Well logging information prediction method and device

Abstract

The present application discloses a method and device for predicting well logging information, which belongs to the field of detection technology. The method includes: obtaining time-lapse seismic attribute information of a target well in multiple periods, which are different from each other; based on the time-lapse seismic attribute information of the target well in multiple periods, using partial least squares regression method to predict well logging information in multiple periods; outputting the predicted well logging information in multiple periods. The method solves the problem that the reservoir geological model established based on this in the related technology is of low accuracy, which affects the guidance of oil reservoir exploitation, and achieves the effect of improving the effectiveness of oil reservoir exploitation guidance.

Description

Technical Field

[0001] The present application relates to the field of detection technology, and in particular to a method and device for predicting well logging information. Background Art

[0002] A reservoir geological model is a model that can reflect the porosity, permeability, fluid characteristics and dynamic characteristics of a reservoir. The analysis of the reservoir geological model can guide the reservoir development plan and the analysis of the remaining oil in the reservoir, thereby achieving effective exploitation of the reservoir. Therefore, the data accuracy when establishing the reservoir model is particularly important for the establishment of the reservoir geological model. Permeability is a value that measures the ability of a fluid to pass through the effective pores of a porous rock under a pressure difference, and can be used to determine the impact of water injection on the reservoir's physical properties.

[0003] A permeability acquisition method in the related technology is to place a controlled acoustic wave source into the well. The sound waves emitted by the sound source cause the vibration of the surrounding particles. The waves generated in the formation propagate in the rock to the receiver, and the changes in the speed, amplitude and frequency of the wave propagating in the rock are recorded to obtain the logging acoustic wave curve, and the obtained data is regressed to obtain the reservoir permeability.

[0004] However, the reservoir permeability obtained by regression calculation of the logging acoustic wave curve in the above-mentioned related technologies reflects the reservoir permeability under the current rock formation state. The reservoir geological model established based on this has low accuracy, which affects the guidance of oil reservoir exploitation.

[0005] Application Contents

[0006] The embodiment of the present application provides a method and device for predicting well logging information. The technical solution is as follows:

[0007] According to one aspect of the present application, a method for predicting well logging information is provided, the method comprising:

[0008] Acquiring time-lapse seismic attribute information of a target well in multiple periods, wherein the multiple periods are different from each other;

[0009] Based on the time-lapse seismic attribute information of the target well in multiple periods, a partial least squares regression method is used to predict the logging information of the multiple periods;

[0010] The predicted logging information of the multiple periods is output.

[0011] Optionally, there are multiple target wells, and based on the time-lapse seismic attribute information of the target wells in multiple periods, a partial least squares regression method is used to predict the logging information of the multiple periods, and the method further includes:

[0012] Calculating the correlation of time-lapse seismic attribute information of any target well among the plurality of target wells in two periods among the plurality of periods;

[0013] When the correlation of the time-lapse seismic attribute information of any target well in the two periods is greater than the correlation threshold, the target well is eliminated;

[0014] Based on the time-lapse seismic attribute information of the target well in the first period, the partial least squares regression method is used to predict the logging information of the first period.

[0015] Optionally, the method further includes:

[0016] Establishing a seismic attribute logging curve of the target well based on the time-lapse seismic attribute information of the target well in multiple periods;

[0017] Output the seismic attribute logging curve of each target well.

[0018] Optionally, the output prediction of the logging information of the multiple periods includes:

[0019] Based on the predicted logging information of the multiple periods, generating a logging curve of the logging information of each period;

[0020] Output the well log curves for each of the stated periods.

[0021] Optionally, the logging information includes permeability.

[0022] The optional time-lapse seismic attribute information includes at least one of the following: root mean square amplitude, instantaneous frequency, dominant frequency, apparent polarity, instantaneous phase, amplitude envelope or instantaneous phase cosine.

[0023] Optionally, the method further includes:

[0024] Based on the logging information of the multiple periods, a reservoir dynamic geological model reflecting the reservoir state of the target well is generated, and the reservoir dynamic geological model is output;

[0025] Alternatively, based on the logging information of the multiple periods, a time-varying numerical simulation model reflecting the time-varying state of the logging data of the target well is generated, and the time-varying numerical simulation model is output.

