Formation pressure prediction model training method, prediction method and related device

Abstract

The invention discloses a formation pressure prediction model training method, a prediction method and a related device. The method comprises the following steps: determining corresponding historical hydrostatic pressure and historical interval transit time of normal compaction according to each group of well drilling and logging data of different well depths of each drilled adjacent well obtained in advance; determining a group of corresponding historical input parameters according to the correlation strength between each group of drilling data and logging data and the formation pressure profile; according to the historical interval transit time and historical hydrostatic pressure of each group of normal compaction and historical input parameters, determining a corresponding historical input vector and a corresponding stratum pressure actual value; inputting each group of historical input vectors into a pre-constructed initial formation pressure prediction model, and determining a corresponding formation pressure prediction value; if the error between each group of formation pressure prediction values and the corresponding formation pressure actual value is smaller than a preset range, obtaining a formation pressure prediction model; otherwise, multiple groups of historical input vectors are divided again, and the model training process is executed again.

Description

Technical Field

[0001] The invention relates to a formation pressure prediction model training method, a prediction method and related devices. Background Art

[0002] With the continuous deepening of oil and gas exploration and development, the focus of oil and gas exploration and development is developing towards deep and ultra-deep layers. However, deep formations often have the characteristics of high temperature, complex formation pressure system and narrow safe drilling fluid density window. If the drilling process measures are not appropriate, complex downhole situations such as overflow and gas invasion are very likely to occur. Therefore, accurate prediction of formation pressure is the basis for reasonable design of drilling fluid density, and it is also the key technology to achieve safe drilling in deep, high-temperature and high-pressure formations.

[0003] Traditional formation pressure prediction is based on empirical formulas. This method of formation pressure prediction has problems such as low prediction accuracy, low compliance and poor reliability under complex formation conditions. In addition, since its empirical parameters are mostly given manually, there is a strong subjectivity, which leads to its low applicability and large prediction error, resulting in unreasonable drilling fluid density design, which in turn greatly increases the risks of overflow and well leakage.

[0004] In recent years, the advantages of artificial intelligence technology in the field of oil and gas exploration and development have gradually become apparent, and it has been widely used in operating condition diagnosis, parameter optimization and other aspects. Summary of the invention

[0005] In order to better predict the formation pressure in the drilling area, the embodiments of the present application provide a formation pressure prediction model training method, a prediction method and related devices.

[0006] In a first aspect, an embodiment of the present application provides a formation pressure prediction model training method, the method comprising:

[0007] Determine each set of historical hydrostatic pressure and historical acoustic time difference of normal compaction at different depths of each completed adjacent well according to each set of drilling and logging data at different depths of each completed adjacent well obtained in advance;

[0008] Determine a corresponding set of historical input parameters according to the strength of correlation between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance;

[0009] According to the formation pressure profile of each completed adjacent well and each group of the normal compaction historical acoustic time difference and historical hydrostatic pressure and the historical input parameters, batch division is performed to determine multiple groups of historical input vectors and corresponding actual values ​​of formation pressure;

[0010] Inputting the plurality of groups of historical input vectors into a pre-built initial formation pressure prediction model to determine a formation pressure prediction value corresponding to each group of historical input vectors;

[0011] Determine whether the error between the predicted formation pressure value corresponding to each set of historical input vectors and the corresponding actual formation pressure value is less than a preset range;

[0012] If yes, the formation pressure prediction model is obtained;

[0013] If not, then the plurality of groups of historical input vectors and the corresponding actual values ​​of formation pressure are re-divided, and the model training process is re-executed on the plurality of groups of historical input vectors after being re-divided.

[0014] In one or some optional implementations of the embodiment of the present application, the determination of a corresponding set of historical input parameters based on the strength of correlation between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance includes:

[0015] Preprocessing each set of drilling and logging data and formation pressure profiles of each completed adjacent well obtained in advance to obtain each set of drilling and logging data and formation pressure profiles of each completed adjacent well after preprocessing;

[0016] Determine the correlation between each parameter in each set of drilling and logging data and the formation pressure profile according to each set of drilling and logging data and the formation pressure profile of each completed adjacent well after the preprocessing;

[0017] Parameters having a correlation with the formation pressure profile greater than a preset threshold are selected from each set of drilling and logging data as a corresponding set of historical input parameters.

[0018] In one or some optional implementations of the embodiment of the present application, the initial formation pressure prediction model includes a BP neural network, a LSTM neural network and a connection layer;

[0019] The step of inputting the plurality of groups of historical input vectors into a pre-built initial formation pressure prediction model to determine a formation pressure prediction value corresponding to each group of historical input vectors includes:

[0020] Using the non-time series data in each group of the historical input vectors as the BP neural network input parameters to obtain the output parameters of the BP neural network;

[0021] The time series data in each group of the historical input vectors are used as the input parameters of the LSTM neural network and the output parameters of the LSTM neural network; wherein the non-time series data in the historical input vectors are the historical input parameters corresponding to each discrete data type, and the time series data in the historical input vectors are the historical input parameters corresponding to the historical hydrostatic pressure, the historical acoustic time difference of normal compaction, and each continuous data type;

[0022] The output parameters of the BP neural network and the output parameters of the LSTM neural network are fused in parallel through the connection layer to obtain the formation pressure prediction value corresponding to each group of the historical input vectors.

