An oil production prediction model establishment method, a productivity prediction method, and a storage medium

Abstract

The present application provides an oil production prediction model establishment method, a production capacity prediction method and a storage medium, comprising: obtaining a training set and a test set after cleaning, wherein the training set and the test set comprise a plurality of groups of historical data, including geological parameters, construction parameters and oil production; inputting the geological parameters and the construction parameters in the training set into an initialized prediction model to obtain a model output value; using a loss function to determine the error term of the model output value and the true value in the test set, the loss function being determined according to a true loss function and a physical loss function; updating the weight coefficients of each node in the current prediction model according to the loss function to obtain a trained prediction model, which realizes the correction of the loss function according to the physical loss function, considers the physical law of the shale oil and gas reservoir, and updates the weight coefficients of each node by the loss function determined by the true loss function and the physical loss function, so that the convergence speed and the prediction accuracy of the model can be improved.

Description

Technical Field

[0001] This invention relates to the field of geological exploration and development technology, and in particular to a method for establishing an oil production prediction model, a method for predicting production capacity, and a storage medium. Background Technology

[0002] Currently, data mining-based methods for predicting production capacity in unconventional oil and gas fields such as shale oil and gas reservoirs have been discussed and researched. In general, researchers follow two main approaches:

[0003] The first method involves fitting univariate nonlinear and multiple linear regressions based on collected historical drilling data, and then using the fitted univariate nonlinear and multiple linear regression methods to predict production capacity.

[0004] The second method is to predict production capacity based on machine learning technology, such as through prediction models. However, in the field of shale oil and gas reservoir prediction, the design of the loss function does not take into account the physical laws of shale oil and gas reservoirs, which leads to problems such as slow model convergence or poor model accuracy. Summary of the Invention

[0005] To address the aforementioned problems in existing technologies, this paper aims to provide a method for establishing an oil production prediction model, a production capacity prediction method, and a storage medium. This addresses the issue that existing technologies fail to consider the physical characteristics of shale oil and gas reservoirs in their loss function design, leading to slow model convergence or poor model accuracy.

[0006] To solve the above-mentioned technical problems, the specific technical solution presented in this paper is as follows:

[0007] On the one hand, this paper provides a method for establishing an oil production prediction model, including:

[0008] Obtain the training set and test set after the cleaning is completed, wherein the training set and test set include several sets of historical data, and the historical data includes geological parameters, construction parameters and corresponding oil production;

[0009] The geological parameters and construction parameters in the training set are input into the initialized prediction model to obtain the model output value;

[0010] The error term between the model output value and the true value in the test set is determined using a loss function, wherein the loss function is determined based on the true loss function and the physical loss function;

[0011] The weight coefficients of each node in the current prediction model are updated according to the loss function to obtain the trained prediction model.

[0012] As an example of this paper, the loss function is determined based on the true loss function and the physical loss function, and further includes:

[0013] The true loss function is determined based on the model output value and the test set.

[0014] The physical loss function is determined based on the model output value and the physical constraints of the target oil well;

[0015] The loss function is determined based on the penalty factor, the number of features, the true loss function, and the physical loss function, wherein the number of features is determined based on the number of groups of historical data in the training set.

[0016] As an embodiment of this paper, determining the physical loss function based on the model output value and the physical constraints of the target oil well further includes:

[0017] According to the physical constraint formula:

[0018]

[0019] Determine the physical constraints Q of the target oil well theory , where n F Let be the number of fractures in the target oil well network, and s be a Laplace space variable. The geological features are determined in Laplace space based on the geological characteristics, which include matrix porosity / permeability, reservoir geometry, crude oil viscosity, crude oil compressibility, production time, bottom hole temperature, initial reservoir pressure, and bottom hole flowing pressure.

[0020] The physical loss function is E th =(Q NN -Q theory ) 2 Q NN The output yield obtained after inputting the test set into the prediction model.

[0021] As an embodiment of this paper, determining the true loss function based on the model output value and the test set further includes:

[0022] The true loss function is E re =(Q NN -Q real ) 2 Q real The actual production of the well network in the test set.

