Dynamic production prediction method and system of fractured gas wells based on physical data drive

Abstract

The present invention belongs to the technical field of production prediction of fractured gas wells, and specifically relates to a method and system for dynamic production prediction of fractured gas wells driven by physical data, aiming to solve the problems of difficulty and inaccurate prediction of gas well production prediction methods in the prior art. The present invention includes: obtaining historical production data, wellbore parameters and gas production of gas wells; calculating the flow rate and viscosity of mixed fluids based on historical production data and wellbore parameters; calculating the Reynolds number based on the flow rate and viscosity of mixed fluids, and calculating the friction coefficient based on wellbore parameters and the Reynolds number; calculating the wellbore friction based on the friction coefficient; calculating the bottom hole flow pressure according to the wellbore friction; training the BiLSTM neural network prediction model based on historical production data and bottom hole flow pressure to obtain a gas production prediction model; bringing the bottom hole flow pressure and historical production data into the gas production prediction model to obtain the corresponding gas production. The present invention is accurate, scientific and reasonable.

Description

Background Art

[0002] As an unconventional energy with abundant reserves and huge environmental protection potential, natural gas has attracted widespread attention. Its development and utilization are of great significance to reducing the use of traditional fossil fuels. However, due to the development of micro-nano pores in natural gas (NG) reservoirs, the flow mechanism of natural gas in these micro-nano pores is very complex, covering multiple modes such as viscous flow, Knudsen diffusion, free molecular flow and surface diffusion, which challenges the traditional seepage theory. Therefore, in order to achieve efficient natural gas extraction, it is crucial to accurately predict the production of fracturing gas wells and improve the prediction accuracy for the development and design of natural gas wells. At present, the production prediction methods of fracturing gas wells are mainly divided into two categories: physical modeling-based and data-driven. The physical modeling-based method relies on geological and engineering principles to establish theoretical models to predict production. However, these methods often rely on idealized assumptions and are difficult to adapt to complex and changeable geological and engineering conditions, so they are limited in generalization ability. On the other hand, traditional data-driven prediction methods usually analyze the production decline trend based on the temporal autocorrelation of production data. Although these models can accurately describe the production decline law, their structure is relatively complex and difficult to solve. In recent years, machine learning methods have gradually become an emerging method for predicting the production of fracturing gas wells. However, traditional machine learning methods have certain limitations. For example, pure data-driven methods rely entirely on historical data for prediction. If the historical data is insufficient or of poor quality, the model's prediction results may be affected, making it difficult to accurately capture future trends and changes. In addition, there is a lack of clear physical explanations, making it difficult to provide useful explanatory information. This can make it difficult to understand the model's prediction results and make decisions based on the model.

[0003] Therefore, the gas well production prediction method in the prior art has the problems of great difficulty and inaccurate prediction results. Summary of the invention

[0004] In order to solve the above-mentioned problems in the prior art, that is, the gas well production prediction methods in the prior art are difficult and the prediction results are inaccurate, the present invention provides a method for predicting the dynamic production of a fractured gas well based on physical data driving, the method comprising:

[0005] Obtain historical production data, wellbore parameters and gas production of gas wells;

[0006] Calculate the mixed fluid flow rate and the mixed fluid viscosity based on the historical production data and the wellbore parameters; calculate the Reynolds number based on the mixed fluid flow rate and the mixed fluid viscosity;

[0007] Calculating the friction coefficient based on the wellbore parameters and the Reynolds number; calculating the wellbore friction based on the friction coefficient; calculating the bottom hole flow pressure based on the wellbore friction;

[0008] Based on the historical production data and the bottom hole flow pressure, a BiLSTM neural network prediction model is trained to obtain a gas production prediction model; the bottom hole flow pressure and historical production data are brought into the gas production prediction model to obtain the corresponding gas production.

[0009] In a preferred embodiment, the method for calculating the flow rate of the mixed fluid is:

[0010] ;

[0011] in, is the mixed fluid flow rate, is the current daily gas production, is the current daily water production, is the inner radius of the pipe column.

[0012] In a preferred embodiment, the method for calculating the viscosity of the mixed fluid is:

[0013] ;

[0014] in, is the viscosity of the mixed fluid, is the liquid viscosity, is the gas viscosity, is the gas-liquid volume ratio.

[0015] In a preferred embodiment, the method for calculating the Reynolds number is:

[0016] ;

[0017] in, is the Reynolds number, is the density of the mixed fluid, is the mixed fluid flow rate, is the inner radius of the pipe string, is the viscosity of the mixed fluid.

[0018] In a preferred embodiment, the method for calculating the friction coefficient based on the wellbore parameters includes:

[0019] exist When , the fluid is in laminar flow state; When , the fluid is in transitional flow state;

[0020] ;

[0021] in, f is the friction coefficient, is the Reynolds number, is the absolute roughness of the inner wall of the pipe string, is the inner radius of the pipe column.