[0026] According to another aspect of the present application, a well logging information prediction device is provided, the device comprising:

[0027] An acquisition module, used to acquire time-lapse seismic attribute information of a target well in multiple periods, wherein the multiple periods are different from each other;

[0028] A prediction module, used for predicting the logging information of the target well in multiple periods by using a partial least squares regression method based on the time-lapse seismic attribute information of the target well in multiple periods;

[0029] The output module is used to output the predicted logging information of the multiple periods.

[0030] Optionally, there are multiple target wells, and the partial least squares regression method is used based on the time-lapse seismic attribute information of the target wells in multiple periods to predict the logging information of the multiple periods, and the device further includes:

[0031] A calculation module, used for calculating the correlation of the time-lapse seismic attribute information of every two target wells among the plurality of target wells in a first period, wherein the first period is any period among the plurality of periods;

[0032] A judgment module, configured to determine any two target wells as associated wells when the correlation of the time-lapse seismic attribute information of any two target wells in the first period is greater than a correlation threshold;

[0033] The prediction submodule predicts the logging information of the first period by using a partial least squares regression method based on the time-lapse seismic attribute information of the associated wells in the first period.

[0034] Optionally, the well logging information prediction device further includes:

[0035] A generating module, configured to generate a reservoir dynamic geological model reflecting the reservoir state of the target well based on the well logging information of the plurality of periods, and output the reservoir dynamic geological model;

[0036] Alternatively, based on the logging information of the multiple periods, a time-varying numerical simulation model reflecting the time-varying state of the logging data of the target well is generated, and the time-varying numerical simulation model is output.

[0037] The technical solution provided by the embodiments of the present application may have the following beneficial effects:

[0038] A logging information prediction method is provided to obtain the time-lapse seismic attribute information of the target well at different periods, and the logging information of different periods predicted by partial least squares regression method is output. The logging information of the target well at different periods is obtained. Since rich logging information is obtained, it provides a basis for establishing an accurate reservoir geological model, thereby providing effective guidance for oil reservoir exploitation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0040] Figure 1 is a flow chart of a well logging information prediction method provided in an embodiment of the present application;

[0041] Figure 2 is a flow chart of another well logging information prediction method provided in an embodiment of the present application;

[0042] Figure 3 A schematic diagram of a correlation analysis provided in an embodiment of the present application;

[0043] Figure 4 A schematic diagram of another correlation analysis provided in an embodiment of the present application;

[0044] Figure 5 A comparison chart of the permeability curves of the target wells provided in the embodiments of the present application;

[0045] Figure 6 It is a structural schematic diagram of a well logging information prediction device provided in an embodiment of the present application;

[0046] Figure 7 Crossplot of original permeability and predicted permeability during the medium and low water content periods;

[0047] Figure 8 The cross-plot of original permeability and predicted permeability during the high water content period;

[0048] Fig. 9 This is the cross-plot of original permeability and predicted permeability during the period of extremely high water content.

[0049] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0050] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0051] At present, high-precision non-coarsening reservoir numerical simulation research has gradually become the focus of application in various oil fields. The establishment of three-dimensional reservoir geological models and four-dimensional reservoir geological models can more accurately judge the reservoir and oil, thereby guiding the development of the reservoir. The permeability data obtained by the acoustic logging method in the relevant technology is the permeability status of the current reservoir. For example, if the reservoir is a loose sandstone reservoir, the physical properties of the loose sandstone reservoir change due to water injection scouring. For the newly put into production wells in the late stage of development on the loose sandstone reservoir, the logging curve obtained by the acoustic logging method in the relevant technology can reflect the reservoir physical properties of the current late stage, but cannot reflect the reservoir physical properties of the early stage of development, resulting in incomplete numerical values ​​for establishing the reservoir geological model, and the accuracy of the reservoir geological model is reduced, resulting in inaccurate judgment of the results of the later numerical simulation and the results of the remaining oil prediction, which cannot effectively guide the exploitation of the reservoir.

