A method and device for predicting production of water-driven oil reservoirs

Abstract

The present invention discloses a method and device for predicting the production of a water-driven oil reservoir. The method comprises: inputting the production-related parameters of the wells to be predicted in the water-driven oil reservoir and the historical time series data of the production and the production-related parameters at the time to be predicted into the production prediction model to obtain the predicted production; the model is obtained by iteratively training the long short-term memory network using the production-related parameters of multiple production wells and the historical time series data of the production, the network comprises an input layer, a first hidden layer, a second hidden layer and an output layer connected in sequence, the first hidden layer comprises a convolution module, a residual module and a maximum pooling layer, which are used for feature extraction and fusion of multi-dimensional time series data, and the second hidden layer is used to capture the time series dependency of the fused time series feature data. The method takes into account the dynamic change characteristics of the production with the development time, combines the residual module to extract the time series features, and combines the long-term memory capacity of the network to establish a water-driven oil reservoir production prediction model to achieve fast and high-precision prediction of the production.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and oil and gas field development, and in particular to a method and device for predicting the production of a water-driven oil reservoir. Background Art

[0002] Since most of the reservoirs in China's oil fields are continental clastic rock deposits with weak natural energy, water drive development is one of the main development methods of oil fields. With the continuous deepening of water drive development, many oil fields have entered the stage of high recovery and high water content. In the high water content period, the vertical and horizontal heterogeneity of the reservoir is serious, the contradiction between layers is aggravated, and the state of the remaining oil occurrence changes. The oil phase changes from a continuous phase to a dispersed phase, the flow capacity of the oil phase decreases significantly, and the water-oil ratio increases rapidly. Therefore, accurately realizing the production prediction of water drive oil reservoirs is of great significance for the continuous potential of old water drive oil fields.

[0003] Production prediction mainly includes traditional methods such as water drive characteristic curve, Logistics, Gompertz, Verhulst and Tong's chart, as well as artificial intelligence methods such as support vector machine, ARIMA, vector autoregression and neural network (LSTM). Summary of the invention

[0004] In order to enrich product types, process routes, and increase selection space, an embodiment of the present invention provides a method and device for predicting water-drive oil reservoir production, which improves BiLSTM based on residual learning to achieve rapid and reasonable prediction of water-drive oil reservoir production.

[0005] In a first aspect, an embodiment of the present invention provides a method for predicting the production of a water-driven oil reservoir, comprising:

[0006] Inputting the production-related parameters of the well to be predicted in the water drive reservoir, the historical time series data of the production, and the production-related parameters at the time to be predicted into the production prediction model, and obtaining the production of the well to be predicted at the time to be predicted according to the output result of the model;

[0007] The production prediction model is obtained by iteratively training a long short-term memory network using production-related parameters of multiple production wells and historical time series data of production. The long short-term memory network includes an input layer, a first hidden layer, a second hidden layer and an output layer connected in sequence. The first hidden layer includes a convolution module, a residual module and a maximum pooling layer, which are used to extract and fuse features of multi-dimensional time series data and output fused time series feature data. The second hidden layer is used to capture the time dependency of the fused time series feature data.

[0008] In a second aspect, an embodiment of the present invention provides a water drive oil reservoir production prediction device, comprising:

[0009] A production prediction model establishment module is used to iteratively train a long short-term memory network using production-related parameters of multiple production wells and historical production time series data to obtain a production prediction model, wherein the long short-term memory network includes an input layer, a first hidden layer, a second hidden layer, and an output layer connected in sequence, wherein the first hidden layer includes a convolution module, a residual module, and a maximum pooling layer, and is used to extract and fuse features of multi-dimensional time series data and output fused time series feature data, and the second hidden layer is used to capture the time series dependency of the fused time series feature data;

[0010] The production prediction module is used to input the production-related parameters of the wells to be predicted in the water-driven oil reservoir, the historical time series data of the production, and the production-related parameters at the time to be predicted into the production prediction model, and obtain the production of the wells to be predicted at the time to be predicted according to the output results of the model.

[0011] In a third aspect, an embodiment of the present invention provides a computer storage medium, wherein the computer storage medium stores computer executable instructions, and when the computer executable instructions are executed by a processor, the above-mentioned water drive reservoir production prediction method is implemented.

[0012] In a fourth aspect, an embodiment of the present disclosure provides a server, 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 water-drive oil reservoir production prediction method when executing the program.