[0022] In a preferred embodiment, the method for calculating the friction coefficient based on the wellbore parameters also includes a method for calculating the friction coefficient when the Reynolds number is greater than or equal to 4000, specifically including:

[0023] When the Reynolds number is greater than or equal to 4000, it is a turbulent state; firstly, the first coefficient is calculated, and the method for calculating the first coefficient includes:

[0024] Assign weights to the Reynolds numbers at the observation points:

[0025] ;

[0026] in, is the weight of the Reynolds number in the observation point within the window, are the estimated points for the fitted polynomial, No. i The Reynolds number of the observation point, is the bandwidth size;

[0027] Calculate fitting polynomial parameters according to the weights of the Reynolds numbers in the observation points;

[0028] ;

[0029] are the parameters of the polynomial; is a weight function based on and The weight is assigned by the distance; n is the number of observation points; It is i The friction coefficient corresponding to the Reynolds number of each observation point;

[0030] When M is the smallest, its corresponding , … is the fitting polynomial parameter; M is the adjustment parameter;

[0031] The first coefficient is calculated based on the fitting polynomial parameter, specifically:

[0032] ;

[0033] in, is the first coefficient, x is the Reynolds number for calculating the friction coefficient in the turbulent stage, is the estimated point of the fitted polynomial, p is the order of the fitted polynomial, , … To fit the polynomial parameters, is the friction coefficient corresponding to the Reynolds number at the observation point;

[0034] Then, a second coefficient is calculated. The method for calculating the second coefficient includes:

[0035] ;

[0036] in, is the second coefficient, is the Reynolds number when the fluid in the wellbore is in the laminar flow stage due to different fluid density and / or fluid velocity and / or fluid viscosity, Reynolds number The corresponding friction coefficient is is the Reynolds number when the fluid in the wellbore is in the turbulent stage due to different fluid density and / or fluid velocity and / or fluid viscosity, Reynolds number The corresponding friction coefficient is The Reynolds number corresponding to the fluid for which the friction coefficient needs to be calculated in the turbulent stage;

[0037] The friction coefficient is calculated based on the first coefficient and the second coefficient:

[0038] ;

[0039] Where y represents the friction coefficient corresponding to the Reynolds number x of the fluid whose friction coefficient needs to be calculated in the turbulent stage;

[0040] In a preferred embodiment, the method for calculating the wellbore friction resistance based on the Reynolds number and the friction resistance coefficient is:

[0041] ;

[0042] in, The friction of the wellbore, is the density of the mixed fluid, is the mixed fluid flow rate, is the inner radius of the pipe string, is the friction coefficient, is the wellbore length.

[0043] In a preferred embodiment, the method for calculating the bottom hole flow pressure according to the wellbore friction is:

[0044] ;

[0045] in, is the bottom hole flowing pressure, is the wellhead pressure, is the acceleration due to gravity, is the density of the liquid, is the wellbore height, is the curvature radius of the deflection section, It is the friction of the wellbore.

[0046] In a second aspect of the present invention, a system for predicting the dynamic production of a fractured gas well driven by physical data is proposed, the system comprising:

[0047] A data acquisition module is used to obtain historical production data, wellbore parameters and gas production of gas wells;

[0048] A first calculation module, for calculating the mixed fluid flow rate and the mixed fluid viscosity based on the historical production data and the wellbore parameters;

[0049] A second calculation module is used to calculate the Reynolds number based on the mixed fluid flow rate and the mixed fluid viscosity, calculate the friction coefficient based on the wellbore parameters and the Reynolds number; and calculate the wellbore friction based on the friction coefficient;

[0050] A third calculation module is used to calculate the bottom hole flow pressure according to the wellbore friction resistance;

[0051] The gas production prediction module is used to train the BiLSTM neural network prediction model based on the historical production data and the bottom hole flow pressure to obtain a gas production prediction model; the bottom hole flow pressure and historical production data are brought into the gas production prediction model to obtain the corresponding gas production.

[0052] The third aspect of the present invention proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned physical data-driven method for dynamic production prediction of fractured gas wells.

[0053] Beneficial effects of the present invention:

[0054] (1) The physical data-driven prediction method of this application is an effective solution. By combining physical modeling with data-driven methods, it can more accurately capture the relationship between complex underground fluid flow characteristics and actual production data, thereby improving prediction accuracy;

[0055] (2) This study will design and evaluate a new neural network model based on actual natural gas well data to improve the accuracy of fractured gas well production prediction;

[0056] (3) The physical data-driven method for predicting the dynamic production of fractured gas wells in this application is easy to implement and is comprehensive, scientific and reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0058] Figure 1 It is a flow chart of a method for predicting dynamic production of a fractured gas well based on physical data drive according to an embodiment of the present invention;

[0059] Figure 2 is a schematic diagram of a daily gas production variation diagram of a natural gas well in an embodiment of the present invention;