[0023] In one or some optional implementations of the embodiments of the present application, based on the pre-obtained drilling and logging data of each group of completed adjacent wells, the historical acoustic time differences of each group of normal compaction at different well depths of each completed adjacent well are determined in the following manner:

[0024] Based on the pre-obtained drilling and logging data of each completed adjacent well, determine the variation of acoustic time difference with vertical depth under each normal compaction condition;

[0025] According to the variation law of acoustic time difference with vertical depth under each normal compaction condition, the normal compaction trend line is fitted, and the historical acoustic time difference of each group of normal compaction at different well depths of each completed adjacent well is determined based on the following formula 1:

[0026] lnΔt=lnΔt0-CH Formula 1;

[0027] Where Δt is the acoustic time difference corresponding to a certain depth point in the completed adjacent well; Δt0 is the surface acoustic time difference; C is the compaction coefficient; and H is the depth of the completed adjacent well.

[0028] In one or some optional implementations of the embodiments of the present application, the historical hydrostatic pressures of each group of different well depths of each completed adjacent well are determined in the following manner based on the pre-obtained drilling and logging data of each group of different well depths of each completed adjacent well:

[0029] According to the depth of each completed adjacent well, the historical hydrostatic pressures of each group at different depths of each completed adjacent well are determined based on the following formula 2:

[0030] P h =ρgh Formula 2;

[0031] Where P h is the historical hydrostatic pressure; ρ is the density of formation water; g is the local gravitational acceleration; h is the depth of the completed adjacent well.

[0032] In a second aspect, an embodiment of the present application provides a formation pressure prediction method, the method comprising:

[0033] Determine the sonic transit time of hydrostatic pressure and normal compaction based on previously acquired drilling and logging data in the wellbore;

[0034] Determining input parameters based on the strength of correlation between the formation pressure obtained in advance during drilling and the drilling data and the logging data;

[0035] Determining an input vector according to the acoustic wave time difference and hydrostatic pressure of the normal compaction and the input parameters;

[0036] The input vector is input into the obtained formation pressure prediction model to determine the predicted value of the formation pressure during drilling.

[0037] In a third aspect, an embodiment of the present application provides a formation pressure prediction model training device, the device comprising:

[0038] The first determination module is used to determine each group of historical hydrostatic pressure and historical acoustic time difference of normal compaction of each completed adjacent well at different depths according to each group of drilling and logging data of each completed adjacent well at different depths obtained in advance;

[0039] The second determination module is used to determine a corresponding set of historical input parameters according to the correlation strength between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance;

[0040] The third determination module is used to divide the batches according to the formation pressure profile of each completed adjacent well and the historical acoustic time difference and historical hydrostatic pressure of each group of normal compaction and the historical input parameters, and determine multiple groups of historical input vectors and corresponding actual values ​​of formation pressure;

[0041] A first prediction module, used for inputting the plurality of groups of historical input vectors into a pre-built initial formation pressure prediction model, and determining a formation pressure prediction value corresponding to each group of historical input vectors;

[0042] An error judgment module is used to judge whether the error between the predicted formation pressure value corresponding to each set of historical input vectors and the corresponding actual formation pressure value is less than a preset range;

[0043] A model determination module is used to obtain a formation pressure prediction model;

[0044] The updating module is used to re-divide the multiple groups of historical input vectors and the corresponding actual values ​​of formation pressure, and re-execute the above model training process on the re-divided multiple groups of historical input vectors.

[0045] In a fourth aspect, an embodiment of the present application provides a formation pressure prediction device, the device comprising:

[0046] A fourth determination module is used to determine the acoustic wave time difference and hydrostatic pressure of normal compaction based on the drilling and logging data obtained in advance during drilling;

[0047] a fifth determination module, for determining input parameters according to the strength of correlation between the formation pressure in the well and the drilling and logging data obtained in advance;

[0048] a sixth determination module, configured to determine an input vector according to the acoustic time difference and hydrostatic pressure of the normal compaction and the input parameters;

[0049] The second prediction module is used to input the input vector into a pre-built formation pressure prediction model to determine the predicted value of the formation pressure during drilling.

[0050] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned formation pressure prediction model training method and / or formation pressure prediction method.

[0051] In a sixth aspect, an embodiment of the present application 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, it implements the formation pressure prediction model training method and / or the formation pressure prediction method as described above.

[0052] In a seventh aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on a computer device, enables the computer device to execute the formation pressure prediction model training method and / or the formation pressure prediction method as described above.

[0053] In an eighth aspect, an embodiment of the present application provides a chip, the chip including a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement the formation pressure prediction model training method as described above, and / or the formation pressure prediction method.

[0054] The beneficial effects of the above technical solution provided by the embodiment of the present application include at least:

[0055] The formation pressure prediction model training method provided in the embodiment of the present application analyzes the historical drilling and logging data and formation pressure profiles of completed adjacent wells, calculates the normal compaction acoustic wave time difference and hydrostatic pressure parameters, and integrates traditional mechanism experience knowledge into the neural network input and structure, thereby improving the robustness, stability and prediction accuracy of the prediction model, making the formation pressure prediction model have a certain degree of interpretability, and effectively improving the stability and generalization ability of the formation pressure prediction model; determining the input parameters related to the formation pressure can reduce the dimension of the input data, reduce the input of irrelevant data, and improve the training efficiency of the formation pressure prediction model; using the prediction model obtained by this method to predict the formation pressure can ensure the stability and reliability of the prediction results, which is conducive to shortening the drilling cycle, reducing drilling costs, improving drilling quality and drilling efficiency, and enhancing the safety and controllability of drilling, and ultimately achieving the goal of ensuring drilling safety and reducing costs and increasing efficiency.