[0023] As an embodiment of this document, determining the loss function based on the penalty factor, the number of features, the true loss function, and the physical loss function further includes:

[0024] The loss function is:

[0025]

[0026] Where n is the number of features and λ is the penalty factor.

[0027] As an embodiment of this paper, after updating the weight coefficients of each node in the current prediction model according to the loss function to obtain the trained prediction model, the process includes:

[0028] Substituting the first penalty factor λ into the loss function, the output output Q is predicted using a prediction model with this loss function. NN ;

[0029] Multiple second penalty factors λ′ are generated using a random perturbation algorithm and incorporated into the loss function. Multiple output quantities Q are then predicted using a prediction model with this loss function. NN ′;

[0030] Determine the output quantity Q NN With output Q NN The size relationship of ′;

[0031] If the output output Q NN Greater than the output output Q NN If ', then the first penalty factor λ will be used as the penalty factor for the current prediction model.

[0032] As an example of this article, the determination of output output Q NN With output Q NN The size relationship of ′ further includes:

[0033] If the output output Q NN Less than the output output Q NN If ′ is obtained, the initial temperature and a random number are acquired, where the random number is 0-1;

[0034] The probability of adoption is determined by the formula. Where T is the initial temperature, and ΔE k For output output Q NN Subtract output quantity Q NN The difference;

[0035] If the adoption probability P k If the value is greater than the random number, then the second penalty factor λ′ will be used as the penalty factor for the current prediction model;

[0036] If the adoption probability P k If the value is less than the random number, then the first penalty factor λ will be used as the penalty factor for the current prediction model.

[0037] As an embodiment of this paper, the step of updating the weight coefficients of each node in the current prediction model according to the loss function to obtain the trained prediction model further includes:

[0038] According to the gradient descent formula

[0039]

[0040] Update the weight coefficients ω of each node. i+1 Where α is the iteration step size coefficient, ω i These are the weighting coefficients before the update.

[0041] On the other hand, this paper also provides a production capacity prediction method, including: a prediction model established using any of the oil production prediction model establishment methods described above, which predicts the oil production of the target oil well based on the geological parameters and construction parameters of the target oil well.

[0042] On the other hand, this document also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the oil production prediction model establishment method described in any one of the claims.

[0043] Using the above technical solution, the method for establishing an oil production prediction model described in this paper obtains a cleaned training set and a test set. The training set and test set include several sets of historical data, including geological parameters, construction parameters, and corresponding oil production. This allows for the acquisition of a large amount of historical oil production data. This historical data includes several sets of parameters, each with a geological parameter, a construction parameter, and their corresponding actual oil production. By inputting the geological parameters and construction parameters from the training set into the initialized prediction model, the model output value is obtained. This allows for the determination of the output value for each pair of geological parameters and their corresponding actual oil production in each training set. The predicted values ​​of the construction parameters in the prediction model are the model output values. An error term between the model output values ​​and the true values ​​in the test set is determined using a loss function. This loss function is determined based on the true loss function and the physical loss function, thus correcting the loss function based on the physical loss function, which takes into account the physical properties of shale oil and gas reservoirs. The trained prediction model is obtained by updating the weight coefficients of each node in the current prediction model according to the loss function. This method, which uses both the true loss function and the physical loss function to jointly determine the loss function and update the weight coefficients of each node, improves the model's convergence speed and accuracy.

[0044] To make the above and other objects, features and advantages of this document more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments or prior art described herein, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this article. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This paper presents an overall system diagram of a method for establishing an oil production prediction model according to an embodiment of the present invention.

[0047] Figure 2 This paper illustrates the steps of establishing an oil production prediction model according to an embodiment of the present invention.

[0048] Figure 3 This diagram illustrates a method for augmenting historical data using a generative adversarial neural network, as described in the embodiments of this paper.

[0049] Figure 4 A schematic diagram of the loss function determination method in the embodiments of this paper is shown;

[0050] Figure 5 A schematic diagram of the penalty factor adjustment method in the embodiments of this paper is shown;

[0051] Figure 6 This document shows a flowchart illustrating the establishment of an oil production prediction model according to an embodiment of the present invention.

