of the formation. Figure 2 The process shown establishes:

[0062] Step 201: Acquire seismic data of the target stratum to form a training data set;

[0063] Step 202: Building a multi-task learning reservoir parameter prediction model based on an artificial neural network model, the reservoir parameter prediction model comprising: an input layer, a basic network and multiple sub-networks, each sub-network being used to output a reservoir parameter;

[0064] Step 203: Using the training data set, the reservoir parameter prediction model is trained to obtain a trained reservoir parameter prediction model.

[0065] Depend on Figure 1 and Figure 2 It can be seen from the shown process that in an embodiment of the present invention, the seismic data of the target formation is obtained; the seismic data of the target formation is input into a pre-established reservoir parameter prediction model to obtain multiple seismic reservoir parameters of the target formation; wherein the reservoir parameter prediction model is used to predict multiple seismic reservoir parameters of the formation based on the seismic data of the formation, and is established according to the following process: seismic work area data of the target formation is obtained to form a training data set; a multi-task learning reservoir parameter prediction model is built based on an artificial neural network model, and the reservoir parameter prediction model includes: an input layer, a basic network and multiple sub-networks, each sub-network is used to output a reservoir parameter; the reservoir parameter prediction model is trained using the training data set to obtain a trained reservoir parameter prediction model. By establishing a multi-task learning reservoir parameter prediction model that can simultaneously predict multiple seismic reservoir parameters, multiple seismic reservoir parameter prediction tasks are put together for learning. By sharing and complementing the learned information, the many-to-many complex nonlinear relationship between seismic attributes and reservoir parameters can be effectively expressed. Compared with the traditional machine learning method that can only predict a single reservoir parameter, it not only reduces the interference of non-target reservoir parameter factors and improves the accuracy of seismic reservoir parameter prediction; and compared with the rock physics model method, it is not limited by the scope of model application and can predict complex reservoirs, thereby improving the accuracy of seismic reservoir parameter prediction.

[0066] In the specific implementation, first obtain the seismic work area data of the target formation to form a training data set for pre-establishing a reservoir parameter prediction model. In the specific implementation, obtain the seismic work area data of the target formation, including: geological reports, core test data and logging data of the target formation within the seismic work area. Based on the acquired seismic work area data, extract multiple formation parameters of the target formation to form a training data set; in the training data set, each row represents a sample, and each column represents a data feature, including elastic features, geological features and reservoir parameters.

[0067] In a specific embodiment, multiple reservoir parameters include: lithology, porosity, gas saturation and mud content. Elastic characteristics include: P-wave velocity, S-wave velocity, density, two-way travel time, P-wave impedance, S-wave impedance, P-wave velocity ratio, Poisson's ratio, bulk modulus, shear modulus, Young's modulus, P-wave modulus, Lame constant and other parameters and the combination of any of the above parameters. Geological characteristics include depth, pressure, temperature, geological age, layer section, sedimentary phase, etc.; reservoir parameters include lithology, reservoir type, fluid type, porosity, gas saturation, oil saturation, mud content, dolomite content, organic carbon content and permeability.

[0068] Those skilled in the art will appreciate that the above-mentioned specific parameters are only examples, and the specific parameters related to seismic reservoir parameter prediction all fall within the protection scope of the present invention and will not be described in detail in the embodiments.

[0069] After forming the training data set, a multi-task learning reservoir parameter prediction model is built based on the artificial neural network model. The reservoir parameter prediction model includes: an input layer, a basic network and multiple sub-networks, each of which is used to output a reservoir parameter. In a specific embodiment, the basic network is connected to the input layer, and multiple sub-networks are connected to the basic network in parallel; wherein the above-mentioned input layer has multiple nodes, each node corresponds to an elastic feature or a geological feature; each sub-network corresponds to a sub-task, and a sub-task is used to determine a reservoir parameter.

[0070] After establishing a reservoir parameter prediction model capable of multi-task collaborative learning, the reservoir parameter prediction model is trained using the above training data set. The specific process is as follows: Figure 3 As shown, including:

[0071] Step 301: Select a cost function for each subtask;

[0072] Step 302: Using the stochastic gradient descent algorithm, the reservoir parameter prediction model is trained with the goal of minimizing the cost function of each subtask.

