Single well index prediction method and system based on long short-term memory neural network

Abstract

The present application discloses a single well index prediction method and system based on long short-term memory neural network, the method comprising: receiving the type information of the single well to be predicted, the type information comprising development mode information and water-containing stage information; inputting the type information of the single well to be predicted into a preset long short-term memory neural network LSTM model, and obtaining the single well production and / or single well water content data corresponding to the type information; wherein the LSTM model is trained using the historical data of the single well corresponding to the type information, the historical data comprising geological data, production data and monitoring data. Thus, the accuracy of the prediction of single well production and water content index is improved.

Description

Technical Field

[0001] The present application relates to the technical field of single well index prediction in oil reservoir development, and in particular to a single well index prediction method, system, device and readable storage medium based on long short-term memory neural network. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the present application as recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.

[0003] Well dynamic analysis, development effect evaluation, and preparation of development plans for new and old areas are important means of oilfield development. Single well production decline and water content prediction are the basis of the above work. At present, single well production decline and water content prediction mainly rely on manual analysis and different software, lack of inheritance and low efficiency. The existing indicator prediction methods are single and poorly applicable, and the prediction accuracy needs to be improved. The prediction results are easily affected by human factors, and the timeliness of methods such as reservoir numerical simulation is not strong.

[0004] As early as 1994, Wu Xingen in my country used ANN to accurately predict the oil production of oil fields. With the rapid development of artificial intelligence technology, domestic and foreign scholars have explored the prediction of other indicators, including experiments on shale gas production, block oil production, block water content, and new well oil production, and have achieved good results. However, the current methods cannot achieve multi-dimensional single well production decline and water content prediction corresponding to different reservoir types, different development methods, different water content stages, etc., and therefore cannot accurately allocate oil well production. Summary of the invention

[0005] To this end, the embodiment of the present application provides a single well index prediction method, system, device and readable storage medium based on long short-term memory neural network, which solves the above technical problems. Taking into account the historical data such as single well static geological data, production data, water injection data, operation data, etc., and using artificial intelligence methods, a single well production decline and water content prediction model for different development methods and different water content stages is established to achieve multi-dimensional single well production decline and water content prediction, and accurately support oil well production prediction and production allocation.

[0006] In a first aspect, an embodiment of the present application provides a single well index prediction method based on a long short-term memory neural network, comprising:

[0007] receiving type information of a single well to be predicted, wherein the type information includes development mode information and water-containing stage information;

[0008] The type information of the single well to be predicted is input into a preset long short-term memory neural network LSTM model to obtain the single well production and / or single well water content data corresponding to the type information; wherein the LSTM model is trained using the historical data of the single well corresponding to the type information, and the historical data includes geological data, production data and monitoring data.

[0009] Optionally, the method further comprises:

[0010] According to the type information of the single well to be predicted, the LSTM model with corresponding type information is selected.

[0011] Optionally, the method comprises:

[0012] According to the single well production threshold, LSTM models are established for different types of single well data sample libraries;

[0013] Train each LSTM model.

[0014] Optionally, the LSTM models are established for different types of single well data sample libraries respectively, including:

[0015] According to the production level of a single well, the threshold is set, and the samples are set as high-production wells, medium-production wells, and low-production wells;

[0016] LSTM models are established for the three samples respectively.

[0017] Optionally, the training for each LSTM model includes:

[0018] For each model, the sample data is divided into training set, test set and validation set;

[0019] The labeled training set and the main controlling influencing factors corresponding to the training set are used as input samples of the LSTM model, and the LSTM model is trained to obtain different types of single well oil production and water content prediction models;

[0020] The trained LSTM model is verified using the test set, and the LSTM model that meets the set conditions is determined to be a qualified LSTM model.

[0021] In a second aspect, the embodiment of the present application further provides a single well index prediction system based on a long short-term memory neural network, the system comprising:

[0022] An information receiving module, used for receiving type information of a single well to be predicted, wherein the type information includes development mode information and water-containing stage information;

[0023] The LSTM model prediction module is used to input the type information of the single well to be predicted into a preset long short-term memory neural network LSTM model to obtain the single well production and / or single well water content data corresponding to the type information; wherein the LSTM model is trained using the historical data of the single well corresponding to the type information, and the historical data includes geological data, production data and monitoring data.