[0052] The embodiments of the present application provide a method and device for predicting well logging information, which can solve the above-mentioned problems. Figure 1 This is a flow chart of a method for predicting well logging information provided by an embodiment of the present application. The method is executed by a computer device, and the steps are as follows:

[0053] Step 101: Obtain time-lapse seismic attribute information of a target well in multiple periods, where the multiple periods are different from each other.

[0054] Step 102: Based on the time-lapse seismic attribute information of the target well in multiple periods, a partial least squares regression method is used to predict the well logging information in multiple periods.

[0055] Step 103: output the predicted logging information for multiple periods.

[0056] In summary, the embodiment of the present application provides a method for predicting well logging information, which obtains time-lapse seismic attribute information of a target well at multiple different periods, and outputs the well logging information of multiple different periods predicted by partial least squares regression method. The well logging information obtained is the well logging information of the target well at different periods. Since rich well logging information is obtained, it provides a basis for establishing an accurate reservoir geological model, thereby providing effective guidance for oil reservoir exploitation.

[0057] Please refer to Figure 2 , which shows a flowchart of another well logging information prediction method provided by an embodiment of the present application, the method can be executed by a computer device, and the method includes:

[0058] Step 201, obtaining time-lapse seismic attribute information of a target well in multiple periods, wherein the multiple periods are different from each other; the time-lapse seismic attribute information includes at least one of the following: root mean square amplitude, instantaneous frequency, main frequency, apparent polarity, instantaneous phase, amplitude envelope or instantaneous phase cosine.

[0059] Time-lapse seismic attribute information can reflect the geometric and physical characteristics of seismic waves, and these characteristics will reflect reservoir characteristics such as the lithology, physical properties of the reservoir and the properties of the fluid in the pores to varying degrees. The time-lapse seismic attribute information of different periods is different. Among them, there are many ways to divide the multiple periods of the target well. For example, according to the chronological order, it is divided into early, middle or late periods, and the duration of each period is the same or different, for example, one month; for example, according to the different water content, it is divided into low water content period, medium-low water content period, medium water content period, high water content period and ultra-high water content period, and the difference in water content of any adjacent periods can be the same or different; for example, the multiple periods can also be determined by artificial experience division or other division methods, and the embodiments of the present application are not limited here. The same time-lapse seismic attribute information of the same target well in different periods may change. There are multiple types of seismic attributes. In the embodiment of the present application, amplitude and frequency seismic attributes are selected to obtain time-shifted seismic attribute information of root mean square amplitude, instantaneous frequency, main frequency, apparent polarity, instantaneous phase, amplitude envelope or instantaneous phase cosine. Other types of seismic attributes can also be selected, which are not limited in the embodiment of the present application.

[0060] The root mean square amplitude can reflect the average change level of the seismic wave amplitude and the average absorption properties of the formation within a specific time window. The value of the value is related to the reservoir properties, rock composition and fluid properties. The root mean square amplitude can identify special lithologies such as bright spots, dark spots, and igneous rocks, and can indicate hydrocarbon displays. It is often used in the analysis of lithological phase changes in formations. The root mean square amplitude is the square root of the average square of the amplitude within the time window. The calculation formula is as follows:

[0061]

[0062] Where a i is the amplitude sampling point value of the i-th seismic trace, and N is the number of sampling points in the time window.

[0063] The amplitude envelope provides information about acoustic impedance differences and can be used to identify bright spots, dark spots, and flat spots. It is often used to delineate high or low amplitude anomalies and determine lateral changes in fluid composition, lithology, and formations in reservoirs. The amplitude envelope is the total instantaneous energy of the composite seismic signal, and is calculated as follows:

[0064]

[0065] Where f(t) is the real part of the seismic trace, and g(t) is the imaginary part of the complex seismic trace, which is obtained by Hilbert transforming f(t).