[0073] In the specific implementation, a cost function is selected for each subtask, where the classification subtask uses the cross entropy function and the regression subtask uses the mean square error function. During the training process, the stochastic gradient descent algorithm is used to optimize the subtasks, with the goal of minimizing the cost function value of each subtask. Through multiple iterations and optimizing the model hyperparameters, a trained reservoir parameter prediction model is obtained.

[0074] After obtaining a trained reservoir parameter prediction model, the seismic data of the target stratum is obtained, wherein the seismic data includes elastic characteristics and geological characteristics, including elastic characteristics such as P-wave impedance, P-wave velocity ratio, density, etc. obtained based on seismic inversion, and geological characteristics such as depth and layer interval obtained by seismic structural interpretation.

[0075] After obtaining the seismic data of the target formation, the seismic data of the target formation is input into the above-mentioned pre-established reservoir parameter prediction model to obtain multiple seismic reservoir parameters of the target formation, such as lithology, porosity, gas saturation, mud content and other parameters of the target formation.

[0076] It can be clearly seen from the above process that the seismic reservoir parameter prediction method provided in the embodiment of the present invention has the following beneficial effects:

[0077] Multiple tasks can be completed simultaneously: only one prediction model is needed to simultaneously complete the prediction of multiple reservoir parameters such as lithology, porosity, gas saturation, and shale content, simplifying the workflow and workload of seismic reservoir prediction;

[0078] Reservoir parameter prediction is more accurate: the introduction of geological characteristics increases the constraints of prior information; solving multiple reservoir parameters at the same time can effectively eliminate the coupling between them, thereby improving the prediction accuracy;

[0079] Efficient application deployment: Parallel computing can be used with high computational efficiency, and the prediction of general three-dimensional seismic work areas can be completed within a few minutes, which is suitable for large-scale industrial applications.

[0080] In a specific embodiment, in order to further improve the accuracy of prediction, the accuracy of model prediction is improved by increasing the diversity of target strata. In specific implementation, Figure 4 As shown in Figure 2, the process of establishing a reservoir parameter prediction model is as follows: Figure 2 On the basis of

[0081] Step 401: amplifying the training data set using a rock physics model to form an amplified training data set;

[0082] Accordingly, step 203 is changed to step 402: using the augmented training data set, the reservoir parameter prediction model is trained to obtain a trained reservoir parameter prediction model.

[0083] In the specific implementation process, the specific implementation process of step 401 is as follows: Figure 5 As shown, including:

[0084] Step 501: determining a rock physics model corresponding to the target stratum according to the target stratum;

[0085] Step 502: randomly generating simulated reservoir parameters using a Monte Carlo simulation method according to the geological characteristics of the target formation;

[0086] Step 503: inputting the simulated reservoir parameters into the rock physics model to obtain simulated elastic characteristics;

[0087] Step 504: an augmented training data subset is formed according to the simulated elastic characteristics, the geological characteristics of the target formation and the simulated reservoir parameters, and the augmented training data subset is appended to the training data set to obtain an augmented training data set.

[0088] In a specific embodiment of the present invention, the rock theoretical model is used to expand the training targets, thereby increasing the diversity of the targets and improving the generalization ability of the model.

[0089] A specific example is given below to illustrate how to predict seismic reservoir parameters in an embodiment of the present invention.

[0090] Figure 6 The scheme flow chart of embodiment 1 specifically includes the following steps:

[0091] 1) Collect various data related to the underground target formation, including geological reports, core data, well logging curves, etc., and extract elastic characteristics (P-wave impedance, P-wave velocity ratio, density, etc.), geological characteristics (depth, layer, etc.) and reservoir parameters (lithology, porosity, gas saturation, mud content, etc.) from them to form a training data set;

[0092] 2) Use artificial neural networks to build a multi-task learning reservoir parameter prediction model, whose network structure contains 1 input layer, 1 basic network and multiple sub-networks. The basic network is connected to the input layer, and multiple sub-networks are connected to the basic network in parallel. The input layer has multiple nodes, each of which corresponds to an elastic feature (P-wave impedance, P-wave velocity ratio, density, etc.) or geological feature (depth, layer, etc.). Each sub-network corresponds to a sub-task and outputs a reservoir parameter (lithology, porosity, gas saturation, mud content, etc.);