[0056] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0057] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0059] Figure 1 A schematic diagram of the steps of the formation pressure prediction model training method provided in an embodiment of the present application;

[0060] Figure 2 A schematic diagram of the structure of the initial formation pressure prediction model provided in the embodiment of the present application;

[0061] Figure 3 A schematic diagram of the flow of formation pressure prediction model training and prediction provided in an embodiment of the present application;

[0062] Figure 4 A schematic diagram of the steps of the formation pressure prediction method provided in an embodiment of the present application;

[0063] Figure 5 A schematic diagram of the structure of a formation pressure prediction model training device provided in an embodiment of the present application;

[0064] Figure 6 A schematic diagram of the structure of a formation pressure prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0065] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0066] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0067] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0068] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0069] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0070] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0071] It should be understood that the size of the serial numbers of the steps in the following embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0072] In order to illustrate the technical solution of the present application, a specific embodiment is provided below for illustration.

[0073] The inventors found that in the prior art, traditional formation pressure prediction is calculated based on empirical formulas. This type of formation pressure prediction method has problems such as low prediction accuracy, low compliance rate and poor reliability under complex formation conditions. In addition, since its empirical parameters are mostly given manually, there is a strong subjectivity, which leads to its low applicability and large prediction error, resulting in unreasonable drilling fluid density design, and further greatly increased risks of overflow, well leakage, etc.

[0074] In recent years, the advantages of artificial intelligence technology in the field of oil and gas exploration and development have gradually become prominent, and it has been widely used in working condition diagnosis, parameter optimization and other aspects. The advantages of artificial intelligence technology in the field of oil and gas exploration and development have gradually become prominent, and it has been widely used in working condition diagnosis, parameter optimization and other aspects, and has achieved good results. However, when applied to formation pressure prediction, due to the lack of mechanism models and the constraints of traditional empirical knowledge, it still lacks stability and reliability. Based on this, the inventor has made the present invention after further research and development, providing a formation pressure prediction model training method, prediction method and related devices.

[0075] Embodiment 1

[0076] The present application embodiment provides a formation pressure prediction model training method, referring to Figure 1 As shown, the method includes:

[0077] S101: Determine each set of historical hydrostatic pressure and historical acoustic time difference of normal compaction of each completed adjacent well at different depths according to each set of drilling and logging data of each completed adjacent well at different depths obtained in advance.

[0078] In the above step S101, based on the pre-obtained drilling and logging data of each group of completed adjacent wells, the historical acoustic time differences of each group of normal compaction at different well depths of each completed adjacent well are determined in the following manner:

[0079] Based on the pre-obtained drilling and logging data of each completed adjacent well, determine the variation of acoustic time difference with vertical depth under each normal compaction condition;

[0080] According to the variation law of acoustic time difference with vertical depth under each normal compaction condition, the normal compaction trend line is fitted, and the historical acoustic time difference of each group of normal compaction at different well depths of each completed adjacent well is determined based on the following formula 1:

[0081] lnΔt=lnΔt0-CH Formula 1;

[0082] Where Δt is the acoustic time difference corresponding to a certain depth point in the completed adjacent well; Δt0 is the surface acoustic time difference; C is the compaction coefficient; and H is the depth of the completed adjacent well.

[0083] In the above step S101, the historical hydrostatic pressures of each group of wells at different depths of each completed adjacent well are determined in the following manner based on the previously obtained drilling and logging data of each group of wells at different depths of each completed adjacent well:

[0084] According to the depth of each completed adjacent well, the historical hydrostatic pressures of each group at different depths of each completed adjacent well are determined based on the following formula 2:

[0085] P h =ρgh Formula 2;

[0086] Where P h is the historical hydrostatic pressure; ρ is the density of formation water; g is the local gravitational acceleration; h is the depth of the completed adjacent well.

[0087] In the embodiment of the present application, it is necessary to collect a large number of drilling and logging data and formation pressure profiles of completed adjacent wells, select a plurality of appropriate completed adjacent wells according to the actual situation, collect drilling and logging data at different well depths of each completed adjacent well, and obtain multiple sets of drilling and logging data and formation pressure profiles of each completed adjacent well to construct multiple sets of sample data. Among them, the completed adjacent well refers to the adjacent well that has completed drilling work in the research block where the drilling is located, and the division range of the completed adjacent wells in the drilling can be determined according to the area size of the research block where the drilling is located. Before constructing the input vector of the formation pressure prediction model, it is necessary to determine the historical hydrostatic pressure and the historical acoustic time difference of normal compaction corresponding to each completed adjacent well at each depth according to the drilling and logging data at different depths of each completed adjacent well, and obtain multiple sets of historical hydrostatic pressure and historical acoustic time difference of normal compaction corresponding to each completed adjacent well. By calculating the normal compaction acoustic time difference and hydrostatic pressure parameters, the empirical knowledge is integrated into the input vector of the formation pressure prediction model, which can improve the robustness of the model.