[0052] Figure 7 The flowchart of the penalty factor update in the embodiments of this article is shown;

[0053] Figure 8 This paper illustrates an apparatus for establishing an oil production prediction model according to an embodiment of the invention;

[0054] Figure 9 A schematic diagram of a computer device as described in this article is shown.

[0055] Explanation of symbols in the attached drawings:

[0056] 101. Database;

[0057] 102. Computing server;

[0058] 801. Acquisition Unit;

[0059] 802. Input Unit;

[0060] 803, Loss Function Unit;

[0061] 804. Update Unit;

[0062] 902. Computer equipment;

[0063] 904, Processor;

[0064] 906. Memory;

[0065] 908. Drive mechanism;

[0066] 910. Input / Output Module;

[0067] 912. Input devices;

[0068] 914. Output devices;

[0069] 916. Presentation equipment;

[0070] 918. Graphical User Interface;

[0071] 920. Network interface;

[0072] 922. Communication link;

[0073] 924. Communication bus. Detailed Implementation

[0074] The technical solutions in the embodiments described below will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments described herein, and not all of the embodiments. Based on the embodiments described herein, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this document.

[0075] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings herein are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0076] like Figure 1 The diagram shows an overall system diagram of a method for establishing an oil production prediction model, including: a database and a computing server;

[0077] Database 101 is used to store historical data in the history of oil well production. In this paper, the historical data includes several sets of corresponding parameters, including geological parameters, construction parameters and their corresponding oil production, as shown in Figure 1.

[0078] Table 1

[0079] Oil well No. 108 in area A 1,8,9… 211,78，79，89… 1000 tons Oil well No. 109 in area B 1,10,9… 201,79，69，19… 900 tons Oil well No. 110 in area C 2,10,10… 111,98，59，29… 780 tons

[0080] In this paper, the geological parameters consist of Young's modulus, total organic carbon, maximum principal stress, minimum principal stress, total gas content, porosity, Poisson's ratio, fracture pressure, brittleness index, and brittle minerals. To save space, they are quantified in Table 1 using several numbers. For specific geological parameters, those skilled in the art can design according to engineering requirements.

[0081] In this paper, the construction parameters consist of liquid strength, total acid volume, total slickwater volume, total linear adhesive volume, sand strength, average pump pressure, average discharge rate, section length, cluster spacing, and number of clusters. To save space, these parameters are quantified using several numbers in Table 1. Those skilled in the art can design specific construction parameters according to project requirements.

[0082] In this article, Figure 1 The geological and construction parameters are provided as examples, therefore no limit is set on the oil production.

[0083] The computing server 102 is used to run several types of prediction models, such as the oil quantity prediction model in this paper. The initialization of this oil quantity prediction model can be completed by assigning random initial values ​​by the computing server, or it can be completed by manually assigning initial values ​​through publicly available literature. This paper does not impose any restrictions on this method. When running the initialized prediction model, the computing server obtains the training set and test set from the database and performs iterative training.

[0084] The computing server 102 is also used to store the loss function of the prediction model that has been initialized. In this paper, the loss function is determined based on the true loss function and the physical loss function. In particular, the physical loss function is set according to physical constraints.

[0085] For example, if there are six sets of historical data in the training set of this paper, after training on these six sets of historical data, the computing server can update the weight coefficients of each node in the prediction model according to the training results. After completing the predetermined number of training iterations, the trained prediction model can be obtained.

[0086] Preferably, after training the training set once, the loss function can be updated once based on the prediction accuracy of the prediction model, as will be described in detail below.

[0087] Currently, data mining-based methods for predicting production capacity in unconventional oil and gas fields such as shale oil and gas reservoirs have been discussed and researched. In general, researchers follow two main approaches:

[0088] One approach is to predict production capacity using machine learning techniques, such as through predictive models. However, in the field of shale oil and gas reservoir prediction, researchers have failed to consider the physical laws of shale oil and gas reservoirs when designing loss functions, leading to problems such as slow model convergence or poor model accuracy.