[0093] 3) Train the multi-task learning reservoir parameter prediction model based on the training data set, and select a cost function for each subtask, where the classification subtask uses the cross entropy function and the regression subtask uses the mean square error function. The stochastic gradient descent algorithm is used for optimization during the training process, and the goal is to minimize the value of the cost function of each subtask. Through multiple iterations and optimization of model hyperparameters, a trained reservoir parameter prediction model is obtained;

[0094] 4) Input the seismic data into the trained reservoir parameter prediction model, and predict multiple seismic reservoir parameters at the same time. The required seismic data include P-wave impedance, P-wave velocity ratio, density, etc. obtained by seismic inversion, and depth, layer, etc. obtained by seismic structural interpretation. The predicted seismic reservoir parameters include lithology, porosity, gas saturation, mud content, etc.

[0095] Figure 7 The scheme flow chart of embodiment 2 specifically includes the following steps:

[0096] 1) Collect various data related to the underground target strata, including geological reports, core data, well logging curves, etc., and extract elastic characteristics (P-wave impedance, P-wave velocity ratio, density, etc.), geological characteristics (depth, layer, etc.) and reservoir parameters (lithology, porosity, gas saturation, mud content, etc.) from them to form a training data set;

[0097] 2) Use the rock physics model to amplify the training data set to form an amplified training data set. For a specific formation, select an applicable rock physics model, then combine the Monte Carlo simulation method to randomly generate simulated data, and append the simulated data to the training data set;

[0098] 3) Use artificial neural networks to build a multi-task learning reservoir parameter prediction model, whose network structure contains 1 input layer, 1 basic network and multiple sub-networks. The basic network is connected to the input layer, and multiple sub-networks are connected to the basic network in parallel. The input layer has multiple nodes, each of which corresponds to an elastic feature (P-wave impedance, P-wave velocity ratio, density, etc.) or geological feature (depth, layer, etc.). Each sub-network corresponds to a sub-task and outputs a reservoir parameter (lithology, porosity, gas saturation, mud content, etc.);

[0099] 4) Train the multi-task learning reservoir parameter prediction model based on the training data set, and select a cost function for each subtask, where the classification subtask uses the cross entropy function and the regression subtask uses the mean square error function. The stochastic gradient descent algorithm is used for optimization during the training process, and the goal is to minimize the value of the cost function of each subtask. Through multiple iterations and optimizing the model hyperparameters, a trained reservoir parameter prediction model is obtained;

[0100] 5) Input the seismic data into the trained reservoir parameter prediction model, and predict multiple seismic reservoir parameters at the same time. The required seismic data include P-wave impedance, P-wave velocity ratio, density, etc. obtained by seismic inversion, and depth, layer, etc. obtained by seismic structural interpretation. The predicted seismic reservoir parameters include lithology, porosity, gas saturation, mud content, etc.

[0101] Taking the Xujiahe Formation tight sandstone gas reservoir in the Sichuan Basin as an example, the methods provided by the above two embodiments are further explained.

[0102] Figure 8 The augmentation effect of the training data set in Example 2 is shown, where the blue dots are logging data and the yellow dots are data simulated using a rock physics model. The number of wells drilled in an actual work area is often limited, and the reservoir lithology, physical properties, and gas content may change dramatically in the lateral direction.

[0103] The well logging data in the figure lacks samples with high gas saturation, which obviously cannot represent the actual underground situation. Combining with rock physics model can overcome this problem and generate some samples that have never been seen in the well, such as high gas saturation. Adding these samples to the training data set increases the diversity of samples and improves the generalization ability of the training model.

[0104] Fig. 9 The above-mentioned multi-task learning reservoir parameter prediction model is shown, and its network structure contains 1 input layer, 1 basic network and 4 sub-networks. The elastic characteristics such as P-wave impedance, P-wave velocity ratio, density, and geological characteristics such as depth and interval are input, and the reservoir parameters such as lithology, porosity, gas saturation, and mud content are output.

[0105] Fig.10 The seismic reservoir parameter prediction results obtained in Example 2 are shown, which are lithology, shale content, porosity, and gas saturation from top to bottom. It can be seen that the intelligent prediction of reservoir parameters such as lithology, physical properties, and oil and gas content is achieved, and the underground complex reservoir is accurately characterized, which can be used for high-precision quantitative prediction of seismic reservoir parameters in the seismic interpretation process.