[0060] Figure 3 It is a schematic diagram of the correlation analysis between various production data of natural gas wells and daily gas production data according to an embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of a sliding window algorithm conversion in an embodiment of the present invention;

[0062] Figure 5 is a schematic diagram of a bidirectional long short-term memory neural unit in an embodiment of the present invention;

[0063] Figure 6 It is a complete structural diagram of a time series prediction network model from input to output in an embodiment of the present invention;

[0064] Figure 7 is a comparison chart of the prediction results and true values ​​of various prediction models in the embodiment of the present invention;

[0065] Figure 8 It is a structural diagram of a computer system of a server for implementing the method, system, and device embodiments of the present application. DETAILED DESCRIPTION

[0066] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0067] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0068] The present invention provides a method for predicting the dynamic production of a fracturing gas well based on physical data drive, the method comprising:

[0069] Obtain historical production data, wellbore parameters and gas production of gas wells;

[0070] Calculating the mixed fluid flow rate and the mixed fluid viscosity based on the historical production data and the wellbore parameters;

[0071] Calculate the Reynolds number based on the mixed fluid flow rate and the mixed fluid viscosity, calculate the friction coefficient based on the wellbore parameters and the Reynolds number; calculate the wellbore friction based on the friction coefficient;

[0072] Calculating bottom hole flow pressure according to the wellbore friction;

[0073] Based on the historical production data and the bottom hole flow pressure, a BiLSTM neural network prediction model is trained to obtain a gas production prediction model; the bottom hole flow pressure and historical production data are brought into the gas production prediction model to obtain the corresponding gas production.

[0074] In order to more clearly explain the method for predicting the dynamic production of a fractured gas well based on physical data drive of the present invention, the following is combined with Figure 1 Each step in the embodiment of the present invention is described in detail.

[0075] The first embodiment of the present invention is a method for predicting the dynamic production of a fractured gas well based on physical data drive, and each step is described in detail as follows:

[0076] Obtain historical production data, wellbore parameters and gas production of gas wells;

[0077] In this embodiment, historical production data and wellbore parameters of M natural gas wells are obtained, wherein the historical production data include: N daily gas productions of a single well in continuous production, and the corresponding daily water production, oil pressure, casing pressure and production time data on that day; the wellbore parameters include: well depth, depth of the deflection point, length of the horizontal section, pipe string straightness, casing diameter, wellbore diameter, absolute roughness of the inner wall of the pipe string, number of wellbore sections, etc.; M is an integer greater than or equal to 2; N is an integer greater than or equal to 100;

[0078] Calculating the mixed fluid flow rate and the mixed fluid viscosity based on the historical production data and the wellbore parameters;

[0079] In this embodiment, the method for calculating the flow rate of the mixed fluid is:

[0080] ;

[0081] in, is the mixed fluid flow rate, is the current daily gas production, is the current daily water production, is the inner radius of the pipe column.

[0082] In this embodiment, the method for calculating the viscosity of the mixed fluid is:

[0083] ;

[0084] in, is the viscosity of the mixed fluid, is the liquid viscosity, is the gas viscosity, is the gas-liquid volume ratio.

[0085] In this embodiment, the state of the fluid is also related to the temperature gradient, the gas state, the gas deviation factor, and the total energy balance equation:

[0086] The new features are constructed based on the original production feature data, and the new production features include gas-liquid ratio, average gas production rate, average liquid production rate and oil-casing pressure difference;

[0087] Calculate the Reynolds number based on the mixed fluid flow rate and the mixed fluid viscosity, calculate the friction coefficient based on the wellbore parameters and the Reynolds number; calculate the wellbore friction based on the friction coefficient;

[0088] In this embodiment, the method for calculating the Reynolds number is:

[0089] ;

[0090] in, is the Reynolds number, is the density of the mixed fluid, is the mixed fluid flow rate, is the inner radius of the pipe string, is the viscosity of the mixed fluid.

[0091] In this embodiment, the method for calculating the friction coefficient based on the wellbore parameters includes:

[0092] exist When , the fluid is in laminar flow state; When , the fluid is in transitional flow state;

[0093] ;

[0094] Where f is the friction coefficient, is the Reynolds number, is the absolute roughness of the inner wall of the pipe string, is the inner radius of the pipe column.