[0024] Optionally, the system further comprises:

[0025] The LSTM model selection module is used to select the LSTM model with corresponding type information according to the type information of the single well to be predicted.

[0026] Optionally, the LSTM model selection module is specifically used to:

[0027] According to the single well production threshold, LSTM models are established for different types of single well data sample libraries;

[0028] Train each LSTM model.

[0029] Optionally, the LSTM models are established for different types of single well data sample libraries respectively, including:

[0030] According to the production level of a single well, the threshold is set, and the samples are set as high-production wells, medium-production wells, and low-production wells;

[0031] LSTM models are established for the three samples respectively.

[0032] Optionally, the training for each LSTM model includes:

[0033] For each model, the sample data is divided into training set, test set and validation set;

[0034] The labeled training set and the main controlling influencing factors corresponding to the training set are used as input samples of the LSTM model, and the LSTM model is trained to obtain different types of single well oil production and water content prediction models;

[0035] The trained LSTM model is verified using the test set, and the LSTM model that meets the set conditions is determined to be a qualified LSTM model.

[0036] In a third aspect, an embodiment of the present application further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the methods described in the first aspect is implemented.

[0037] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for executing any of the methods described in the first aspect.

[0038] In summary, the embodiment of the present application provides a single well index prediction method, system, device and readable storage medium based on long short-term memory neural network, by receiving the type information of the single well to be predicted, the type information includes development mode information and water content stage information; the type information of the single well to be predicted is input into the preset long short-term memory neural network LSTM model, and the single well production and / or single well water content data corresponding to the type information are obtained; wherein the LSTM model is trained using the historical data of the single well corresponding to the type information, and the historical data includes geological data, production data and monitoring data. Thereby improving the accuracy of the prediction of single well production and water content index. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0040] Figure 1 A schematic diagram of a single well index prediction method based on a long short-term memory neural network provided in an embodiment of the present application;

[0041] Figure 2a , Figure 2b and Figure 2c A schematic diagram of constructing a sample according to single well production data provided in an embodiment of the present application;

[0042] Figure 3 A schematic diagram of an LSTM neural network provided in an embodiment of the present application;

[0043] Figure 4 This is a schematic diagram of the LSTM neural network memory unit structure provided in the embodiments of the present application;

[0044] Figure 5a , Figure 5b and Figure 5c A comparison chart of the single well oil production prediction effect based on the LSTM neural network provided in the embodiments of the present application;

[0045] Figure 6a , Figure 6b and Figure 6c A comparison chart of the single well water content prediction effect based on the LSTM neural network provided in the embodiments of the present application;

[0046] Figure 7 This is a block diagram of a single well index prediction system based on a long short-term memory neural network provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. The described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] The principles and spirit of the present application are explained in detail below with reference to several representative implementations of the present application.

[0049] Although the present application provides method operation steps or device structures as shown in the following embodiments or drawings, more or fewer operation steps or module units may be included in the method or device based on routine or no creative labor. In the steps or structures where there is no necessary causal relationship logically, the execution order of these steps or the module structure of the device is not limited to the execution order or module structure shown in the embodiments or drawings of the present application. When the method or module structure is applied to an actual device or terminal product, it can be executed sequentially or in parallel according to the method or module structure shown in the embodiments or drawings.

[0050] Figure 1 The present invention provides a schematic flow chart of a single well index prediction method based on a long short-term memory neural network, which comprises the following steps:

[0051] Step 101: receiving type information of a single well to be predicted, wherein the type information includes development mode information and water-containing stage information.

[0052] Step 102: Input the type information of the single well to be predicted into a preset long short-term memory neural network LSTM model to obtain the single well production and / or single well water content data corresponding to the type information; wherein the LSTM model is trained using the historical data of the single well corresponding to the type information, and the historical data includes geological data, production data and monitoring data.

[0053] In a possible implementation, the method further includes: selecting an LSTM model corresponding to the type information according to the type information of the single well to be predicted.

[0054] In a possible implementation, the method includes: establishing LSTM models for different types of single-well data sample libraries according to a single-well production threshold; and training each LSTM model.