[0106] The implementation of the above specific application is only an example, and the remaining implementation methods will not be described in detail one by one.

[0107] Based on the same inventive concept, the embodiment of the present invention further provides a seismic reservoir parameter prediction device. Since the principle of the problem solved by the seismic reservoir parameter prediction device is similar to that of the seismic reservoir parameter prediction method, the implementation of the seismic reservoir parameter prediction device can refer to the implementation of the seismic reservoir parameter prediction method, and the repeated parts will not be repeated. The specific structure is as follows: Fig.11 As shown:

[0108] A data acquisition module 1101 is used to acquire seismic data of a target formation;

[0109] The multiple seismic reservoir parameter determination module 1102 is used to input the seismic data of the target formation into a pre-established reservoir parameter prediction model to obtain multiple seismic reservoir parameters of the target formation.

[0110] Wherein, the reservoir parameter prediction model is used to predict multiple seismic reservoir parameters of the formation based on the seismic data of the formation;

[0111] The multi-seismic reservoir parameter determination module 1102 includes: a model pre-establishment unit, which is used to:

[0112] Obtain seismic work area data of the target stratum to form a training data set;

[0113] A multi-task learning reservoir parameter prediction model is built based on an artificial neural network model. The reservoir parameter prediction model includes: an input layer, a basic network and multiple sub-networks, each of which is used to output a reservoir parameter;

[0114] The reservoir parameter prediction model is trained using the training data set to obtain a trained reservoir parameter prediction model.

[0115] In a specific embodiment, the above-mentioned earthquake work area data includes:

[0116] Geological reports, core test data and well logging data of target strata within the seismic work area.

[0117] In a specific embodiment, the model pre-establishment unit is specifically used to:

[0118] Based on the seismic work area data, multiple formation parameters of the target formation are extracted to form a training data set;

[0119] In the training data set, each row represents a sample, and each column represents a data feature, including elastic features, geological features, and reservoir parameters.

[0120] In specific implementation, the base network is connected to the input layer, and multiple sub-networks are then connected to the base network in parallel;

[0121] Among them, the input layer has multiple nodes, each node corresponds to an elastic feature or geological feature;

[0122] Each subnetwork corresponds to a subtask, and a subtask is used to determine a reservoir parameter.

[0123] Furthermore, the model pre-establishment unit is specifically used for:

[0124] Choose a cost function for each subtask;

[0125] The reservoir parameter prediction model is trained using the stochastic gradient descent algorithm with the goal of minimizing the cost function of each subtask.

[0126] In a specific embodiment, the model pre-establishment unit is further used to:

[0127] Amplifying the training data set using a rock physics model to form an augmented training data set;

[0128] Accordingly, the model pre-establishment unit is specifically used for:

[0129] The reservoir parameter prediction model is trained by using the augmented training data set to obtain a trained reservoir parameter prediction model.

[0130] Furthermore, the model pre-establishment unit is specifically used for:

[0131] According to the target stratum, a rock physics model corresponding to the target stratum is determined;

[0132] According to the geological characteristics of the target formation, the simulated reservoir parameters are randomly generated using the Monte Carlo simulation method;

[0133] The simulated reservoir parameters are input into the rock physics model to obtain simulated elastic characteristics;

[0134] An augmented training data subset is formed according to the simulated elastic characteristics, the geological characteristics of the target formation and the simulated reservoir parameters, and the augmented training data subset is appended to the training data set to obtain the augmented training data set.

[0135] An embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned seismic reservoir parameter prediction method when executing the computer program.

[0136] An embodiment of the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for executing the above-mentioned seismic reservoir parameter prediction method.