[0095] In this embodiment, the method for calculating the friction coefficient based on the wellbore parameters also includes a method for calculating the friction coefficient when the Reynolds number is greater than or equal to 4000, specifically including:

[0096] When the Reynolds number is greater than or equal to 4000, it is a turbulent state; firstly, the first coefficient is calculated, and the method for calculating the first coefficient includes:

[0097] Assign weights to the Reynolds numbers at the observation points:

[0098] ;

[0099] in, is the weight of the Reynolds number in the observation point within the window, are the estimated points for the fitted polynomial, The Reynolds number of the ith observation point, is the bandwidth size;

[0100] Calculate fitting polynomial parameters according to the weights of the Reynolds numbers in the observation points;

[0101] ;

[0102] are the parameters of the polynomial; is a weight function based on and The weight is assigned by the distance; n is the number of observation points; is the friction coefficient corresponding to the Reynolds number of the ith observation point;

[0103] When M is the smallest, its corresponding , … is the fitting polynomial parameter; M is the adjustment parameter;

[0104] The first coefficient is calculated based on the fitting polynomial parameter, specifically:

[0105] ;

[0106] in, is the first coefficient, x is the Reynolds number for calculating the friction coefficient in the turbulent stage, is the estimated point of the fitted polynomial, p is the order of the fitted polynomial, , … To fit the polynomial parameters, The Reynolds number at the observation point, is the friction coefficient corresponding to the Reynolds number of the ith observation point

[0107] Then, a second coefficient is calculated. The method for calculating the second coefficient includes:

[0108] ;

[0109] in, is the second coefficient, is the Reynolds number when the fluid in the wellbore is in the laminar flow stage due to different fluid density and / or fluid velocity and / or fluid viscosity, Reynolds number The corresponding friction coefficient is is the Reynolds number when the fluid in the wellbore is in the turbulent stage due to different fluid density and / or fluid velocity and / or fluid viscosity, Reynolds number The corresponding friction coefficient is The Reynolds number corresponding to the fluid for which the friction coefficient needs to be calculated in the turbulent stage;

[0110] The friction coefficient is calculated based on the first coefficient and the second coefficient:

[0111] ;

[0112] Where y represents the Reynolds number corresponding to the fluid whose friction coefficient needs to be calculated in the turbulent stage x The corresponding friction coefficient.

[0113] The method for calculating the wellbore friction based on the Reynolds number and the friction coefficient is:

[0114] ;

[0115] in, The friction of the wellbore, is the density of the mixed fluid, is the mixed fluid flow rate, is the inner radius of the pipe string, is the friction coefficient, is the wellbore length.

[0116] Calculating bottom hole flow pressure according to the wellbore friction;

[0117] In this embodiment, the method for calculating the bottom hole flow pressure according to the wellbore friction is:

[0118] ;

[0119] in, is the bottom hole flowing pressure, is the wellhead pressure, is the acceleration due to gravity, is the density of the liquid, is the wellbore height, is the curvature radius of the deflection section, It is the friction of the wellbore.

[0120] Based on the historical production data and the bottom hole flow pressure, a BiLSTM neural network prediction model is trained to obtain a gas production prediction model; the bottom hole flow pressure and historical production data are brought into the gas production prediction model to obtain the corresponding gas production.

[0121] In this embodiment, the specific training method is: 8 feature data items, including bottom hole pressure, gas-liquid ratio, average gas production rate, average liquid production rate, casing pressure difference, daily gas production, daily liquid production and production time from the 1st day to the 50th day, are used as the input of the first sample, and the daily gas production on the 51st day is the output of the first sample, and the sliding window moves for one day. The 8 feature data items from the 2nd day to the 51st day are used as the input of the second sample, and the daily gas production on the 52nd day is the output of the second sample. Similarly, the sliding window is continuously performed until all production data are traversed.

[0122] After completing the model training with the sliding window data of this structure, the corresponding gas production prediction model can be obtained. The production data of the initial 50 days of production of the new well can be input into the model to complete the production prediction of the new well, and dynamic prediction can be performed continuously with the production time.

[0123] A gas production prediction model is obtained by training a BiLSTM neural network prediction model based on the historical production data and the bottom hole flow pressure. A physical model can be constructed by bringing the bottom hole flow pressure and historical production data into the gas production prediction model. By simulating the flow of gas, water and mixed fluids under different conditions, the temperature and pressure changes in different sections of the well are iteratively calculated, and the bottom hole flow pressure is calculated by iterating from the wellhead casing pressure to the bottom of the well from top to bottom.

[0124] In this embodiment, the production data and wellbore parameters of M natural gas wells are obtained and substituted into the physical model to calculate the bottom hole pressure corresponding to the daily production data, and combined with the historical production data; new production data features are constructed according to the existing production data, and a new production data sequence is constructed with the original production features; the combined historical production data are cleaned and preprocessed to obtain model input data, and the model input data is divided into a training data set, a verification data set and a test data set, and the model output is set to the daily gas production; a time series prediction model BiLSTM bidirectional long short-term memory neural network is constructed; the initial hyperparameters of the time series prediction model are obtained, and the initial hyperparameters are substituted into the hyperparameter optimization algorithm to obtain optimized hyperparameters; according to the optimized hyperparameters, the time series prediction model is trained to determine the final time series prediction model; the test data set is input into the final time series prediction model for model verification, and if the model accuracy does not meet the target requirements to obtain the corresponding gas production, the model training is performed again.