[0089] To address the aforementioned issues, this paper provides a method for establishing an oil production prediction model that takes into account the physical characteristics of shale oil and gas reservoirs, thereby improving the convergence speed and prediction accuracy of the model. Figure 2 This is a schematic diagram illustrating the steps of an oil production prediction model establishment method provided in this embodiment. This specification provides the operational steps of the method described in the embodiments or flowcharts, but based on conventional or non-inventive labor, more or fewer operational steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or device products, the methods shown in the embodiments or accompanying drawings can be executed sequentially or in parallel. Specifically, as shown... Figure 2 As shown, the method may include:

[0090] Step 201: Obtain the training set and test set after the cleaning is completed. The training set and test set include several sets of historical data, including geological parameters, construction parameters and corresponding oil production.

[0091] Step 202: Input the geological parameters and construction parameters from the training set into the initialized prediction model to obtain the model output value.

[0092] Step 203: Use a loss function to determine the error term between the model output value and the true value in the test set, wherein the loss function is determined based on the true loss function and the physical loss function.

[0093] Step 204: Update the weight coefficients of each node in the current prediction model according to the loss function to obtain the trained prediction model.

[0094] By acquiring cleaned training and test sets, which include several sets of historical data including geological parameters, construction parameters, and corresponding oil production, a large amount of historical oil production data is obtained. This historical data includes several sets of parameters, each with a geological parameter, a construction parameter, and their corresponding actual oil production. By inputting the geological and construction parameters from the training set into the initialized prediction model, the model output value is obtained. This allows for obtaining the predicted value (model output value) for each pair of geological and construction parameters in the training set for each pair. An error term between the model output value and the true value in the test set is determined using a loss function, which is based on a true loss function and a physical loss function. This achieves loss function correction based on the physical loss function, which considers the physical laws of shale oil and gas reservoirs. By updating the weight coefficients of each node in the current prediction model according to the loss function, the trained prediction model is obtained. This allows for updating the weight coefficients of each node by jointly determining the loss function using the true and physical loss functions, improving the model's convergence speed and accuracy.

[0095] As an embodiment of this paper, before step 201, obtaining the training set and test set after the cleaning is completed, wherein the training set and test set include several sets of historical data, the historical data including geological parameters, construction parameters and corresponding oil production, includes:

[0096] Clean the historical data in the database, for example, by using data alignment or removing NULL values.

[0097] As an embodiment of this paper, step 202, inputting both the geological parameters and the construction parameters from the training set into the initialized prediction model to obtain the model output value, further includes:

[0098] The model output value can also be calculated based on the forward computation formula (activation function) of the ANN neural network, where the activation function includes SELU, tanh, Sigmoid and linear.

[0099] As an example of this article, such as Figure 3 The diagram shows a method for augmenting historical data using a generative adversarial neural network.

[0100] Step 301: Initialize the parameters of the generator G(z) and the discriminator D(x), fix the parameters of the generator G(z), and train the discriminator D(x) in a loop so that the discriminator D(x) can correctly distinguish between the real dataset and the fake dataset.

[0101] Initialize a generator G(z): In this paper, the generator model can be a CNN, RNN, LSTM, etc. Its role is to capture the distribution characteristics of the original historical dataset and introduce random noise to generate new fake data. The selected parameter fields can be all the parameters of the original sample dataset or a combination of some parameters, including reservoir thickness, formation pressure, bottom hole pressure, production dynamic curve, etc.

[0102] Initialize a discriminator model D(x): In this paper, the discriminator can be a model such as CNN, RNN, LSTM, etc. Its role is to determine the probability that the fake historical data generated by the generator belongs to the original sample real historical dataset.

[0103] The objective function of the generator is:

[0104]

[0105] The objective function of the discriminator is:

[0106]

[0107] Where D(x) is the output of the discriminator, a real number in the range of 0 to 1, representing the probability that the discriminated data is real data, p data p represents the distribution of real data. z The distribution of the generated data is represented by E, where E represents the expected value of the distribution specified in the subscript. The goal of the discriminator is to maximize the objective function V(D,G) when both real and fake data are input simultaneously; the goal of the generator is to minimize the objective function V(D,G) of the discriminator.

[0108] Step 302: Update the generator and train the generator and discriminator simultaneously until a Nash equilibrium is formed.

[0109] In this paper, the generator is updated so that the discriminator D(x) has a prediction accuracy of 50%, which means it is difficult to distinguish whether the dataset comes from real samples or is fabricated.