[0055] In a possible implementation, the LSTM models are established for different types of single well data sample libraries, including: setting thresholds according to the production of single wells, setting samples as high-production wells, medium-production wells and low-production wells; and establishing LSTM models for the three types of samples.

[0056] In a possible implementation, the training for each LSTM model includes: for each model, dividing the sample data into a training set, a test set, and a validation set; using the labeled training set and the main controlling influencing factors corresponding to the training set as input samples of the LSTM model, training the LSTM model to obtain different types of single-well oil production and water content prediction models; using the test set to verify the trained LSTM model, and determining the LSTM model that meets the set conditions as a qualified LSTM model.

[0057] Comprehensively consider the single well geological data, production dynamics, measures data, dynamic monitoring and other related data, and sort out the data items involved in the analysis. Establish a preprocessing method for each data item, including outlier processing, missing value replacement, etc. Combined with professional knowledge and expert experience, determine all types of production decline and water content increase covered by Changyuan water drive single wells, and establish a single well data sample library based on different types, such as low water content well samples, medium water content well samples, high water content well samples, ultra-high water content well samples, fracturing well samples, and plugging well samples; the sample data is divided into training set, test set and validation set.

[0058] Through correlation analysis, the main key parameters affecting different types of samples are identified; the labeled training data set and its corresponding main controlling factors are used as input samples of the long short-term memory neural network, and the neural network is trained to obtain different types of single well oil production and water cut prediction models. Then, the trained neural network model is verified using the test data set. If the accuracy requirements are met, the trained single well oil production prediction and water cut prediction models are obtained; the trained neural network model is used to automatically predict the trend of different types of single well oil production or water cut.

[0059] The single well production and water content prediction method based on long short-term memory neural network can realize the rapid prediction of single well production and water content in different development modes and different water content stages, greatly improve the accuracy of single well index prediction, and make the single well production and water content prediction accuracy reach 90-95%. In addition, this method can make the prediction model obtained after each machine learning be precipitated, so that the single well index prediction technology has learning ability and obtains higher single well index prediction accuracy, which is of great significance to reservoir dynamic analysis and development planning.

[0060] The following is a further explanation of the single well index prediction method based on the long short-term memory neural network provided in the embodiment of the present application, which specifically includes the following steps:

[0061] Step 1: Collect single well data and calibrate different types of single well samples.

[0062] Using the monthly production curve of a single well, sample calibration is performed by selecting a sample time period, and the results of the sample calibration are classified and saved. Old well samples are divided into low water-content well samples, medium water-content well samples, high water-content well samples, and extra-high water-content well samples. Measures well samples are divided into perforation repair samples and fracturing samples.

[0063] Step 2: Data preprocessing. The preprocessing methods of various types of original single well data are as follows:

[0064] Outlier processing: According to the reasonable value range of different features formulated by business experts, thresholds are set for feature values ​​in the data for filtering, and unreasonable values ​​exceeding the threshold are corrected by adjacent data or well average data.

[0065] Missing value processing: Supplement reasonable values ​​for missing data in single well production data.

[0066] Data normalization: In the field of machine learning, different feature vectors often have different dimensions and dimensional units, which will affect the results of data analysis. In order to eliminate the dimensional impact between indicators, data normalization is required to solve the comparability between data indicators. After the original data is normalized, each indicator is at the same order of magnitude, which is suitable for comprehensive comparative evaluation. Using the Z-score normalization method, the data is processed to conform to the standard normal distribution, that is, the mean is 0 and the standard deviation is 1.

[0067] The specific standardization method is as follows: Assume that the discrete sequence of a feature is X = {x1, x2....x n}, the series average is The normalized calculation formula is shown in the following formula (1):

[0068]

[0069] Where: x* Normalized features for oil well production data, is the average value of a characteristic of the oil well, and σ is the standard deviation of the data.

[0070] Correlation analysis: In order to reduce the number of data features, reduce the correlation between features, make full use of features with high correlation with oil production, calculate the Pearson correlation coefficient between features, and optimize the features.