[0088] S102: Determine a corresponding set of historical input parameters according to the strength of correlation between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance.

[0089] In the above step S102, the determination of a corresponding set of historical input parameters based on the correlation strength between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance includes:

[0090] Preprocessing each set of drilling and logging data and formation pressure profiles of each completed adjacent well obtained in advance to obtain each set of drilling and logging data and formation pressure profiles of each completed adjacent well after preprocessing;

[0091] Determine the correlation between each parameter in each set of drilling and logging data and the formation pressure profile according to each set of drilling and logging data and the formation pressure profile of each completed adjacent well after the preprocessing;

[0092] Parameters having a correlation with the formation pressure profile greater than a preset threshold are selected from each set of drilling and logging data as a corresponding set of historical input parameters.

[0093] In an embodiment of the present application, after obtaining multiple groups of sample data, i.e., multiple groups of drilling data and logging data and formation pressure profiles of completed adjacent wells, preprocessing is required first. The preprocessing method can be outlier and missing value processing. The specific processing method can refer to the detailed description in the prior art and is not specifically limited here. After preprocessing, it is necessary to calculate the correlation between each parameter and the formation pressure profile in each group of drilling and logging data of each completed adjacent well after preprocessing. The correlation coefficient used to judge the correlation is mainly used to analyze and measure the correlation between each variable and the target variable, and can be calculated by the Pearson correlation analysis method. Among them, the Pearson correlation coefficient calculation formula is as shown in the following formula 3:

[0094]

[0095] In the formula, ρ XY is the Pearson correlation coefficient of variables X and Y; Cov(X,Y) is the covariance of variables X and Y; D(X) is the variance of variable X; D(Y) is the variance of variable Y.

[0096] After calculating the correlation between each parameter and the formation pressure profile in each group of drilling and logging data of each completed adjacent well after preprocessing through the above formula 3, the parameters with correlation greater than the preset threshold are screened out as the input parameters. Among them, the preset threshold is set according to the actual situation, for example, it can be set to 0.3, and the correlation coefficient greater than 0 and less than 0.3 can be regarded as no correlation; greater than 0.3 and less than 0.5 is considered to have a low correlation; greater than 0.5 and less than 0.8 is considered to have a moderate correlation; greater than 0.8 and less than 1.0 is considered to be highly correlated. Determining the input parameters related to formation pressure can reduce the dimension of input data, reduce the input of irrelevant data, and improve the training efficiency of the formation pressure prediction model.

[0097] In the embodiment of the present application, after calculating the correlation between each parameter and the formation pressure profile in each group of drilling and logging data of each completed adjacent well after pre-processing, the change law of each parameter and the formation pressure profile can also be analyzed and screened. For example, the following change law can be analyzed: where the formation pressure is high, the acoustic time difference of the formation will be large, the resistivity will be high, the porosity will be larger, and the density will be lower. In addition, the drilling speed will be higher and the hook load will be lower. The formation with low formation pressure is opposite to the above law. Analyzing the change law of each parameter and the formation pressure profile can be used to discover some abnormally high-pressure or low-pressure formations.

[0098] In the embodiment of the present application, after determining the historical input parameters, it is necessary to analyze the sequence and continuity of each historical input parameter. According to the analysis results, each historical input parameter can be divided into discrete data and continuous data, among which discrete data such as lithology, layer, formation lithology, formation layer and drilling fluid type are used as non-time series data; continuous data such as acoustic wave time difference, depth, DC index, drilling speed, acoustic logging density logging, resistivity logging, gamma logging, natural potential logging, caliper logging, neutron logging, well depth, drilling pressure, drilling speed, rotation speed, standpipe pressure, pump displacement, hook load, drilling fluid density, inlet and outlet density, inlet and outlet conductivity, inlet and outlet temperature and inlet and outlet flow are used as time series data, and the historical hydrostatic pressure and normal compaction historical acoustic wave time difference obtained in step S101 are also time series data.

[0099] S103: Divide the formation pressure profiles of the completed adjacent wells into batches according to the historical acoustic time difference and historical hydrostatic pressure of each group of normal compaction and the historical input parameters, and determine multiple groups of historical input vectors and corresponding actual formation pressure values.

[0100] In the embodiment of the present application, according to the correlation strength between each group of drilling and logging data of each completed adjacent well and the formation pressure profile, the historical input parameters of the formation pressure prediction model are optimized through step S102, and then the historical input parameters of each completed adjacent well are combined with the historical acoustic time difference and historical hydrostatic pressure of normal compaction at different depths, and then combined with the actual values ​​of formation pressure corresponding to different depths in the formation pressure profile, multiple groups of data sets are obtained, and the multiple groups of data sets are batch-divided to construct historical input vectors. Among them, the batch division method can be by generating a column of random numbers of fixed length, for each random number, obtaining a data set indexed by the random number and dividing it into a batch (i.e., a group of historical input vectors and the actual value of formation pressure corresponding to the group of historical input vectors), thereby generating multiple batches, i.e., determining multiple groups of historical input vectors and corresponding actual values ​​of formation pressure. Among them, since the neural network calculation process is a matrix operation, the parameter input in the neural network is in the form of a vector, so when constructing the parameters for input model training, the form is an input vector.