[0110] Specifically, the discriminator gradient update is as follows:

[0111]

[0112] The generator gradient update is as follows:

[0113]

[0114] Where m represents the number of samples, Z represents the noise or random samples input to the network G, and θ d θ gThese are the target update parameters for the generator and discriminator, respectively (representing the magnitude of the influence of features on the predicted values).

[0115] like Figure 4 A schematic diagram of the loss function determination method is shown in this embodiment. The loss function is determined based on the true loss function and the physical loss function, and further includes:

[0116] Step 401: Determine the true loss function based on the model output value and the test set.

[0117] In this step, the true loss function is E. re =(Q NN -Q real ) 2 Q real The actual production of the well network in the test set.

[0118] Q NN To obtain the output output after inputting the test set into the prediction model, the difference E is obtained by subtracting the actual output from the output output. re and the difference E re It is used as a loss function to supplement subsequent predictions, thereby improving the prediction accuracy of the prediction model.

[0119] Step 402: Determine the physical loss function based on the model output value and the physical constraints of the target oil well.

[0120] In this step, according to the physical constraint formula:

[0121]

[0122] Determine the physical constraints Q of the target oil well theory , where n F Let be the number of fractures in the target oil well network, and s be a Laplace space variable. The geological features are determined in Laplace space based on the geological characteristics, which include matrix porosity / permeability, reservoir geometry, crude oil viscosity, crude oil compressibility, production time, bottom hole temperature, initial reservoir pressure, and bottom hole flowing pressure.

[0123] The physical loss function is E th =(Q NN -Q theory ) 2 Q NN The output yield obtained after inputting the test set into the prediction model.

[0124] That is, by taking into account the physical conditions of the area where the target oil well is located, the theoretical production rate Q is obtained. theoryIn this way, artificially defined outputs can be introduced, thereby adjusting the loss function to accelerate the convergence speed of the prediction model.

[0125] Step 403: Determine the loss function based on the penalty factor, the number of features, the true loss function, and the physical loss function, wherein the number of features is determined based on the number of historical data sets in the training set.

[0126] In this step, the loss function is:

[0127]

[0128] Where n is the number of features and λ is the penalty factor.

[0129] In this paper, the number of features is determined based on the number of historical data sets in the training set. That is, if there are six historical data sets in the training set, the number of features is six; if there are seven historical data sets, the number of features is seven. In this paper, the number of features corresponds to the number of historical data sets in the training set.

[0130] In this paper, the loss function is related to the penalty factor λ. Therefore, in order to improve the performance of the loss function and enhance the generalization performance and robustness of the predicted data, this paper presents a method for adjusting the penalty factor, which can prevent the loss function from falling into a local optimum and failing to find the global optimum.

[0131] As an example of this article, such as Figure 5 The diagram shown illustrates the penalty factor adjustment method, including:

[0132] Step 501: Substitute the first penalty factor λ into the loss function, and predict the output Q using the prediction model with this loss function. NN .

[0133] In this step, the current penalty factor λ can be substituted into the loss function to obtain the corresponding Q. NN .

[0134] Step 502: Generate multiple second penalty factors λ′ using a random perturbation algorithm and input them into the loss function. Then, predict multiple output quantities Q using a prediction model with this loss function. NN ′.

[0135] In this step, the second penalty factor λ′ is generated based on the first penalty factor λ and a random number generated by the random perturbation algorithm. Specifically, the random perturbation algorithm generates a random number and then adds the random number to the first penalty factor λ to obtain the second penalty factor λ′.

[0136] To better adjust the loss function, second penalty factors λ′ can be generated in batches. Each second penalty factor λ′ is then imported into the loss function in batches, resulting in several adjusted loss functions. After training the prediction model with the adjusted loss functions, a trained prediction model corresponding to each second penalty factor λ′ is obtained. At this point, the same set of historical data is imported into different trained prediction models, yielding multiple output values ​​Q. NN ′.

[0137] Step 503: Determine the output output Q NN With output Q NN The size relationship of ′.