[0125] In this embodiment, the training data includes bottom hole flowing pressure, gas-liquid ratio, average gas production rate, average liquid production rate, oil-casing pressure difference, daily gas production, daily liquid production and production time, and the data types are shown in Table 1;

[0126] Table 1

[0127]

[0128] According to empirical formulas and relevant physical background, production data has the highest correlation with bottom hole pressure, but other characteristics also have an impact on production. However, there is no explicit formula to clearly define the contribution of each characteristic to production. The characteristics of the neural network model can effectively eliminate this impact.

[0129] The combined production data sequence is preprocessed to obtain a preprocessed production data sequence, wherein the preprocessing operation includes the following steps: deleting the production data whose production time, daily gas production, daily water production, oil pressure and casing pressure are 0; sorting the undeleted historical production data in chronological order of production date; and normalizing each production data;

[0130] The normalized historical production data is divided into a training data set, a validation data set, and a test data set according to a ratio of 6:2:2;

[0131] The BiLSTM bidirectional long short-term memory neural network settings for building a time series prediction model are shown in Table 2; the training options and parameter settings of the model are shown in Table 3;

[0132] Table 2

[0133]

[0134] Table 3

[0135]

[0136] BiLSTM performs both forward and reverse processing on the input data. This means that it not only uses the sequential information from front to back, but also uses the sequential information from back to front to capture more comprehensive contextual information. BiLSTM mainly achieves this by introducing memory units and gating mechanisms.

[0137] The gating mechanism mainly includes an input gate and a forget gate, and the relevant calculation formula is:

[0138] ;

[0139] ;

[0140] in, is the activation value of the input gate, is the weight matrix of the input gate, is the bias term, represents the output of the forget gate, is the first relevant parameter, is the second relevant parameter, express sigmoid function, is the hidden state at the previous moment, is the input at the current moment;

[0141] The model input data is characterized in that the normalized data is substituted into the sliding window algorithm to obtain the model input data. The sliding window algorithm is to perform the required operation on an array or string of a given specific window size. The input normalized data to be processed is a three-dimensional vector, and the three-dimensional vector structure is (sample number, feature number, window size). The output data is a two-dimensional vector, which can be (sample number, preset feature), where the sample number is the total sequence length of the data minus the value of the sliding window size, and the preset feature number refers to a specific feature. The feature number is the 8 features of the production data sequence obtained by the new combination, and the sample number and window size vary according to the specific project implementation settings.

[0142] The initial hyperparameters of the constructed neural network structure model are obtained, and the initial hyperparameters are substituted into the hyperparameter optimization algorithm, and finally the final time series prediction model is determined according to the optimized hyperparameters; the hyperparameter optimization algorithm is a Bayesian optimization algorithm, an iterative algorithm for global optimization problems; the core formula of Bayesian optimization includes the mean function and covariance function of the Gaussian process, as well as the confidence or expected improvement for selecting the next sampling point.

[0143] According to the optimization of hyperparameters from the neural network model and the training and learning of the final time series prediction model based on the training data set and the verification data set, the test data set will be input into the final time series prediction model to obtain the prediction results of the daily production data for the next M days, where M is a natural number greater than 0. The bidirectional long short-term memory unit layer added to the model extracts time series data from both directions during the model training process, so that the trained time series prediction model is more suitable for the prediction of long series, enhances the anti-interference and data extraction capabilities, and thus improves the prediction accuracy.

[0144] This application also provides an example of a method for predicting the dynamic production of a fractured gas well based on physical data drive according to the present invention, specifically:

[0145] The method for predicting the production of a fractured gas well based on a physical-data driven neural network provided in this example includes the following steps:

[0146] S1, obtain historical production data and wellbore parameters of the fractured gas well.

[0147] like Figure 2 This is a schematic diagram of the daily production data of the fracturing gas well in the embodiment of the present application, where the horizontal axis is the date and the vertical axis is the daily gas production, in units of m³ / day, reflecting the change in daily gas production of the fracturing gas well selected in the embodiment with the development time.

[0148] S2, substitute the historical production data and wellbore parameters into the constructed improved physical model to calculate the bottom hole flowing pressure corresponding to the daily production data.

[0149] S3, calculating new production data features based on the original production data features.

[0150] S4, combining the combined production data and bottom hole flow pressure data, and analyzing the correlation between them.

[0151] like Figure 3 As shown, in this embodiment, the principal component analysis method is used to analyze the daily production value of fracturing gas and other actual production-related data. The numbers in the figure reflect the degree of correlation between this data and the daily gas production data. It can be seen from the figure that the daily production of fracturing gas wells is related to the production data such as oil pressure, casing pressure, wellhead pressure, bottom hole flow pressure, gas-liquid ratio and production time in the historical production data that change with the development process. In this embodiment, daily water production, oil pressure, casing pressure, bottom hole flow pressure and production time data are selected as model data input features, and daily gas production is used as the target prediction value.