[0101] S104: Inputting the plurality of groups of historical input vectors into a pre-built initial formation pressure prediction model, and determining a formation pressure prediction value corresponding to each group of historical input vectors.

[0102] In the present application, refer to Figure 2 As shown, the initial formation pressure prediction model includes a BP neural network, a LSTM neural network and a connection layer;

[0103] In the above step S104, the inputting of the plurality of groups of historical input vectors into the pre-built initial formation pressure prediction model to determine the formation pressure prediction value corresponding to each group of historical input vectors specifically includes:

[0104] Using the non-time series data in each group of the historical input vectors as the BP neural network input parameters to obtain the output parameters of the BP neural network;

[0105] The time series data in each group of the historical input vectors are used as the input parameters of the LSTM neural network and the output parameters of the LSTM neural network; wherein the non-time series data in the historical input vectors are the historical input parameters corresponding to each discrete data type, and the time series data in the historical input vectors are the historical input parameters corresponding to the historical hydrostatic pressure, the historical acoustic time difference of normal compaction, and each continuous data type;

[0106] The output parameters of the BP neural network and the output parameters of the LSTM neural network are fused in parallel through the connection layer to obtain the formation pressure prediction value corresponding to each group of the historical input vectors.

[0107] In the embodiment of the present application, after obtaining multiple groups of historical input vectors, each group of historical input vectors is input into the initial formation pressure prediction model for training to obtain the formation pressure prediction value corresponding to each group of historical input vectors. The initial formation pressure prediction model is obtained by parallel fusion of the BP neural network and the LSTM neural network through the connection layer, and the BP neural network and the LSTM neural network are selected to receive non-sequential and sequential data respectively.

[0108] For each set of input vectors, the preprocessed non-time series data needs to be used as the BP neural network input parameter, and the preprocessed time series data needs to be used as the LSTM neural network input parameter. The preprocessing method for non-time series data can be 0-1 unique hot encoding; the preprocessing method for time series data can be normalization.

[0109] In the embodiment of the present application, two non-time series data, lithology and stratigraphic position, are taken as examples and one-hot encoding is performed in the following manner:

[0110] The row vector is set based on the number of lithology categories. Assuming that the number of lithology categories is n, an n-dimensional row vector is set. Each column in the row vector represents a type of lithology. When the lithology appears, the column representing the lithology is recorded as 1, and the remaining columns are recorded as 0. In this way, the lithology is encoded and converted into a 0-1 code that can be recognized by the computer.

[0111] The row vector is set based on the number of layer categories. Assuming the number of layer categories is m, an m-dimensional row vector is set. Each column in the row vector represents a layer. When the layer appears, the column representing the layer is recorded as 1, and the remaining columns are recorded as 0. In this way, the layers are encoded and converted into 0-1 codes that can be recognized by computers.

[0112] In an embodiment of the present application, the inventors considered the time series characteristics of each historical input parameter when constructing the model, selected a BP neural network based on the non-time series data in the historical input parameters, selected an LSTM neural network based on the time series data in the historical input parameters, and constructed an initial formation pressure prediction model based on the BP neural network and the LSTM neural network.

[0113] Since the LSTM time-series neural network is not suitable for processing discrete data such as lithology and formation, but is suitable for processing sequential data such as drilling and logging; the BP neural network can process discrete data such as formation and lithology, but cannot extract sequence information from sequential data. Therefore, the initial formation prediction model is obtained by constructing the BP neural network and the LSTM neural network, and then fusing the BP neural network and the LSTM neural network in parallel through the connection layer. The neural network structure is reasonably designed according to the time-series of the data, which is more in line with the formation pressure calculation process.

[0114] When each set of historical input vectors is input into the initial formation pressure prediction model for training, since the historical input vectors are constructed by combining the input parameters with the historical acoustic time difference and historical hydrostatic pressure of normal compaction, when the initial formation pressure prediction model formed by the parallel fusion of the BP neural network and the LSTM sequential neural network is input, the non-time series data in the historical input vectors, such as lithology and horizon, will be input through the BP neural network; the time series data in the historical input vectors, such as depth, dc index, and the historical acoustic time difference and historical hydrostatic pressure of normal compaction, will be input through the LSTM neural network, and finally the corresponding formation pressure prediction value will be output through an output layer composed of neurons. By integrating empirical knowledge into the neural network input and structure, the formation pressure prediction model has a certain degree of interpretability, which effectively improves the stability and generalization ability of the formation pressure prediction model.

[0115] In the embodiment of the present application, the structure of the initial formation pressure prediction model can be adjusted according to the actual situation, for example Figure 2 The figure shows a schematic diagram of the structure of an initial formation pressure prediction model. The upper part of the figure is a BP neural network with non-sequential input, and the lower part is a LSTM neural network with sequential input. The BP neural network and the LSTM neural network are fused in parallel by a neuron layer, and finally a neuron is set as the output layer to output the formation pressure prediction value. Among them, the number of layers of the BP neural network and the LSTM neural network is set according to the actual situation.

[0116] In addition, during the training process, it is necessary to construct an error loss function based on the sample data and the initial formation pressure prediction model, and use the back propagation algorithm to solve the parameters in the initial formation pressure prediction model based on the error loss function.