[0138] In this step, since the prediction model will inevitably produce errors, and previous calculations have shown that these errors are likely to be smaller than the actual output value, the output output Q is compared. NN With output Q NN The relationship between the sizes of the two values, i.e., which one is the smallest, proves that the deviation from the actual output value is the smallest.

[0139] Step 504, if the output output Q NN Greater than the output output Q NN If ', then the first penalty factor λ will be used as the penalty factor for the current prediction model.

[0140] In this step, because the output output Q NN Greater than the output output Q NN If '', then it proves that the output output Q is... NN The deviation from the actual output is minimal, therefore the output output Q will be [value]. NN The corresponding first penalty factor λ is used as the penalty factor for the current prediction model. This ensures that the prediction model with this loss function has a strong fitting effect.

[0141] Specifically, when the output output Q corresponds to the randomly generated second penalty factor λ′ NN The fitting effect of ' is better than the output Q corresponding to the first penalty factor λ. NN This proves that the loss function with the first penalty factor λ is not globally optimal. When the current loss function is not globally optimal, it is necessary to find the penalty factor corresponding to the globally optimal loss function.

[0142] As an example of this article, step 503, which involves determining the output output Q, NN With output Q NN The size relationship of ′ further includes:

[0143] If the output output Q NN Less than the output output Q NNIf ′ is obtained, the initial temperature and a random number are acquired, where the random number is 0-1;

[0144] The probability of adoption is determined by the formula. Where T is the initial temperature, and ΔE k For output output Q NN Subtract output quantity Q NN The difference;

[0145] If the adoption probability P k If the value is greater than the random number, then the second penalty factor λ′ will be used as the penalty factor for the current prediction model;

[0146] If the adoption probability P k If the value is less than the random number, then the first penalty factor λ will be used as the penalty factor for the current prediction model.

[0147] To avoid the loss function gradually increasing or decreasing and becoming stuck in a local optimum, we determine whether a random number falls within the adoption probability obtained from the initial temperature. In this paper, the adoption probability can be 0.5. That is, by using two non-fixed numbers, the loss function can jump around a wide range to escape the local optimum.

[0148] As an embodiment of this paper, the step of updating the weight coefficients of each node in the current prediction model according to the loss function to obtain the trained prediction model further includes:

[0149] According to the gradient descent formula

[0150]

[0151] Update the weight coefficients ω of each node. i+1 Where α is the iteration step size coefficient, ω i These are the weighting coefficients before the update.

[0152] In this step, to improve the fitting effect of the prediction model, the weight coefficients in the prediction model can be updated using a loss function.

[0153] In this paper, the weights can also be updated based on gradient-based optimization algorithms, including but not limited to SGD, RMSprop, AdaGrad, Adam, and Nadam.

[0154] To enable those skilled in the art to better understand the training process of the prediction model presented in this paper, this paper provides the following... Figure 6 The flowchart shown illustrates the establishment of an oil production prediction model.

[0155] Step 601: Obtain the cleaned training and test sets. In this step, data cleaning can be performed through data completion, data alignment, and removal of UNLL values.

[0156] Step 602: Import the training set into the initialized prediction model. In this step, you can import historical data from one, two, or three training sets into the prediction model at a time.

[0157] Step 603: Determine whether all historical data in the training set has been imported into the prediction model. If yes, proceed to step 604; otherwise, return to step 602. In this step, importing all historical data from the training set completes one training iteration.

[0158] Step 604: Update the weight coefficients of each node in the prediction model according to the current loss function. In this step, the weight coefficients of each node in the hidden layer of the prediction model can be updated.

[0159] Step 605: Determine whether to update the penalty factor in the loss function. If yes, proceed to step 606; otherwise, the prediction model training is complete. This step can improve the iterative effect of the loss function.

[0160] Step 606: Update the penalty factor; the prediction model training is complete. In this step, as... Figure 7 The flowchart shown is for updating the penalty factor. Through this... Figure 7 Complete the penalty factor update. Figure 7 include:

[0161] Step 701: Set the initial temperature.

[0162] Step 702: Randomly generate a penalty factor λ and input it into the prediction model to obtain Q. NN .

[0163] Step 703: The perturbation generates a new penalty factor λ′, which is then substituted into the prediction model to obtain Q′. NN .