[0137] In summary, the seismic reservoir parameter prediction method and device provided by the embodiments of the present invention have the following advantages:

[0138] By acquiring seismic data of a target formation; inputting the seismic data of the target formation into a pre-established reservoir parameter prediction model to obtain a plurality of seismic reservoir parameters of the target formation; wherein the reservoir parameter prediction model is used to predict a plurality of seismic reservoir parameters of a formation based on the seismic data of the formation, and is established according to the following process: acquiring seismic work area data of the target formation to form a training data set; building a multi-task learning reservoir parameter prediction model based on an artificial neural network model, the reservoir parameter prediction model comprising: an input layer, a basic network and a plurality of sub-networks, each sub-network being used to output a reservoir parameter; using the training data set to train the reservoir parameter prediction model to obtain a trained reservoir parameter prediction model. By establishing a multi-task learning reservoir parameter prediction model that can simultaneously predict multiple seismic reservoir parameters, multiple seismic reservoir parameter prediction tasks are put together for learning. By sharing and complementing the learned information, the many-to-many complex nonlinear relationship between seismic attributes and reservoir parameters can be effectively expressed. Compared with the traditional machine learning method that can only predict a single reservoir parameter, it not only reduces the interference of non-target reservoir parameter factors and improves the accuracy of seismic reservoir parameter prediction; and compared with the rock physics model method, it is not limited by the scope of model application and can predict complex reservoirs, thereby improving the accuracy of seismic reservoir parameter prediction. The use of rock theoretical models to amplify training samples increases the diversity of samples, thereby improving the generalization ability of the model and further improving the accuracy of seismic reservoir parameter prediction.

[0139] Although the present invention provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or the drawings (for example, in a parallel processor or multi-threaded processing environment).

[0140] It should be understood by those skilled in the art that the embodiments of this specification may be provided as methods, devices (systems) or computer program products. Therefore, the embodiments of this specification may take the form of complete hardware embodiments, complete software embodiments or embodiments combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0141] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0142] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0143] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0144] Each embodiment in this specification is described in a progressive manner, and the same and similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such a process, method, article or device. The orientation or position relationship indicated by the terms "upper", "lower", etc. is based on the orientation or position relationship shown in the drawings, which 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 therefore cannot be understood as a limitation of the present invention. Unless otherwise expressly specified and limited, the terms "installed", "connected" and "connected" should 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 a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other in the absence of conflict. The present invention is not limited to any single aspect, nor to any single embodiment, nor to any combination and / or permutation of these aspects and / or embodiments. Moreover, each aspect and / or embodiment of the present invention can be used alone or in combination with one or more other aspects and / or embodiments thereof.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for predicting seismic reservoir parameters, characterized in that: include: Acquiring seismic data of target formations; Inputting the seismic data of the target formation into a pre-established reservoir parameter prediction model to obtain multiple seismic reservoir parameters of the target formation; The reservoir parameter prediction model is used to predict multiple seismic reservoir parameters of a formation based on seismic data of the formation, and is established according to the following process: Acquire seismic data of the target formation to form a training data set; in the training data set, each row represents a sample, and each column represents a data feature, including elastic features, geological features, and reservoir parameters; A reservoir parameter prediction model based on multi-task learning is built based on an artificial neural network model, and the reservoir parameter prediction model includes: an input layer, a basic network and multiple sub-networks, each sub-network is used to output a reservoir parameter; wherein the input layer has multiple nodes, each node corresponds to an elastic feature or a geological feature; each sub-network corresponds to a sub-task, and a sub-task is used to determine a reservoir parameter; Using the training data set, training the reservoir parameter prediction model to obtain a trained reservoir parameter prediction model; The process of establishing the reservoir parameter prediction model also includes: determining a rock physics model corresponding to the target stratum according to the target stratum; randomly generating simulated reservoir parameters using a Monte Carlo simulation method according to the geological characteristics of the target stratum; inputting the simulated reservoir parameters into the rock physics model to obtain simulated elastic characteristics; forming an augmented training data subset according to the simulated elastic characteristics, the geological characteristics of the target stratum and the simulated reservoir parameters, and appending the augmented training data subset to the training data set to obtain an augmented training data set; Using the training data set to train the reservoir parameter prediction model to obtain a trained reservoir parameter prediction model includes: using the amplified training data set to train the reservoir parameter prediction model to obtain a trained reservoir parameter prediction model.

2. The method for predicting seismic reservoir parameters according to claim 1, characterized in that: The earthquake work area data include: Geological reports, core test data and well logging data of target strata within the seismic work area.