[0117] The error loss function may be the mean square error (MSE), which can be used to measure the deviation between the predicted value and the true value. The mean square error calculation formula is as follows:

[0118]

[0119] Where, MSE is the mean square error; true True value of the sample; y predict is the sample prediction value; N is the number of samples.

[0120] S105: Determine whether the error between the predicted formation pressure value of each set of historical input vectors and the corresponding pre-obtained actual formation pressure value is less than a preset range. If so, execute step S106; if not, execute step S107.

[0121] In the embodiment of the present application, after obtaining the predicted value of each formation pressure through step S104, it is necessary to compare each predicted formation pressure value with the corresponding actual formation pressure value to determine whether the error between the two is less than a preset range. If so, step S106 is executed. At this time, the initial formation pressure prediction model has been trained and the formation pressure prediction model is obtained; if not, step S107 is executed, and the historical input vector needs to be updated, that is, the batches of multiple groups of historical input vectors are re-divided, and the above-mentioned model training process is re-executed on the historical input vectors of each batch after re-division.

[0122] S106: Obtain a formation pressure prediction model.

[0123] S107: Re-dividing the plurality of groups of historical input vectors and the corresponding actual values ​​of formation pressure, and re-executing the above model training process with the re-divided plurality of groups of historical input vectors.

[0124] In an embodiment of the present application, when the error between the predicted formation pressure value and the corresponding actual formation pressure value is greater than a preset range, it is necessary to re-divide multiple groups of historical input vectors, and a column of random numbers of fixed length can be regenerated during the re-division, thereby determining multiple groups of re-divided historical input vectors and corresponding actual formation pressure values. The historical input vectors after the re-division batches are then input into the trained initial formation pressure prediction model for re-training. The above model training process is repeated until the error between the trained formation pressure prediction value and the corresponding actual formation pressure value is less than a preset range, and / or the error between the trained formation pressure prediction value and the corresponding actual formation pressure value tends to be stable and unchanged, then the model training is completed, and a trained formation pressure prediction model is obtained.

[0125] In the present application embodiment, Figure 3 As shown in FIG. 1 , the training and prediction process of the formation pressure prediction model is shown in the figure. The training process of the formation pressure prediction model is shown on the left side. By obtaining multiple sets of drilling and logging data and formation pressure data of multiple completed adjacent wells, the historical acoustic wave time difference of normal compaction is obtained by fitting the normal compaction trend line, and the historical hydrostatic pressure value is calculated according to the empirical formula. Then, according to the correlation strength between each set of drilling and logging data of each completed adjacent well and the formation pressure profile, the historical input parameters of the formation pressure prediction model are optimized. Then, the historical input parameters are combined with the obtained historical acoustic wave time difference and historical hydrostatic pressure of normal compaction to construct the historical input vector. Finally, , input the historical input vector into the initial formation pressure prediction model for training, and judge whether the model training is completed based on the error between the predicted formation pressure value and the actual formation pressure value. If the training is completed, update the historical input vector and re-train; the right side of the figure shows the prediction process of the formation pressure prediction model. Before predicting the formation pressure at a certain depth of the well, it is necessary to construct the input vector in the same way as the historical input vector, and then input the input vector into the trained formation pressure prediction model to predict the bottom layer pressure prediction value of the well. Multiple predictions at different depths of the well can obtain the formation pressure profile of the well.

[0126] The formation pressure prediction model established by drilling adjacent wells is used to accurately predict the formation pressure, thereby achieving the prediction accuracy of formation pressure under complex formation conditions, which has important guiding significance for reducing drilling risks and achieving safe drilling.

[0127] The formation pressure prediction model training method provided in the embodiment of the present application analyzes the historical drilling and logging data and formation pressure profiles of completed adjacent wells, calculates the normal compaction acoustic wave time difference and hydrostatic pressure parameters, and integrates traditional mechanism experience knowledge into the neural network input and structure, thereby improving the robustness, stability and prediction accuracy of the prediction model, making the formation pressure prediction model have a certain degree of interpretability, and effectively improving the stability and generalization ability of the formation pressure prediction model; determining the input parameters related to the formation pressure can reduce the dimension of the input data, reduce the input of irrelevant data, and improve the training efficiency of the formation pressure prediction model; using the prediction model obtained by this method to predict the formation pressure can ensure the stability and reliability of the prediction results, which is conducive to shortening the drilling cycle, reducing drilling costs, improving drilling quality and drilling efficiency, and enhancing the safety and controllability of drilling, and ultimately achieving the goal of ensuring drilling safety and reducing costs and increasing efficiency.

[0128] Embodiment 2

[0129] The present application embodiment provides a formation pressure prediction method, referring to Figure 4 As shown, the method includes:

[0130] S201: Determine the acoustic time difference of hydrostatic pressure and normal compaction based on the drilling and logging data obtained in advance during drilling;

[0131] S202: Determine input parameters according to the correlation strength between the formation pressure obtained in advance during drilling and the drilling data and the logging data;

[0132] S203: determining an input vector according to the acoustic wave time difference and hydrostatic pressure of the normal compaction and the input parameters;

[0133] S204: Inputting the input vector into the acquired formation pressure prediction model to determine the predicted value of the formation pressure during drilling.

[0134] The specific implementation of the above-mentioned formation pressure prediction process can refer to the method of obtaining the formation pressure prediction value based on the historical input vector prediction in Example 1, which will not be repeated here.