[0164] Step 704, Determine Q NN With Q′ NN The size relationship, when Q NN Greater than Q′ NN Then proceed to step 705, when Q NN Less than Q′ NN Then proceed to step 706.

[0165] Step 705: Receive the new penalty factor λ′.

[0166] Step 706: Perform calculations according to the specific embodiment of step 503.

[0167] Step 707: Determine whether the number of iterations has reached the iteration threshold. If it has, proceed to step 708. If it has not, return to step 703.

[0168] Step 708: Determine whether a new loss function has been obtained. If so, end the process. If not, adjust the initial temperature and return to step 701.

[0169] like Figure 8 The apparatus shown is for establishing an oil production prediction model, comprising:

[0170] The acquisition unit 801 is used to acquire the training set and test set after the cleaning is completed. The training set and test set include several sets of historical data, including geological parameters, construction parameters and corresponding oil production.

[0171] Input unit 802 is used to input the geological parameters and construction parameters in the training set into the initialized prediction model to obtain the model output value;

[0172] Loss function unit 803 is used to determine the error term between the model output value and the true value in the test set using a loss function, wherein the loss function is determined based on the true loss function and the physical loss function;

[0173] The update unit 804 is used to update the weight coefficients of each node in the current prediction model according to the loss function, so as to obtain the prediction model after training.

[0174] The acquisition unit acquires a large amount of historical oil production data, including several sets of parameters, each with a geological parameter, a construction parameter, and their corresponding actual oil production. The input unit obtains the predicted values ​​(i.e., model output values) for each pair of geological and construction parameters in the training set. The loss function unit corrects the loss function based on a physical loss function that considers the physical properties of shale oil and gas reservoirs. The update unit updates the weight coefficients of each node by jointly determining the loss function using both the true and physical loss functions, thereby improving the model's convergence speed and accuracy.

[0175] This paper also provides a production capacity prediction method, including: a prediction model established using any of the oil production prediction model establishment methods described above, which predicts the oil production of the target oil well based on the geological parameters and construction parameters of the target oil well.

[0176] like Figure 9As shown in the embodiments of this document, a computer device 902 provides a method for establishing an oil production prediction model as described herein. The computer device 902 may include one or more processors 904, such as one or more central processing units (CPUs), each of which may implement one or more hardware threads. The computer device 902 may also include any memory 906 for storing information of any kind, such as code, settings, data, etc. Without limitation, for example, the memory 906 may include any type of RAM, any type of ROM, flash memory, hard disk, optical disk, etc. More generally, any memory can use any technology to store information. Furthermore, any memory may provide volatile or non-volatile retention of information. Furthermore, any memory may represent a fixed or removable component of the computer device 902. In one case, when the processor 904 executes associated instructions stored in any memory or combination of memories, the computer device 902 may perform any operation of the associated instructions. The computer device 902 also includes one or more drive mechanisms 908 for interacting with any memory, such as hard disk drive mechanisms, optical disk drive mechanisms, etc.

[0177] Computer device 902 may also include an input / output module 910 (I / O) for receiving various inputs (via input device 912) and providing various outputs (via output device 914). A specific output mechanism may include a presentation device 916 and an associated graphical user interface (GUI) 918. In other embodiments, the input / output module 910 (I / O), input device 912, and output device 914 may be omitted, and the device may function solely as a computer device within a network. Computer device 902 may also include one or more network interfaces 920 for exchanging data with other devices via one or more communication links 922. One or more communication buses 924 couple the components described above together.

[0178] Communication link 922 can be implemented in any way, such as via a local area network (LAN), a wide area network (WAN) (e.g., the Internet), a point-to-point connection, or any combination thereof. Communication link 922 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc., governed by any protocol or combination of protocols.

[0179] Corresponding to Figures 2-7 In addition to the methods described above, this embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described methods.

[0180] This embodiment also provides a computer-readable instruction, wherein when a processor executes the instruction, the program therein causes the processor to perform the following: Figures 2-7 The method shown.

[0181] It should be understood that in the various embodiments of this document, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this document.

[0182] It should also be understood that, in the embodiments herein, the term "and / or" is merely a description of the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following associated objects have an "or" relationship.