[0066] Instantaneous phase, instantaneous phase describes the angle between the phase vector and the real axis, and is often used in conjunction with phase cosine. Instantaneous phase helps to strengthen the continuity of the event axis and can be used to identify discontinuities and small faults. When performing phase reversal, it can be used to determine the location of oil and gas. Instantaneous phase formula:

[0067]

[0068] The instantaneous frequency is used to calculate and estimate the attenuation of seismic waves. It is related to the propagation of waves and the sedimentary environment and is a physical property. It is usually used to indicate the contrast between seismic attenuation and faults, and can also be used to indicate the gas-water or gas-oil contact surface. The instantaneous frequency is the time derivative of the instantaneous phase and is an indication of the acceleration of the phase. The calculation formula is as follows:

[0069]

[0070] The main frequency is the frequency at the maximum point of the spectrum curve of the seismic data, which is helpful for seismic low-frequency search. The change of the main frequency of the seismic wave is caused by the change of local lithology and fluid, so it is often used to predict the attenuation of high-frequency components caused by oil and gas in the formation, the change of rock grain size, the cyclicity of the formation, etc. The main frequency can be obtained by Fourier transforming the seismic wave. The specific calculation process can refer to the relevant technology, and the embodiment of this application will not be repeated here.

[0071] Apparent polarity is the polarity of the instantaneous phase of the regional amplitude, reflecting the polarity corresponding to the change in reflection coefficient. It is often used in conjunction with reflection intensity to check the lateral change of polarity along the reflection layer, improve the quality of layer calibration, seismic interpretation, etc. Reflection at the top of the reservoir shows negative polarity, and reflection from the oil-gas or gas-water interface shows positive polarity.

[0072] The instantaneous phase cosine is a property derived from the instantaneous phase. It is often used to improve the display of instantaneous phase variations. It is also used for phase tracking and checking the quality of seismic profile comparison and interpretation. It is often used in conjunction with the instantaneous phase.

[0073] Step 202: Calculate the correlation of time-lapse seismic attribute information of any target well among the multiple target wells in two periods among the multiple periods.

[0074] For example, the two periods may be two adjacent periods among multiple periods, or two non-adjacent periods. For example, the two periods may be two adjacent periods selected from the early, middle, and late periods, or two non-adjacent periods, and the time-shift seismic attribute information of the two periods is obtained, and a correlation analysis is performed on the time-shift seismic attribute information of the two periods. Exemplarily, the instantaneous phase cosine of the early stage and the instantaneous phase cosine of the middle stage of a target well are obtained, and a correlation analysis of the target well is performed, and the analysis can be completed by correlation analysis software. The specific analysis process can refer to the relevant technology, and the embodiments of the present application are not limited here.

[0075] Step 203: When the correlation of the time-lapse seismic attribute information of any target well in two periods is greater than the correlation threshold, the target well is eliminated.

[0076] After obtaining the results of the correlation analysis, it is determined whether the correlation of the time-lapse seismic attribute information of the target well in two periods is greater than the correlation threshold. If the correlation is greater than the correlation threshold, it indicates that the correlation of at least two periods of the target well is poor, and it is impossible to provide an effective law of reservoir physical property changes in the subsequent steps. Therefore, the target well with a correlation greater than the correlation threshold is eliminated, which can improve the prediction accuracy of logging information in subsequent periods. After the target well with a correlation greater than the correlation threshold is eliminated, multiple updated target wells are obtained, and the subsequent steps are processed based on the updated multiple target wells, that is, the processed multiple target wells do not include the target well with a correlation greater than the correlation threshold.

[0077] For example, Figure 3 A schematic diagram of a correlation analysis provided in an embodiment of the present application, in which the analysis diagram is extended in a straight line form, indicates that the correlation between the time-lapse seismic attribute information of the target well in two periods is good, that is, the correlation is less than the correlation threshold. Figure 4 A schematic diagram of another correlation analysis provided in an embodiment of the present application, where the lines in the analysis diagram are disorderly and irregular, indicating that the correlation between the time-lapse seismic attribute information of the target well in two periods is poor, that is, the correlation is greater than the correlation threshold, and the target well is eliminated. The correlation threshold can be set according to the actual situation of the target well.

[0078] Step 204: Based on the time-lapse seismic attribute information of the target well in multiple periods, a partial least squares regression method is used to predict the well logging information in multiple periods.

[0079] For example, assuming that the first period can be any period among multiple periods, step 204 includes: based on the time-lapse seismic attribute information of the target well in the first period, using partial least squares regression method, predicting the well logging information of the first period, the well logging information including permeability.