[0152] S5, clean and preprocess the historical production data to obtain model input data, and divide the model input data into a training data set, a verification data set, and a test data set.

[0153] Specifically, in an embodiment of the present application, historical production data is cleaned and preprocessed to obtain model input data, including:

[0154] Delete the data that does not meet the preset filtering conditions in the historical production data, and save the deleted data that does not meet the preset filtering conditions as data to be processed.

[0155] In this embodiment, the preset screening conditions may include whether the data is missing, whether the format and content are wrong. Accordingly, the data to be processed does not include historical production data with missing data, wrong format and wrong content.

[0156] S6, normalizing the data to be processed to obtain normalized data to be processed.

[0157] In this embodiment, the normalization process is proposed for the convenience of data processing, and the data is mapped to the range of 0 to 1 for processing. The formula used in the normalization process is:

[0158] ;

[0159] z 1 is the historical production data of the input fracturing gas well, The historical production data of the fractured gas wells is input x The normalized result is is the minimum value in the input data. is the maximum value in the input data.

[0160] For example, taking the daily oil production in the input data as an example, the input data is (15, 30, 45, 60, 75, 30, 30) cubic meters per day at 60 o'clock, and the normalized result value corresponding to 60 is:

[0161] ;

[0162] By analogy, the normalized result values ​​are (0, 0.25, 0.5, 0.75, 1, 0.25, 0.25, 0.25).

[0163] S7, substituting the normalized data to be processed into the sliding window algorithm to obtain the model input data, which is data in vector form.

[0164] In this embodiment, the sliding window algorithm is to perform the required operation on an array or string of a given specific window size. In this embodiment, the input normalized data to be processed can be a three-dimensional vector, the three-dimensional vector structure is (sample number, feature number, window size), and the output data is a two-dimensional vector, the two-dimensional vector can be (sample number, preset feature), where the sample number is the total sequence length of the data minus the value of the sliding window size, and the preset feature number refers to a specific feature. For example, in the embodiment of the present application, the preset feature can be the daily gas production.

[0165] Reference Figure 4 , Figure 4 This is a schematic diagram of the sliding window algorithm conversion in the embodiment of the present application. The specific conversion example of the sliding window algorithm in this implementation case is as follows:

[0166] The window size is n=50. A total of eight production characteristic data including bottom hole flowing pressure, gas-liquid ratio, average gas production rate, average liquid production rate, casing pressure difference, daily gas production, daily liquid production and production time from the 1st day to the 50th day are the input of the first sample. The daily gas production on the 51st day is the output of the first sample, and the sliding window moves for one day.

[0167] The eight production feature data from the 2nd day to the 51st day are the input of the second sample, and the daily gas production on the 52nd day is the output of the second sample. Similarly, the sliding window is continuously performed until all production data are traversed.

[0168] After the model training is completed with the sliding window data of this structure, the production data of the initial 50 days of production of the new well can be input into the model to complete the production forecast of the new well, and dynamic forecasts can be performed continuously over the production time.

[0169] The BiLSTM bidirectional long short-term memory neural network is constructed as a time series prediction model. The neural network structure diagram is as follows Figure 5 , Figure 6 shown.

[0170] The parameters that need to be optimized in this implementation case are: the number of layers and corresponding nodes of the bidirectional long short-term memory neural network, the maximum number of training times, the learning rate, and the learning batch size. The Bayesian optimization algorithm is used to optimize and solve the global optimal solution. The optimization process is as follows: Construct the initial network structure of the BiLSTM neural network prediction model, initialize the Bayesian optimization algorithm, perform fitness calculations, and update the positions of the discoverer and the follower in parallel, obtain the current updated position, and if the new position is better than the old position, update the old position to check the iteration end condition, and output the best fitness value. Substitute the output hyperparameters into the BiLSTM neural network prediction model to update the set values ​​and train. At the same time, use oil pressure, casing pressure, daily water production, bottom hole flow pressure, and production time as constraints, and finally output the daily gas production prediction results to obtain the optimized hyperparameters of the BiLSTM neural network prediction model; the learning rate is 0.0034, the number of long and short memory unit layers is 2, and the number of nodes is 512 and 1024 respectively. The optimal solution is used as the initial parameters based on the BiLSTM neural network. Then the test data is input into the prediction model to predict the daily gas production, and finally the prediction results are output. In order to evaluate the performance of the BiLSTM neural network prediction model, this implementation case adopts the following indicators for comprehensive evaluation:

[0171] Mean absolute percentage error (MAPE), whose value range is [0, +∞);

[0172] ;

[0173] in, Refers to the actual daily gas production value. is the value of daily gas production predicted by the model, N is the number of days for predicting daily gas production, and MAPE is the mean absolute percentage error. The smaller the value, the better the fitting effect.