[0071] Step 3: Classify the samples. Figure 2a , Figure 2b and Figure 2c ,Due to the large difference in production between individual wells due to reasons such as the block and the year of well opening, it makes model prediction more difficult. Therefore, according to the level of production, thresholds are set and samples are set as high-yield, medium-yield and low-yield wells. Model the three types of samples separately.

[0072] Step 4: Determine the structure of the long and short time neural network based on the sample characteristics. Figure 3 , the long short-term memory neural network (LSTM) was selected. LSTM is a neural network with memory function (reference Figure 4 ) can fully consider the changing trends between samples over time.

[0073] In the embodiment of the present application, the LSTM neural network includes 4 LSTM layers. A Dropput layer is added to each LSTM layer to eliminate the problem of overfitting of the LSTM model. The Tanh activation function is used to increase the nonlinear fitting ability of the model. MSE is used as the target optimization function, and the Adam optimizer is used to optimize the target function.

[0074] Step 5: Select 90% of the well samples as the training set, train the LSTM model, adjust the network parameters to ensure that the prediction accuracy is met, and obtain the trained LSTM neural network.

[0075] Step 6: Use 10% of the labeled data as a validation set and use the validation set to evaluate the accuracy of the network model.

[0076] Step 7: After model evaluation and verification, the single-well production and single-well water content models with good results are automatically predicted for the test data, and the prediction results are manually verified.

[0077] Figure 5a , Figure 5b and Figure 5cThe following is a comparison chart of the production prediction effect of actual single well data, which are the production predictions of low-yield wells in three water-containing stages. The blue line is the actual monthly production data of a single well, and the orange line is the fitted monthly production of a single well. By comparison, it can be seen that the trained LSTM model is more accurate in predicting monthly production. Figure 6a , Figure 6b and Figure 6c The comparison chart of water cut prediction effect of actual single well data shows the water cut prediction of low-yield wells in three water cut stages. The blue line is the actual water cut of the single well, and the orange line is the fitted monthly water cut of the single well. By comparison, it can be seen that the water cut change trend predicted by the LSTM model is consistent with the actual water cut trend.

[0078] It can be seen that the single well production and water content prediction method based on long short-term memory neural network provided by the embodiment of the present application greatly improves the single well index prediction accuracy, greatly reducing the prediction workload of reservoir developers. In addition, the prediction model obtained after each machine learning can be precipitated, so that the single well index prediction technology has learning ability and obtains higher single well index prediction accuracy, which is of great significance to reservoir dynamic analysis and development planning.

[0079] In summary, the embodiment of the present application provides a single well index prediction method based on a long short-term memory neural network, by receiving the type information of the single well to be predicted, the type information includes development mode information and water content stage information; the type information of the single well to be predicted is input into a preset long short-term memory neural network LSTM model, and the single well production and / or single well water content data corresponding to the type information are obtained; wherein the LSTM model is trained using the historical data of the single well corresponding to the type information, and the historical data includes geological data, production data and monitoring data. Thereby improving the accuracy of the prediction of single well production and water content indicators. Comprehensively considering the historical data such as single well static geological data, production data, water injection data, operation data, etc., using artificial intelligence methods, a single well production decline and water content prediction model for different development methods and different water content stages is established, realizing multi-dimensional single well production decline and water content prediction, accurately supporting oil well production prediction and production allocation, improving the level of dynamic management of oil well production, and improving the accuracy of single well measures decision-making, and finally realizing the improvement of the final recovery rate and production efficiency of the oil field. By utilizing the transferable learning features of artificial intelligence, it is promoted and applied in water-driven sandstone reservoirs while expanding its application to other reservoir types.

[0080] Based on the same technical concept, the embodiment of the present application also provides a block diagram of a single well index prediction system based on a long short-term memory neural network, such as Figure 7 , including:

[0081] The information receiving module 701 is used to receive the type information of the single well to be predicted, wherein the type information includes development mode information and water-containing stage information.

[0082] The LSTM model prediction module 702 is used to input the type information of the single well to be predicted into a preset long short-term memory neural network LSTM model to obtain the single well production and / or single well water content data corresponding to the type information; wherein the LSTM model is trained using the historical data of the single well corresponding to the type information, and the historical data includes geological data, production data and monitoring data.