[0080] By obtaining the changes in reservoir parameters caused by the changes in reservoir properties at different times due to water injection, which can be reflected by the permeability, a basis is provided for the law of changes in reservoir properties caused by water injection in loose sandstone reservoirs in the high water content period. Well logging information can also be other information that can reflect the laws of reservoir properties under different geological conditions, which is not limited in the embodiments of this application.

[0081] Partial least squares is used to find the basic relationship between two matrices (X and Y). The partial least squares regression method can establish a regression model between multiple independent variables and dependent variables, integrating the advantages of principal component analysis, canonical correlation analysis, and linear regression analysis in related technologies, and this method can effectively eliminate multiple correlations between factors. The process of using the partial least squares regression method to predict the permeability of the first period in the embodiment of the present application includes:

[0082] Step A1: using the known time-lapse seismic attribute information of the target period and the known well logging information of the target period, a target regression equation is determined by partial least squares regression method. The target regression equation is used to represent the relationship between the seismic attribute information and the well logging information of the same period.

[0083] Step A11: Set the independent variable X and the dependent variable Y. The independent variable X includes p elements, and each element has n sample data. The dependent variable Y includes n elements, and both p and n are positive integers. Then the independent variable X and the dependent variable Y satisfy the following formula:

[0084]

[0085]

[0086] For example, the p elements included in the independent variable X are determined based on the time-lapse seismic attribute information of the target well in the known target period, and the p elements include p of the root mean square amplitude, instantaneous frequency, main frequency, apparent polarity, instantaneous phase, amplitude envelope or instantaneous phase cosine, and the n sample data of each element are obtained by sampling the element of the element in the first period n times. For example, assuming x 1 is the root mean square amplitude, then x 11 to x n1 is the n root mean square amplitude values ​​obtained by sampling the root mean square amplitude of the first period n times. The dependent variable Y includes n elements, that is, n well logging information, such as n permeability values. The n well logging information corresponds one-to-one with the aforementioned n sample data. For example, the n well logging information is the well logging information obtained by sampling at the same sampling interval as the sample data during the target period. For example, each well logging information has the same sampling time as the corresponding sample data.

[0087] Optionally, after obtaining the independent variable X and the dependent variable Y, the independent variable X and the dependent variable Y may be standardized to obtain a standardized matrix X and a standardized vector Y, so as to reduce the computational complexity. It is worth noting that the standardization process may not be performed in the actual calculation process. For the convenience of the reader's understanding, the following embodiments are described by taking the standardization process as an example:

[0088] The normalized matrix X is denoted as E 0, which is an n×p prediction matrix, and the standardized vector Y is recorded as F 0 , which is an n×1 response matrix. For example, the element x* in the i-th row and j-th column of the standardized matrix ij Satisfies the following formula:

[0089]

[0090] in,

[0091] x ij Indicates E 0 The element in the i-th row and j-th column of E 0 The element in the jth column of It represents the average value of the elements in the jth column, and var represents the integer part.

[0092] Step A12: Based on the independent variable X and the dependent variable Y, obtain the target regression equation.

[0093] Step A121: From the independent variable matrix E 0 With the dependent variable F 0 Extract the principal component, assuming that the extracted principal component is the first principal component t 1 and u 1 (t 1 and u 1 YesE 0 With F 0 Linear combination). According to the principle of principal component analysis, t 1 and u 1 Carry as much original information in each data set as possible, and the components of the independent variables t 1 and the dependent variable u 1 The correlation reaches the maximum. That is, find t 1 and u 1 The covariance is maximized. For example, the first principal component t 1 and u 1 Satisfy the following formulas respectively:

[0094] t 1 =Xw 1 +w 1 x 1 +w 2 x 2 +…w p x p ;

[0095]

[0096]

[0097]

[0098] Among them, cov means covariance, E 01 To E 0p are the normalized values ​​from x1 to xp respectively.

[0099] Step A122: Establish a regression relationship equation between the principal component and the independent variable and the dependent variable. For example, the regression relationship equation satisfies the following first regression formula:

[0100] E 0 =t 1 p 1 +E 1 ;

[0101] F 0 =u 1 r 1 +F 1 ;

[0102] Among them, p1 and r1 represent the regression coefficients, E′ 0 For E 0 The residual matrix, F′ 0 F 0 The residual matrix of .