[0013] The beneficial effects of the above technical solution provided by the embodiment of the present invention include at least:

[0014] Traditional LSTM can only perform autocorrelation learning on a single time series variable. The water-drive oil reservoir production prediction method provided in the embodiment of the present invention utilizes a BiLSTM network to extract and fuse features of multi-dimensional time series data by embedding a convolution module CNN and a residual module ResNet. It not only has the multi-dimensional feature extraction and fusion capabilities that the LSTM neural network lacks, but also effectively alleviates problems such as network performance degradation and gradient diffusion. It can retain long-term dependencies and important features of current input data when processing sequence data, so that it has better performance and prediction accuracy in sequence data prediction tasks such as oil well production prediction. It can cope with the challenge of large temporal variations in production data of oil fields in the high water content stage, and plays an important role in the adjustment strategy of oil field production and development.

[0015] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0016] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0018] Figure 1 Schematic diagram of the structure of the BiLSTM network in the first embodiment of the present invention;

[0019] Figure 2 Schematic diagram of the structure of the residual module in the first embodiment of the present invention;

[0020] Figure 3 Schematic diagram of the structure of the BiLSTM network layer in the first embodiment of the present invention;

[0021] Figure 4 This is a cell unit structure diagram of BiLSTM in Example 1 of the present invention;

[0022] Figure 5 Schematic diagram of the structure of a water drive oil reservoir production prediction device in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0024] It should be understood that the terms described in the present invention are only for describing special embodiments and are not intended to limit the present invention. In addition, for the numerical range in the present invention, it should be understood that each intermediate value between the upper and lower limits of the scope is also specifically disclosed. Each smaller range between the intermediate value in any stated value or stated range and any other stated value or intermediate value in the described range is also included in the present invention. The upper and lower limits of these smaller ranges can be independently included or excluded in the scope.

[0025] Unless otherwise indicated, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art to which the invention belongs. Although the present invention describes only preferred methods and materials, any methods and materials similar or equivalent to those described herein may also be used in the implementation or testing of the present invention. All documents mentioned in this specification are incorporated by reference to disclose and describe the methods and / or materials related to the documents. In the event of a conflict with any incorporated document, the content of this specification shall prevail.

[0026] The inventor found in his work that using traditional mathematical models to solve production has certain limitations in terms of model assumptions. Specifically, when using the traditional empirical model method to predict the high water-cut period production of water-driven oil fields, the water-driven characteristic curve is upturned, and the prediction result has a large error, which is not suitable for describing the law of production decline during the high water-cut period. Although machine learning and deep learning models have strong nonlinear capabilities and are suitable for the construction of prediction models, most of the application research in the oil and gas field directly applies general models to production prediction problems, resulting in insufficient ability to extract data features, affecting the accuracy of production prediction.

[0027] In order to solve the problem of low production prediction accuracy in the prior art, an embodiment of the present invention provides a method and device for predicting water-drive oil reservoir production, which predicts water-drive oil reservoir production by improving BiLSTM based on residual learning, thereby improving prediction accuracy.

[0028] Embodiment 1

[0029] Embodiment 1 of the present invention provides a method for establishing a water-drive oil reservoir production prediction model, by iteratively training a long short-term memory network using production-related parameters of multiple production wells and historical time series data of production, to obtain a water-drive oil reservoir production prediction model.

[0030] The long short-term memory network BiLSTM here embeds hidden layers including convolution modules, residual modules and maximum pooling layers on the basis of the traditional network structure.

[0031] See also Figure 1 As shown, the final long short-term memory network BiLSTM includes an input layer, a first hidden layer, a second hidden layer and an output layer connected in sequence. The first hidden layer includes a convolution module, a residual module and a maximum pooling layer, which are used to extract and fuse features of multi-dimensional time series data and output fused time series feature data. The second hidden layer is used to capture the temporal dependency of the fused time series feature data.

[0032] This model applies the convolutional module CNN and the residual module ResNet to the multi-dimensional time series data feature extraction and fusion part, and combines BiLSTM to capture the time series dependency of reservoir production data. The input layer receives the relevant data of the oil well, the hidden layer extracts the multi-dimensional features and time series features of the production prediction data set, and the output layer outputs the production prediction results of the oil well.

[0033] Furthermore, the first hidden layer of the hidden layer includes:

[0034] (1) Convolutional module CNN.