[0135] Embodiment 3

[0136] Based on the same inventive concept, the present application also provides a formation pressure prediction model training device, referring to Figure 5 As shown, the device comprises:

[0137] The first determination module 101 is used to determine each set of historical hydrostatic pressure and historical acoustic time difference of normal compaction of each completed adjacent well at different well depths according to each set of drilling and logging data of each completed adjacent well at different well depths obtained in advance;

[0138] The second determination module 102 is used to determine a corresponding set of historical input parameters according to the correlation strength between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance;

[0139] The third determination module 103 is used to divide the batches according to the formation pressure profile of each completed adjacent well and each group of the normal compaction historical acoustic wave time difference and historical hydrostatic pressure and the historical input parameters, and determine multiple groups of historical input vectors and corresponding actual values ​​of formation pressure;

[0140] A first prediction module 104 is used to input the plurality of groups of historical input vectors into a pre-built initial formation pressure prediction model to determine a formation pressure prediction value corresponding to each group of historical input vectors;

[0141] The error judgment module 105 is used to judge whether the error between the formation pressure prediction value corresponding to each group of historical input vectors and the corresponding formation pressure actual value is less than a preset range;

[0142] A model determination module 106 is used to obtain a formation pressure prediction model;

[0143] The updating module 107 is used to re-divide the multiple groups of historical input vectors and the corresponding actual values ​​of formation pressure, and re-execute the above model training process on the re-divided multiple groups of historical input vectors.

[0144] Embodiment 4

[0145] Based on the same inventive concept, the present application also provides a formation pressure prediction device, referring to Figure 6 As shown, the device comprises:

[0146] The fourth determination module 201 is used to determine the acoustic wave time difference and hydrostatic pressure of normal compaction based on the drilling and logging data obtained in advance during drilling;

[0147] A fifth determination module 202 is used to determine input parameters according to the pre-obtained correlation strength between the formation pressure in the well and the drilling and logging data;

[0148] A sixth determination module 203, configured to determine an input vector according to the acoustic wave time difference and hydrostatic pressure of the normal compaction and the input parameters;

[0149] The second prediction module 204 is used to input the input vector into a pre-built formation pressure prediction model to determine a predicted value of formation pressure during drilling.

[0150] Embodiment 5

[0151] Based on the same inventive concept, an embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, it implements the formation pressure prediction model training method described in the above embodiment one, and / or the formation pressure prediction method described in the above embodiment two.

[0152] Embodiment 6

[0153] Based on the same inventive 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. When the processor executes the computer program, it implements the formation pressure prediction model training method described in the above-mentioned embodiment 1, and / or the formation pressure prediction method described in the above-mentioned embodiment 2.

[0154] Embodiment 7

[0155] Based on the same inventive concept, an embodiment of the present application also provides a computer program product comprising instructions. When the computer program product is run on a computer device, the computer device executes the formation pressure prediction model training method described in the above embodiment one, and / or the formation pressure prediction method described in the above embodiment two.

[0156] Embodiment 8

[0157] Based on the same inventive concept, an embodiment of the present application also provides a chip, the chip includes a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement the formation pressure prediction model training method described in the above embodiment one, and / or the formation pressure prediction method described in the above embodiment two.

[0158] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0159] The present invention 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 block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to 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.

[0160] 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.

[0161] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0162] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A formation pressure prediction model training method, characterized in that: include: Determine each set of historical hydrostatic pressure and historical acoustic time difference of normal compaction at different depths of each completed adjacent well according to each set of drilling and logging data at different depths of each completed adjacent well obtained in advance; Determine a corresponding set of historical input parameters according to the strength of correlation between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance; According to the formation pressure profile of each completed adjacent well and each group of the normal compaction historical acoustic time difference and historical hydrostatic pressure and the historical input parameters, batch division is performed to determine multiple groups of historical input vectors and corresponding actual values ​​of formation pressure; Inputting the plurality of groups of historical input vectors into a pre-built initial formation pressure prediction model to determine a formation pressure prediction value corresponding to each group of historical input vectors; Determine whether the error between the predicted formation pressure value corresponding to each set of historical input vectors and the corresponding actual formation pressure value is less than a preset range; If yes, the formation pressure prediction model is obtained; If not, then the plurality of groups of historical input vectors and the corresponding actual values ​​of formation pressure are re-divided, and the model training process is re-executed on the plurality of groups of historical input vectors after being re-divided.

2. The method according to claim 1, characterized in that The method of determining a corresponding set of historical input parameters based on the correlation strength between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance includes: Preprocessing each set of drilling and logging data and formation pressure profiles of each completed adjacent well obtained in advance to obtain each set of drilling and logging data and formation pressure profiles of each completed adjacent well after preprocessing; Determine the correlation between each parameter in each set of drilling and logging data and the formation pressure profile according to each set of drilling and logging data and the formation pressure profile of each completed adjacent well after the preprocessing; Parameters having a correlation with the formation pressure profile greater than a preset threshold are selected from each set of drilling and logging data as a corresponding set of historical input parameters.