[0103] Step A123: Establish a regression equation based on the determined regression relationship equation. If the regression equation reaches the expected accuracy, the regression process is stopped, that is, the action is stopped; if the regression equation does not reach the expected accuracy, the residual matrix corresponding to the independent variable is replaced by the independent variable to obtain a new independent variable; the residual matrix corresponding to the dependent variable is replaced by the dependent variable to obtain a new dependent variable, and steps A121 to A123 are repeated until the accuracy of the obtained regression equation is greater than the expected accuracy, and the regression equation with an accuracy greater than the expected accuracy is determined as the target regression equation.

[0104] For example, using the residual matrix E 1 and F 1 Replace E respectively 0 and F 0 After that, we get the new independent variable E 1 and the new dependent variable F 1 ; Execute step A121, from E 1 and F 1 Extract the principal component from the , assuming that the extracted principal component is the second principal component t 2 and u 2 ; Execute step A122, establish a regression relationship equation between the principal component and the independent variable and the dependent variable, and the regression relationship equation satisfies the following second regression formula:

[0105] E 1 =t 2 p2 +E 2 ;

[0106] F 1 =u 2 r 2 +F 2 ;

[0107] The definition of each parameter in the second regression formula may refer to the definition of each parameter in the aforementioned first regression formula.

[0108] Substitute the second regression formula into the first regression formula to establish the regression equation:

[0109] E 0 =t 1 p 1 +t 2 p 2 +E 2 ;

[0110] F 0 =t 1 r 1 +t 2 r 2 +F 2 ;

[0111] If the regression equation reaches the expected accuracy, the regression process is stopped, that is, the action is stopped; if the regression equation does not reach the expected accuracy, the residual matrix corresponding to the independent variable is replaced by the independent variable to obtain a new independent variable; the dependent variable is replaced by the residual matrix corresponding to the dependent variable to obtain a new dependent variable, and A121 to step A123 are executed again until the accuracy of the obtained regression equation is greater than the expected accuracy, and the regression equation with an accuracy greater than the expected accuracy is determined as the target regression equation.

[0112] Optionally, the partial least squares regression method generally uses a cross-validation test method to determine the number of principal components to be extracted, that is, to consider whether the prediction accuracy of the regression equation will be improved after adding a new component. When executing the aforementioned step A12, a cross-validation test can be used to determine whether the regression equation has achieved the expected accuracy. When the multiple principal components obtained meet the cross-validation, it is determined that the regression equation established by the multiple principal components has achieved the expected accuracy; when the multiple principal components obtained do not meet the cross-validation, it is determined that the regression equation established by the multiple principal components has not achieved the expected accuracy. The specific implementation is as follows:

[0113] First, the acquired data set (ie, the data in the time-lapse seismic attribute information of the aforementioned target period) is divided into a training data set and a validation data set. The training data set is used to derive the transformation, and the validation data set is used to measure the final prediction error.

[0114] Use the training data set without sample point i and the regression equation containing h principal components, where h is greater than or equal to 1, and find the fitting value of sample point i Repeat the above steps to obtain the fitting value for each sample point to obtain y j The sum of squared prediction errors of y j The prediction error sum of squares PRESS h,j Satisfies the following formula:

[0115]

[0116] Then the prediction error sum of squares of Y is PRESS h Satisfies the following formula:

[0117]

[0118] At the same time, the validation data set containing all sample points is used to fit the regression equation containing h principal components, and the predicted value of the i-th sample point is recorded as Get y j The error sum of squares SS hj and the error sum of squares SS of Y hJ , where SS hJ and SS hj Satisfy the following formulas respectively:

[0119]

[0120] Define cross validity Q 2 h Satisfies the following formula:

[0121]

[0122] In an alternative example, when the cross-validation Q 2 h is less than a first specified threshold, such as If the marginal contribution of the component is effective, that is, it meets the cross validity, then the first h principal components are extracted to establish the final regression equation, which is the regression equation that achieves the expected accuracy. 2 h Not less than the first specified threshold t h The marginal contribution of the component is invalid and does not meet the cross-validity requirement. Therefore, the regression equation established by the first h principal components does not achieve the expected accuracy.