[0183] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this document.

[0184] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0185] In the embodiments provided herein, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or units, or they may be electrical, mechanical, or other forms of connection.

[0186] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments described herein, depending on actual needs.

[0187] Furthermore, the functional units in the various embodiments of this document can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0188] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this paper, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this paper. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0189] This document uses specific embodiments to illustrate the principles and implementation methods of this document. The descriptions of the embodiments above are only for the purpose of helping to understand the methods and core ideas of this document. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this document. Therefore, the content of this specification should not be construed as a limitation of this document.

Claims

1. A method for establishing an oil production prediction model, characterized in that, include: Obtain the training set and test set after the cleaning is completed, wherein the training set and test set include several sets of historical data, and the historical data includes geological parameters, construction parameters and corresponding oil production; The geological parameters and construction parameters in the training set are input into the initialized prediction model to obtain the model output value; The error term between the model output value and the true value in the test set is determined using a loss function, wherein the loss function is determined based on the true loss function and the physical loss function; The weight coefficients of each node in the current prediction model are updated according to the loss function to obtain the prediction model after training. The loss function is determined based on the true loss function and the physical loss function, and further includes: The true loss function is determined based on the model output value and the test set. The physical loss function is determined based on the model output value and the physical constraints of the target oil well; The loss function is determined based on the penalty factor, the number of features, the true loss function, and the physical loss function, wherein the number of features is determined based on the number of groups of historical data in the training set; The step of determining the physical loss function based on the model output value and the physical constraints of the target oil well further includes: According to the physical constraint formula: The theoretical production rate formed by determining the physical constraints of the target oil well ,in The number of fractures in the target oil well network. For Laplace space variables, The geological features are determined in Laplace space based on the geological characteristics, which include matrix porosity / permeability, reservoir geometry, crude oil viscosity, crude oil compressibility, production time, bottom hole temperature, initial reservoir pressure, and bottom hole flowing pressure. The physical loss function is: ,in The output yield obtained after inputting the test set into the prediction model.

2. The method for establishing an oil production prediction model according to claim 1, characterized in that, Determining the true loss function based on the model output value and the test set further includes: The true loss function is: ,in The actual production of the well network in the test set.

3. The method for establishing an oil production prediction model according to claim 2, characterized in that, The step of determining the loss function based on the penalty factor, the number of features, the true loss function, and the physical loss function further includes: The loss function is: in, For the number of features, This is a penalty factor.

4. The method for establishing an oil production prediction model according to claim 3, characterized in that, After updating the weight coefficients of each node in the current prediction model according to the loss function to obtain the trained prediction model, the process includes: The first penalty factor Substituting the values ​​into the loss function, the output output is predicted using a prediction model with that loss function. ; Multiple second penalty factors are generated through a random perturbation algorithm. The values ​​are then substituted into the loss function, and multiple output quantities are predicted using a prediction model with that loss function. Determine output output With output Size relationship; If output Greater than the output output Then the first penalty factor will be... As a penalty factor in the current prediction model.

5. The method for establishing an oil production prediction model according to claim 4, characterized in that, The judgment output output With output The size relationship further includes: If output Less than the output output Then, the initial temperature and a random number are obtained, where the random number is 0-1; The probability of adoption is determined by the formula. Where T is the initial temperature. For output output Subtract output The difference; If the adoption probability If the value is greater than the random number, then the second penalty factor will be applied. As a penalty factor for the current prediction model; If the adoption probability If the number is less than the specified random number, then the first penalty factor will be applied. As a penalty factor in the current prediction model.

6. The method for establishing an oil production prediction model according to claim 1, characterized in that, The step of updating the weight coefficients of each node in the current prediction model according to the loss function to obtain the trained prediction model further includes: According to the gradient descent formula Update the weight coefficients of each node ,in This is the iteration step size coefficient. These are the weighting coefficients before the update.

7. A capacity forecasting method, characterized in that, The method uses a prediction model established by the oil production prediction model establishment method as described in any one of claims 1-6 to predict the oil production of the target oil well based on the geological parameters and construction parameters of the target oil well.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the oil production prediction model establishment method according to any one of claims 1-6.

Images