3. The method according to claim 1, characterized in that The initial formation pressure prediction model includes a BP neural network, an LSTM neural network and a connection layer; The step of inputting the plurality of groups of historical input vectors into a pre-built initial formation pressure prediction model to determine a formation pressure prediction value corresponding to each group of historical input vectors includes: Using the non-time series data in each group of the historical input vectors as the input parameters of the BP neural network to obtain the output parameters of the BP neural network; The time series data in each group of the historical input vectors are used as the input parameters of the LSTM neural network and the output parameters of the LSTM neural network; wherein the non-time series data in the historical input vectors are the historical input parameters corresponding to each discrete data type, and the time series data in the historical input vectors are the historical input parameters corresponding to the historical hydrostatic pressure, the historical acoustic time difference of normal compaction, and each continuous data type; The output parameters of the BP neural network and the output parameters of the LSTM neural network are fused in parallel through the connection layer to obtain the formation pressure prediction value corresponding to each group of the historical input vectors.

4. The method according to claim 1, characterized in that Based on the previously obtained drilling and logging data of each completed adjacent well, the historical acoustic time differences of each group of normal compaction at different well depths of each completed adjacent well are determined in the following manner: Based on the pre-obtained drilling and logging data of each completed adjacent well, determine the variation of acoustic time difference with vertical depth under each normal compaction condition; According to the variation law of acoustic time difference with vertical depth under each normal compaction condition, the normal compaction trend line is fitted, and the historical acoustic time difference of each group of normal compaction at different well depths of each completed adjacent well is determined based on the following formula 1: lnΔt=lnΔt0-CH Formula 1; Where Δt is the acoustic time difference corresponding to a certain depth point in the completed adjacent well; Δt0 is the surface acoustic time difference; C is the compaction coefficient; and H is the depth of the completed adjacent well.

5. The method according to claim 1, characterized in that Based on the previously obtained groups of drilling and logging data of different well depths of each completed adjacent well, the groups of historical hydrostatic pressures of different well depths of each completed adjacent well are determined in the following manner: According to the depth of each completed adjacent well, the historical hydrostatic pressures of each group of different depths of each completed adjacent well are determined based on the following formula 2: P h =ρgh Formula 2; Where P h is the historical hydrostatic pressure; ρ is the density of formation water; g is the local gravitational acceleration; h is the depth of the completed adjacent well.

6. A formation pressure prediction method, characterized in that: include: Determine the sonic time difference of hydrostatic pressure and normal compaction based on previously acquired drilling and logging data in the wellbore; Determining input parameters based on the strength of correlation between the formation pressure obtained in advance during drilling and the drilling data and the logging data; Determining an input vector according to the acoustic wave time difference and hydrostatic pressure of the normal compaction and the input parameters; The input vector is input into the obtained formation pressure prediction model to determine the predicted value of the formation pressure during drilling.

7. A formation pressure prediction model training device, characterized in that: include: The first determination module is used to determine each group of historical hydrostatic pressure and historical acoustic time difference of normal compaction of each completed adjacent well at different depths according to each group of drilling and logging data of each completed adjacent well at different depths obtained in advance; The second determination module is used to determine a corresponding set of historical input parameters according to the correlation strength between each set of drilling data and logging data of each completed adjacent well and the formation pressure profile obtained in advance; The third determination module is used to divide the batches according to the formation pressure profile of each completed adjacent well and the historical acoustic time difference and historical hydrostatic pressure of each group of normal compaction and the historical input parameters, and determine multiple groups of historical input vectors and corresponding actual values ​​of formation pressure; A first prediction module, used for inputting the plurality of groups of historical input vectors into a pre-built initial formation pressure prediction model, and determining a formation pressure prediction value corresponding to each group of historical input vectors; An error judgment module is used to judge whether the error between the predicted formation pressure value corresponding to each set of historical input vectors and the corresponding actual formation pressure value is less than a preset range; A model determination module is used to obtain a formation pressure prediction model; The updating module is used to re-divide the multiple groups of historical input vectors and the corresponding actual values ​​of formation pressure, and re-execute the above model training process on the re-divided multiple groups of historical input vectors.

8. A formation pressure prediction device, characterized in that: include: A fourth determination module is used to determine the acoustic wave time difference and hydrostatic pressure of normal compaction based on the drilling and logging data obtained in advance during drilling; a fifth determination module, for determining input parameters according to the strength of correlation between the formation pressure in the well and the drilling and logging data obtained in advance; a sixth determination module, configured to determine an input vector according to the acoustic time difference and hydrostatic pressure of the normal compaction and the input parameters; The second prediction module is used to input the input vector into a pre-built formation pressure prediction model to determine the predicted value of the formation pressure during drilling.

9. A computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal, the terminal executes the formation pressure prediction model training method as described in any one of claims 1 to 5, and / or the formation pressure prediction method as described in claim 6.

10. A computer device, characterized in that: It comprises 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, it implements the formation pressure prediction model training method as described in any one of claims 1 to 5, and / or the formation pressure prediction method as described in claim 6.

11. A computer program product comprising instructions, which, when executed on a computer device, enables the computer device to execute the formation pressure prediction model training method as described in any one of claims 1 to 5, and / or the formation pressure prediction method as described in claim 6.

12. A chip, comprising a processor and a communication interface, wherein the communication interface and the processor are coupled, and the processor is used to run a computer program or instruction to implement the formation pressure prediction model training method as described in any one of claims 1 to 5, and / or the formation pressure prediction method as described in claim 6.

Images