3. The method for predicting seismic reservoir parameters according to claim 2, characterized in that: Obtain the seismic data of the target stratum to form a training data set, including: Extracting multiple formation parameters of the target formation based on the seismic work area data to form a training data set; Elastic characteristics include: longitudinal wave velocity, shear wave velocity, density, two-way travel time, longitudinal wave impedance, shear wave impedance, longitudinal to shear wave velocity ratio, Poisson's ratio, bulk modulus, shear modulus, Young's modulus, longitudinal wave modulus, Lame constant, and any combination of the above parameters; Geological characteristics include: depth, pressure, temperature, geological age, intervals, and sedimentary facies; Reservoir parameters include: lithology, reservoir type, fluid type, porosity, gas saturation, oil saturation, mud content, dolomite content, organic carbon content and permeability.

4. The method for predicting seismic reservoir parameters according to claim 3, characterized in that: The base network is connected to the input layer, and the multiple sub-networks are connected to the base network in parallel.

5. The method for predicting seismic reservoir parameters according to claim 4, characterized in that: Using the training data set, training the reservoir parameter prediction model includes: Choose a cost function for each subtask; The reservoir parameter prediction model is trained by using a stochastic gradient descent algorithm with the goal of minimizing the value of the cost function of each subtask.

6. A seismic reservoir parameter prediction device, characterized in that: include: A data acquisition module, used to acquire seismic data of a target formation; A multiple seismic reservoir parameter determination module is used to input the seismic data of the target formation into a pre-established reservoir parameter prediction model to obtain multiple seismic reservoir parameters of the target formation; Wherein, the reservoir parameter prediction model is used to predict multiple seismic reservoir parameters of the formation based on the seismic data of the formation; The multi-seismic reservoir parameter determination module includes: a model pre-establishment unit, which is used to: Acquire seismic data of the target formation to form a training data set; in the training data set, each row represents a sample, and each column represents a data feature, including elastic features, geological features, and reservoir parameters; A reservoir parameter prediction model based on multi-task learning is built based on an artificial neural network model, and the reservoir parameter prediction model includes: an input layer, a basic network and multiple sub-networks, each sub-network is used to output a reservoir parameter; wherein the input layer has multiple nodes, each node corresponds to an elastic feature or a geological feature; each sub-network corresponds to a sub-task, and a sub-task is used to determine a reservoir parameter; Using the training data set, training the reservoir parameter prediction model to obtain a trained reservoir parameter prediction model; The model pre-establishment unit is further used to: determine the rock physics model corresponding to the target stratum according to the target stratum; randomly generate simulated reservoir parameters using the Monte Carlo simulation method according to the geological characteristics of the target stratum; input the simulated reservoir parameters into the rock physics model to obtain simulated elastic characteristics; form an augmented training data subset according to the simulated elastic characteristics, the geological characteristics of the target stratum and the simulated reservoir parameters, and append the augmented training data subset to the training data set to obtain an augmented training data set; The model pre-establishment unit is specifically used to: use the amplified training data set to train the reservoir parameter prediction model to obtain a trained reservoir parameter prediction model.

7. The seismic reservoir parameter prediction device according to claim 6, characterized in that: The earthquake work area data include: Geological reports, core test data and well logging data of target strata within the seismic work area.

8. The seismic reservoir parameter prediction device according to claim 7, characterized in that: The model pre-establishment unit is specifically used for: Extracting multiple formation parameters of the target formation based on the seismic work area data to form a training data set; Elastic characteristics include: longitudinal wave velocity, shear wave velocity, density, two-way travel time, longitudinal wave impedance, shear wave impedance, longitudinal to shear wave velocity ratio, Poisson's ratio, bulk modulus, shear modulus, Young's modulus, longitudinal wave modulus, Lame constant, and any combination of the above parameters; Geological characteristics include: depth, pressure, temperature, geological age, intervals, and sedimentary facies; Reservoir parameters include: lithology, reservoir type, fluid type, porosity, gas saturation, oil saturation, mud content, dolomite content, organic carbon content and permeability.

9. The seismic reservoir parameter prediction device according to claim 8, characterized in that: The base network is connected to the input layer, and the multiple sub-networks are connected to the base network in parallel.

10. The seismic reservoir parameter prediction device according to claim 9, characterized in that: The model pre-establishment unit is specifically used for: Choose a cost function for each subtask; The reservoir parameter prediction model is trained by using a stochastic gradient descent algorithm with the goal of minimizing the value of the cost function of each subtask.

11. 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, the method according to any one of claims 1 to 5 is implemented.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images