[0174] In order to verify the convergence of the BiLSTM neural network prediction model, this implementation case uses the LOSS value of model training to compare the BiLSTM neural network prediction model and the LSTM model.

[0175] After the optimization and training process, the test data is input into the optimized model for prediction. In order to make the prediction method proposed in the present invention more convincing, a multi-layer perceptron (MLP), a LSTM neural network, a RNN neural network and a neural network based on an attention mechanism are used as comparative experiments. Figure 7It shows the comparison curve between the prediction results of different models and the actual values. It can be seen from the figure that MLP, LSTM and RNN are basically close to the real curve, but the prediction effect is not good at peaks and troughs; and the prediction effect of the BiLSTM neural network model is better than that of MLP, LSTM and RNN models, especially at peaks and troughs.

[0176] Although the various steps in the above embodiment are described in the above-mentioned order, those skilled in the art can understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple changes are within the scope of protection of the present invention.

[0177] A second embodiment of the present invention is a system for predicting the dynamic production of a fracturing gas well based on physical data drive, the system comprising:

[0178] A data acquisition module is used to obtain historical production data, wellbore parameters and gas production of gas wells;

[0179] A first calculation module is used to calculate the mixed fluid flow rate and the mixed fluid viscosity based on the historical production data and the wellbore parameters; and calculate the Reynolds number based on the mixed fluid flow rate and the mixed fluid viscosity;

[0180] A second calculation module is used to calculate the friction coefficient based on the wellbore parameters and the Reynolds number; calculate the wellbore friction based on the friction coefficient; and calculate the bottom hole flow pressure based on the wellbore friction;

[0181] The gas production prediction module is used to train the BiLSTM neural network prediction model based on the historical production data and the bottom hole flow pressure to obtain a gas production prediction model; the bottom hole flow pressure and historical production data are brought into the gas production prediction model to obtain the corresponding gas production.

[0182] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process and related instructions of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0183] It should be noted that the physical data-driven dynamic production prediction system for fractured gas wells provided in the above embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be decomposed or combined. For example, the modules in the above embodiment can be combined into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the modules or steps, and are not regarded as improper limitations of the present invention.

[0184] An electronic device according to the third embodiment of the present invention comprises: at least one processor; and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned physical data-driven method for dynamic production prediction of fractured gas wells.

[0185] A fourth embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the above-mentioned physical data-driven method for predicting the dynamic production of fractured gas wells.

[0186] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process and related instructions of the electronic device and storage medium described above can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0187] Those skilled in the art should be able to appreciate that the modules and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two, and the programs corresponding to the software modules and method steps can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the technical field. In order to clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are performed in electronic 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 to exceed the scope of the present invention.

[0188] Reference below Figure 8 , which shows a schematic diagram of the structure of a computer system of a server for implementing the method, system, and device embodiments of the present application. Figure 8 The server shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0189] like Figure 8As shown, the computer system includes a central processing unit 601, which can perform various appropriate actions and processes according to the program stored in the read-only memory 602 or the program loaded from the storage part 608 to the random access memory 603. Various programs and data required for system operation are also stored in the random access memory 603. The central processing unit 601, the read-only memory 602 and the random access memory 603 are connected to each other through a bus 604. An input / output interface 605 (I / O interface) is also connected to the bus 604.

[0190] The following components are connected to the input / output interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed so that a computer program read therefrom is installed into the storage section 608 as needed.

[0191] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit 601, the above-mentioned functions defined in the method of the present application are executed. It should be noted that the above-mentioned computer-readable medium of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, an apparatus or a device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, an apparatus or a device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0192] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0193] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0194] The terms "first", "second", etc. are used to distinguish similar objects rather than to describe or indicate a particular order or sequence.

[0195] The term "comprise" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that includes a list of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article, or apparatus / device.