[0035] The convolution module is used to extract feature data from multi-dimensional time series data through convolution calculation, and obtain one-dimensional time series feature data through dimensionality reduction and downsampling processing, so as to realize the extraction of potential features in the multi-factor production prediction data set (production-related parameters of multiple production wells and historical time series data of production).

[0036] (2) Residual module ResNet.

[0037] The residual module stacks convolutional layers and adds skip connections to prevent the gradient vanishing and exploding problems caused by the increase in the number of layers. Figure 2 shown.

[0038] The residual module is used to extract features from one-dimensional time series feature data by stacking convolutional layers, and fuse the extracted features with the original multi-dimensional time series data through jump connections to obtain and output the fused time series feature data.

[0039] Furthermore, the residual module includes forward pass and backward pass:

[0040] The forward pass is used to pass the input through stacked convolutional layers, batch normalization layers, and activation functions to obtain a feature representation, which is then added to the original input through a skip connection to form the final output;

[0041] Backward transfer is used to propagate the gradient through each layer of the residual block during the back propagation process, and also directly propagates it to the input through the skip connection to prevent the gradient dissipation and explosion problems.

[0042] The formula of the residual module is as follows:

[0043]

[0044] in, It is The output of the residual block, is the output of the forward pass, It is The feature weight matrix of the residual block is a nonlinear transformation function.

[0045] (3) Max pooling layer.

[0046] It is used to perform dimensionality reduction processing on the fused time series feature data to obtain one-dimensional fused time series feature data.

[0047] The second hidden layer of the hidden layer includes two BiLSTM network layers and an expansion layer. The second hidden layer is used to complete the feature extraction of the one-dimensional fusion time series feature data from the front and back directions, and flatten the extracted features.

[0048] The BiLSTM network layer is composed of a forward LSTM and a backward LSTM. Figure 3 As shown in the figure, the weights and thresholds between neurons are shared in the forward and backward directions and have effective memory, which can learn the nonlinear characteristics of time series data. It can effectively alleviate the problem of information being forgotten during long-term processing, mine more characteristic information of time series data, and capture details that LSTM easily overlooks.

[0049] assumed For BiLSTM The hidden state at the moment is calculated as follows:

[0050]

[0051]

[0052] Where: and is the calculation process of LSTM, yes Input at the moment, is time t, The hidden layer state of the forward LSTM network at that moment, , for time, Hidden state of the backward LSTM network at time instant, hidden state of BiLSTM that is and The calculation formula is as follows:

[0053]

[0054] The LSTM network uses three gates to protect and control the cell state, filter and update information, see the attached Figure 4 shown.

[0055] (1) Forget gate: acts on the cell state and selectively forgets the information in the memory cells.

[0056]

[0057] (2) Input gate: acts on the cell state and selectively records new information into the new cell state.

[0058]

[0059]

[0060] (3) Cell status:

[0061]

[0062] (4) Output gate: acts on the input and hidden layer output, so that the final output includes both the cell state and the input, and updates the result to the next hidden layer.

[0063]

[0064]

[0065] Where: The input information at the current moment; is the connection weight of the forget gate; is the bias coefficient of the forget gate; is the connection weight of the input gate; is the bias coefficient of the input gate; is the connection weight of the memory unit; is the bias coefficient of the memory unit; is the connection weight of the output gate; is the bias coefficient of the output gate.

[0066] The above yield-related parameters can be determined by Pearson correlation coefficient analysis. For two random variables , The Pearson correlation coefficient is shown as follows:

[0067]

[0068] In the formula, x and y are the sample values ​​of two variables, and are the sample means of the two variables.

[0069] The yield-related parameters screened out by Pearson correlation analysis include at least one of the following parameters:

[0070] Porosity, permeability, oil saturation, bottom hole flowing pressure, daily production time, casing pressure and connected well injection volume.

[0071] In the input layer, the production-related parameters and historical time series data of multiple production wells are input, and a multi-factor production-related historical time series data set is formed through the data preprocessing layer, and the input multi-factor time series matrix is ​​constructed in chronological order.

[0072] Furthermore, data preprocessing includes data cleaning, normalization, and data partitioning.

[0073] Data cleaning: The collected historical time series data related to production must first be cleaned, mainly to deal with missing values, outliers and duplicate data. The residual error , if the following equation is satisfied, then Should be eliminated.