[0083] In a possible implementation, the system further includes: an LSTM model selection module, configured to select an LSTM model corresponding to type information according to type information of a single well to be predicted.

[0084] In a possible implementation, the LSTM model selection module is specifically used to: establish LSTM models for different types of single well data sample libraries according to a single well production threshold; and train each LSTM model.

[0085] In a possible implementation, the LSTM models are established for different types of single well data sample libraries, including: setting thresholds according to the production of single wells, setting samples as high-production wells, medium-production wells and low-production wells; and establishing LSTM models for the three types of samples.

[0086] In a possible implementation, the training for each LSTM model includes: for each model, dividing the sample data into a training set, a test set, and a validation set; using the labeled training set and the main controlling influencing factors corresponding to the training set as input samples of the LSTM model, training the LSTM model to obtain different types of single-well oil production and water content prediction models; using the test set to verify the trained LSTM model, and determining the LSTM model that meets the set conditions as a qualified LSTM model.

[0087] Based on the same technical concept, an embodiment of the present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the above methods when executing the computer program.

[0088] Based on the same technical concept, an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program for executing any of the above methods.

[0089] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0090] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0091] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0092] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0093] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A single well index prediction method based on long short-term memory neural network, characterized in that: include: receiving type information of a single well to be predicted, wherein the type information includes development mode information and water-containing stage information; According to the type information of the single well to be predicted, a long short-term memory neural network LSTM model corresponding to the type information is selected from a plurality of pre-established long short-term memory neural network LSTM models; Input the type information of the single well to be predicted into the selected long short-term memory neural network LSTM model to obtain the single well production and / or single well water content data corresponding to the type information; wherein the pre-established multiple long short-term memory neural network LSTM models are trained using historical data of single wells with different types of information, and the historical data include geological data, production data and monitoring data; Furthermore, the method further comprises: According to the single well production threshold, LSTM models are established for different types of single well data sample libraries; Train each LSTM model; The LSTM models are established for different types of single well data sample libraries, including: According to the production level of a single well, the threshold is set, and the samples are set as high-production wells, medium-production wells, and low-production wells; LSTM models are established for the three samples respectively; Wherein, the single well is an oil well.

2. The method according to claim 1, characterized in that The training for each LSTM model includes: For each model, the sample data is divided into training set, test set and validation set; The labeled training set and the main controlling influencing factors corresponding to the training set are used as input samples of the LSTM model, and the LSTM model is trained to obtain different types of single well oil production and water content prediction models; The trained LSTM model is verified using the test set, and the LSTM model that meets the set conditions is determined to be a qualified LSTM model.

3. A single well index prediction system based on long short-term memory neural network, characterized in that: The system comprises: An information receiving module, used for receiving type information of a single well to be predicted, wherein the type information includes development mode information and water-containing stage information; An LSTM model selection module is used to select a LSTM model corresponding to the type information from a plurality of pre-established LSTM models according to the type information of the single well to be predicted; An LSTM model prediction module is used to input the type information of the single well to be predicted into the selected long short-term memory neural network LSTM model to obtain the single well production and / or single well water content data corresponding to the type information; wherein the pre-established multiple long short-term memory neural network LSTM models are pre-trained using the historical data of the single wells corresponding to different types of information, and the historical data include geological data, production data and monitoring data; Furthermore, the LSTM model selection module is specifically used for: According to the single well production threshold, LSTM models are established for different types of single well data sample libraries; Train each LSTM model; The LSTM models are established for different types of single well data sample libraries, including: According to the production level of a single well, the threshold is set, and the samples are set as high-production wells, medium-production wells, and low-production wells; LSTM models are established for the three samples respectively; Wherein, the single well is an oil well.

4. The system according to claim 3, characterized in that The training for each LSTM model includes: For each model, the sample data is divided into training set, test set and validation set; The labeled training set and the main controlling influencing factors corresponding to the training set are used as input samples of the LSTM model, and the LSTM model is trained to obtain different types of single well oil production and water content prediction models; The trained LSTM model is verified using the test set, and the LSTM model that meets the set conditions is determined to be a qualified LSTM model.

5. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to claim 1 or 2 is implemented.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the method according to claim 1 or 2.

Images