[0196] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A method for predicting the dynamic production of a fractured gas well based on physical data drive, characterized in that: The method comprises: Obtain historical production data, wellbore parameters and gas production of gas wells; Calculate the mixed fluid flow rate and the mixed fluid viscosity based on the historical production data and the wellbore parameters; calculate the Reynolds number based on the mixed fluid flow rate and the mixed fluid viscosity; Calculating the friction coefficient based on the wellbore parameters and the Reynolds number; calculating the wellbore friction based on the friction coefficient; calculating the bottom hole flow pressure based on the wellbore friction; The method for calculating the friction coefficient based on the wellbore parameters also includes a method for calculating the friction coefficient when the Reynolds number is greater than or equal to 4000, specifically including: When the Reynolds number is greater than or equal to 4000, it is a turbulent state; firstly, the first coefficient is calculated, and the method for calculating the first coefficient includes: Assign weights to the Reynolds numbers at the observation points: ； in, For the window i The weight of the Reynolds number of each observation point, are the estimated points for the fitted polynomial, No. i The Reynolds number of the observation point, is the bandwidth size; Calculate fitting polynomial parameters according to the weights of the Reynolds numbers in the observation points; ； are the parameters of the polynomial; is a weight function based on and The distance is used to assign weights; n is the number of observation points; p is the order of the fitting polynomial; It is i The friction coefficient corresponding to the Reynolds number of each observation point; When M is the smallest, its corresponding , … is the fitting polynomial parameter; M is the adjustment parameter; The first coefficient is calculated based on the fitting polynomial parameter, specifically: ； in, is the first coefficient, x is the Reynolds number for calculating the friction coefficient in the turbulent stage, are the estimated points for the fitted polynomial, , … is the fitting polynomial parameter; Then, a second coefficient is calculated. The method for calculating the second coefficient includes: ； in, is the second coefficient, is the Reynolds number when the fluid in the wellbore is in the laminar flow stage due to different fluid density and / or fluid velocity and / or fluid viscosity, Reynolds number The corresponding friction coefficient is is the Reynolds number when the fluid in the wellbore is in the turbulent stage due to different fluid density and / or fluid velocity and / or fluid viscosity, Reynolds number The corresponding friction coefficient is The Reynolds number corresponding to the fluid for which the friction coefficient needs to be calculated in the turbulent stage; The friction coefficient is calculated based on the first coefficient and the second coefficient: ； Where y represents the Reynolds number corresponding to the fluid whose friction coefficient needs to be calculated in the turbulent stage x The corresponding friction coefficient; Based on the historical production data, the bottom hole flow pressure and the gas production, a BiLSTM neural network prediction model is trained to obtain a gas production prediction model; the bottom hole flow pressure and historical production data of the past time are brought into the gas production prediction model to obtain the gas production in the future time.

2. The method for predicting the dynamic production of a fracturing gas well based on physical data drive according to claim 1, characterized in that: The method for calculating the flow rate of the mixed fluid is: ； in, is the mixed fluid flow rate, is the current daily gas production, is the current daily water production, is the inner radius of the pipe column.

3. The method for predicting the dynamic production of a fractured gas well based on physical data drive according to claim 2, characterized in that: The method to calculate the viscosity of the mixed fluid is: ； in, is the viscosity of the mixed fluid, is the liquid viscosity, is the gas viscosity, is the gas-liquid volume ratio.

4. The method for predicting the dynamic production of a fractured gas well based on physical data drive according to claim 3, characterized in that: The method to calculate the Reynolds number is: ； in, is the Reynolds number, is the density of the mixed fluid, is the mixed fluid flow rate, is the inner radius of the pipe string, is the viscosity of the mixed fluid.

5. The method for predicting the dynamic production of a fractured gas well based on physical data drive according to claim 4, characterized in that: The method for calculating the friction coefficient based on the wellbore parameters includes: exist When , the fluid is in laminar flow state; When , the fluid is in transitional flow state; ； in, f is the friction coefficient, is the Reynolds number, is the absolute roughness of the inner wall of the pipe string, is the inner radius of the pipe column.

6. The method for predicting the dynamic production of a fractured gas well based on physical data drive according to claim 5, characterized in that: The method for calculating the wellbore friction based on the Reynolds number and the friction coefficient is: ； in, The friction of the wellbore, is the density of the mixed fluid, is the mixed fluid flow rate, is the inner radius of the pipe string, is the friction coefficient, is the wellbore length.

7. The method for predicting the dynamic production of a fractured gas well based on physical data drive according to claim 6, characterized in that: The method for calculating the bottom hole flow pressure according to the wellbore friction is: ； in, is the bottom hole flowing pressure, is the wellhead pressure, is the acceleration due to gravity, is the density of the liquid, is the wellbore height, is the curvature radius of the deflection section, It is the friction of the wellbore.

8. A system for predicting the dynamic production of a fractured gas well based on physical data drive, which is based on the method for predicting the dynamic production of a fractured gas well based on physical data drive as described in any one of claims 1 to 7, and is characterized in that: The system comprises: A data acquisition module is used to obtain historical production data, wellbore parameters and gas production of gas wells; A first calculation module is used to calculate the mixed fluid flow rate and the mixed fluid viscosity based on the historical production data and the wellbore parameters; and calculate the Reynolds number based on the mixed fluid flow rate and the mixed fluid viscosity; A second calculation module is used to calculate the friction coefficient based on the wellbore parameters and the Reynolds number; calculate the wellbore friction based on the friction coefficient; and calculate the bottom hole flow pressure based on the wellbore friction; The gas production prediction module is used to train the BiLSTM neural network prediction model based on the historical production data and the bottom hole flow pressure to obtain a gas production prediction model; the bottom hole flow pressure and historical production data are brought into the gas production prediction model to obtain the corresponding gas production.

9. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to implement the physical data-driven dynamic production prediction method for fractured gas wells as described in any one of claims 1-7.

Images