[0074]

[0075] Here, x represents the mean and δ represents the standard deviation.

[0076] Data standardization: Considering the dimensional differences and different value ranges between data features, data standardization is necessary to enhance the stability and accuracy of model training. This process makes the data follow a standard normal distribution by performing a linear transformation based on the mean and standard deviation of the data. The standardization formula is:

[0077]

[0078] In the formula, and are the minimum and maximum values ​​of the sample data respectively.

[0079] Data partitioning: When training a model, the collected data set needs to be divided into training set, test set, and validation set in order to evaluate the performance and generalization ability of the model.

[0080] During the iterative training of the long short-term memory network using historical data, the back-propagation method is used to adjust the network parameters, thereby enhancing the model's ability to predict the relationship between oil well output and other data.

[0081] The data are sorted using time series analysis methods, and the daily production data of the current and previous n-1 time points are continuously selected as input, and the data at the n+1th time point is used as the output target of the model prediction.

[0082] The long short-term memory network BiLSTM of the embodiment of the present invention combines the advantages of the above-mentioned convolution module CNN, residual module ResNet and BiLSTM network layer, combines the characteristics of oil well production data and production-related parameter characteristics, and establishes a water-drive reservoir production prediction model based on residual learning to improve BiLSTM. The model can scientifically and effectively predict production. The model structure consists of an input layer, a first hidden layer, a second hidden layer and an output layer. In the input layer, the relevant production data of the oil well is input, and after the production-related data is deleted and standardized through the data preprocessing layer, a multi-factor production prediction data set is formed, and the input multi-factor time series matrix is ​​constructed in chronological order. In the first hidden layer, the convolution and pooling operations in CNN are used to extract the preliminary features of the multi-factor yield time series, reduce the length of the time series data, and reduce the difficulty of subsequent network analysis; the residual module ResNet is added to CNN to deepen the depth of the network structure and improve the prediction effect of the model; in the second hidden layer, the preliminary features of the multi-factor yield time series are input into BiLSTM to complete the extraction of time series features in the forward and backward directions, and the BiLSTM is extracted to the final features for flattening; it is connected to the fully connected layer of the output layer, and finally the predicted output value of the yield is obtained.

[0083] Traditional LSTM can only perform autocorrelation learning on a single time series variable. The water-drive oil reservoir production prediction method provided in the embodiment of the present invention utilizes a BiLSTM network to extract and fuse features of multi-dimensional time series data by embedding a convolution module CNN and a residual module ResNet. It not only has the multi-dimensional feature extraction and fusion capabilities that the LSTM neural network lacks, but also effectively alleviates problems such as network performance degradation and gradient diffusion. It can retain long-term dependencies and important features of current input data when processing sequence data, so that it has better performance and prediction accuracy in sequence data prediction tasks such as oil well production prediction. It can cope with the challenge of large temporal variations in production data of oil fields in the high water content stage, and plays an important role in the adjustment strategy of oil field production and development.

[0084] Embodiment 2

[0085] Embodiment 2 of the present invention provides a method for predicting production of a water-driven oil reservoir, wherein production-related parameters of a well to be predicted in the water-driven oil reservoir, historical time series data of production, and production-related parameters at a time to be predicted are input into a production prediction model, and the production of the well to be predicted at a time to be predicted is obtained based on an output result of the model.

[0086] The yield prediction model is a model established by the method in the above-mentioned embodiment 1. The yield prediction model is used to predict the yield based on the input data using the forward propagation technology.

[0087] In order to verify the effect of the embodiment of the present invention, a data table is presented to prove the accuracy of the model. Through the gated recurrent neural network model test on the test set, 8 samples are selected for display. The results in Table 1 below show that the deviation between the predicted value and the actual value is small, indicating that the model has good prediction performance.

[0088] Table 1 Comparison of prediction results

[0089]

[0090] The embodiment of the present invention improves BiLSTM based on residual learning to predict the production of water-driven oil reservoirs, solves the problem that the accuracy of production prediction in the late stage of high water content in water-driven oil reservoirs is difficult to ensure, and takes into account the problems that the water-driven characteristic curve does not take into account the time series, the mathematical model is complicated to solve, and the model feature extraction ability is insufficient. Considering the dynamic change characteristics of production with development time, the residual module is combined to extract time series feature information, and the long-term memory information ability of the BiLSTM network is combined to establish a water-driven oil reservoir production prediction model. This method overcomes the limitations of other methods such as relying on engineer experience and taking a long time, and can quickly predict production, providing a basis for the formulation of subsequent development adjustment plans and the potential of residual oil.

[0091] Based on the inventive concept of the present invention, an embodiment of the present invention further provides a water drive oil reservoir production prediction device, the structure of which is as follows: Figure 5 As shown, including:

[0092] The production prediction model establishment module 51 is used to iteratively train the long short-term memory network using the production-related parameters of multiple production wells and the historical time series data of the production to obtain a production prediction model. The long short-term memory network includes an input layer, a first hidden layer, a second hidden layer and an output layer connected in sequence. The first hidden layer includes a convolution module, a residual module and a maximum pooling layer, which are used to extract and fuse features of multi-dimensional time series data and output the fused time series feature data. The second hidden layer is used to capture the time dependency of the fused time series feature data.

[0093] The production prediction module 52 is used to input the production-related parameters of the well to be predicted in the water drive reservoir, the historical time series data of the production, and the production-related parameters at the time to be predicted into the production prediction model, and obtain the production of the well to be predicted at the time to be predicted according to the output result of the model;

[0094] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0095] Based on the inventive concept of the present invention, an embodiment of the present invention further provides a computer storage medium, wherein the computer storage medium stores computer executable instructions, and when the computer executable instructions are executed by a processor, the above-mentioned water drive reservoir production prediction method is implemented.

[0096] Based on the inventive concept of the present invention, an embodiment of the present invention further provides a server, 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 water drive reservoir production prediction method when executing the program.

[0097] Unless otherwise specifically stated, terms such as processing, computing, calculating, determining, displaying, etc. may refer to the actions and / or processes of one or more processing or computing systems, or similar devices, which operate and convert data represented as physical (e.g., electronic) quantities within registers or memories of a processing system into other data similarly represented as physical quantities within memories, registers, or other such information storage, transmission, or display devices of the processing system. Information and signals may be represented using any of a variety of different techniques and methods. For example, data, instructions, commands, information, signals, bits, symbols, and chips mentioned throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, light fields or particles, or any combination thereof.

[0098] It should be understood that the specific order or hierarchy of steps in the disclosed process is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process can be rearranged without departing from the scope of protection of the present disclosure. The attached method claims present the elements of the various steps in an exemplary order and are not intended to be limited to the specific order or hierarchy described.

[0099] In the above detailed description, various features are grouped together in a single embodiment to simplify the disclosure. This method of disclosure should not be interpreted as reflecting an intention that an embodiment of the claimed subject matter requires more features than those set forth in each claim. On the contrary, as reflected in the appended claims, the invention is in a state of having less than all the features of the disclosed individual embodiments. Therefore, the appended claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.

[0100] Those skilled in the art will also appreciate that the various illustrative logic blocks, modules, circuits, and algorithmic steps described in conjunction with the embodiments herein can all be implemented as electronic hardware, computer software, or a combination thereof. In order to clearly illustrate the interchangeability between hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above around their functions. Whether such functions are implemented as hardware or software depends on specific applications and the design constraints imposed on the entire system. A skilled person can implement the described functions in an alternative manner for each specific application, but such implementation decisions should not be interpreted as departing from the scope of protection of the present disclosure.

[0101] The steps of the method or algorithm described in conjunction with the embodiments herein may be directly embodied as hardware, a software module executed by a processor, or a combination thereof. The software module may be located in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a mobile disk, a CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is connected to the processor so that the processor can read information from the storage medium and can write information to the storage medium. Of course, the storage medium may also be an integral part of the processor. The processor and the storage medium may be located in an ASIC. The ASIC may be located in a user terminal. Of course, the processor and the storage medium may also be present in a user terminal as discrete components.

[0102] For software implementation, the techniques described in this application can be implemented with modules (e.g., procedures, functions, etc.) that perform the functions described in this application. These software codes can be stored in a memory unit and executed by a processor. The memory unit can be implemented within the processor or outside the processor. In the latter case, it is coupled to the processor in a communication manner via various means, which are well known in the art.

[0103] The above description includes examples of one or more embodiments. Of course, it is impossible to describe all possible combinations of components or methods for the purpose of describing the above embodiments, but it should be recognized by those skilled in the art that the various embodiments can be further combined and arranged. Therefore, the embodiments described herein are intended to cover all such changes, modifications and variations that fall within the scope of protection of the appended claims. In addition, with respect to the term "comprising" used in the specification or claims, the word is covered in a manner similar to the term "including", just as "including," explained as a transitional word in the claims. In addition, any term "or" used in the specification of the claims is intended to represent "non-exclusive or". The terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance.

Claims

1. A method for predicting the production of a water-driven oil reservoir, characterized in that: include: Inputting the production-related parameters of the well to be predicted in the water drive reservoir, the historical time series data of the production, and the production-related parameters at the time to be predicted into the production prediction model, and obtaining the production of the well to be predicted at the time to be predicted according to the output result of the model; The production prediction model is obtained by iteratively training a long short-term memory network using production-related parameters of multiple production wells and historical time series data of production; the long short-term memory network includes an input layer, a first hidden layer, a second hidden layer and an output layer connected in sequence; the input layer is used to construct a multi-factor time series matrix of the input multi-dimensional time series data in chronological order; the first hidden layer includes a convolution module, a residual module and a maximum pooling layer; the convolution module is used to extract feature data from the multi-dimensional time series matrix through convolution calculation, and obtain one-dimensional time series feature data through dimensionality reduction and downsampling processing; The residual module stacks convolution layers, and prevents gradient vanishing and explosion problems by adding jump connections. The residual module is used to extract features from the one-dimensional time series feature data by stacking convolution layers, and fuse the extracted features with the original multi-dimensional time series matrix through jump connections to obtain fused time series feature data; the maximum pooling layer is used to perform dimensionality reduction processing on the fused time series feature data to obtain one-dimensional fused time series feature data; the second hidden layer includes two BiLSTM network layers and an expansion layer, the two BiLSTM network layers are used to capture the time dependency of the one-dimensional fused time series feature data, and complete the feature extraction of the one-dimensional fused time series feature data from the front and back directions, and the expansion layer is used to flatten the extracted features.

2. The method according to claim 1, characterized in that The yield-related parameters include at least one of the following parameters: Porosity, permeability, oil saturation, bottom hole flowing pressure, daily production time, casing pressure and connected well injection volume.

3. The method according to claim 1, characterized in that During the iterative training of the long short-term memory network, the back-propagation method is used to adjust the network parameters.

4. The method according to claim 1, characterized in that The yield prediction model is used to predict the yield based on input data using a forward propagation technique.

5. A water drive oil reservoir production prediction device, characterized in that: include: The production prediction model establishment module is used to iteratively train the long short-term memory network using the production-related parameters of multiple production wells and the historical time series data of the production to obtain the production prediction model; the long short-term memory network includes an input layer, a first hidden layer, a second hidden layer and an output layer connected in sequence; the input layer is used to construct a multi-factor time series matrix of the input multi-dimensional time series data in chronological order; the first hidden layer includes a convolution module, a residual module and a maximum pooling layer; the convolution module is used to extract feature data from the multi-dimensional time series matrix through convolution calculation, and obtain one-dimensional time series feature data through dimensionality reduction and downsampling processing; The residual module stacks convolutional layers, and prevents gradient dissipation and explosion problems by adding jump connections. The residual module is used to extract features from the one-dimensional time series feature data by stacking convolutional layers, and fuse the extracted features with the original multi-dimensional time series matrix through jump connections to obtain fused time series feature data; the maximum pooling layer is used to perform dimensionality reduction processing on the fused time series feature data to obtain one-dimensional fused time series feature data; the second hidden layer includes two BiLSTM network layers and an expansion layer, and the two BiLSTM network layers are used to capture the time dependency of the one-dimensional fused time series feature data, complete the feature extraction of the one-dimensional fused time series feature data from the front and back directions, and the expansion layer is used to flatten the extracted features; The production prediction module is used to input the production-related parameters of the wells to be predicted in the water-driven oil reservoir, the historical time series data of the production, and the production-related parameters at the time to be predicted into the production prediction model, and obtain the production of the wells to be predicted at the time to be predicted according to the output results of the model.

6. A computer storage medium, characterized in that: The computer storage medium stores computer executable instructions, and when the computer executable instructions are executed by a processor, the water drive reservoir production prediction method according to any one of claims 1 to 4 is implemented.

7. A server, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the water drive reservoir production prediction method according to any one of claims 1 to 4 when executing the program.

Images