[0123] In another alternative example, when is less than a second specified threshold, such as t hIf the marginal contribution of the component is effective, that is, it meets the cross-validation requirement, the first h principal components are extracted to establish the final regression equation, which is the regression equation that achieves the expected accuracy. Not less than the second specified threshold, t h The marginal contribution of the component is invalid and does not meet the cross-validity requirement. Therefore, the regression equation established by the first h principal components does not achieve the expected accuracy.

[0124] Step A2: for the time-lapse seismic attribute information of each period in the multiple periods of the target well, use the target regression equation to predict the well logging information of the period.

[0125] For the time-lapse seismic attribute information of each period in multiple periods of the target well, the time-lapse seismic attribute information of the period is substituted into the target regression equation as an independent variable, and the obtained dependent variable is the well logging information of the period.

[0126] It is worth noting that the aforementioned step 204 may also be implemented in other ways. In an optional example, the computer device may obtain a target regression equation established in advance using a partial least squares regression method, and directly substitute the time-lapse seismic attribute information of each period of multiple periods of the target well as an independent variable into the target regression equation, and the obtained dependent variable is the well logging information of the corresponding period.

[0127] In another optional example, the computer device inputs the time-lapse seismic attribute information of each of the multiple periods of the target well into the artificial intelligence model, and receives the logging information of the multiple periods output by the artificial intelligence model. The artificial intelligence model is used to predict the logging information of the received period based on the received time-lapse seismic attribute information of each period using partial least squares regression method.

[0128] After the computer device obtains the logging information of multiple periods, the predicted logging information of multiple periods can be output. To facilitate the understanding of the reader, the following embodiments are described by taking the example that the predicted logging information of multiple periods can be represented by curves.

[0129] Step 205: Generate a logging curve of the logging information of each period based on the predicted logging information of multiple periods.

[0130] By representing the logging information in the form of curves, the changes in the logging information can be seen more intuitively, making it easier to discover the changing patterns of the logging information.

[0131] Step 206: Output the logging curve of each period.

[0132] The generated curve is output so that subsequent relevant technical personnel can more intuitively compare the data with other data.

[0133] Step 207: Establish a seismic attribute logging curve of the target well based on the time-lapse seismic attribute information of the target well in multiple periods.

[0134] The time-lapse seismic attribute information of multiple periods obtained in step 201 is represented in the form of a curve. Step 207 can also be completed after step 201.

[0135] Step 208: output the seismic attribute logging curve of each target well.

[0136] The time-lapse seismic attribute information of multiple periods can be split into different periods to output seismic attribute logging curves, or the time-lapse seismic attribute information of multiple periods can be continuously output as seismic attribute logging curves.

[0137] Figure 5 A comparison chart of the permeability curves of the target wells provided in the embodiment of the present application. For the convenience of comparison, the logging curves generated by the logging information of multiple periods predicted can be compared with the seismic attribute logging curves, so as to determine that the numerical error of the permeability predicted by the partial least squares regression method is within the allowable range.

[0138] Step 209: Generate a reservoir dynamic geological model reflecting the reservoir status of the target well based on logging information of multiple periods, and output the reservoir dynamic geological model; or, generate a time-varying numerical simulation model reflecting the time-varying situation of logging data of the target well based on logging information of multiple periods, and output the time-varying numerical simulation model.

[0139] The geological model can be used as an important tool to guide the preparation of reservoir development plans and residual oil analysis. Therefore, the accuracy of the geological model is required to be high. Affected by water injection, the reservoir properties of loose sandstone reservoirs often change. This causes the logging curves obtained for newly commissioned wells in the late development period to often fail to reflect the reservoir properties in the early development period, thereby reducing the accuracy of the geological model, affecting the results of later numerical simulations, and causing inaccurate predictions of residual oil. In the embodiment of the present application, the partial least squares regression method is used to establish a multi-period logging permeability geological model corresponding to time-lapse seismic attribute information of multiple periods, which greatly improves the prediction accuracy of reservoir properties, so that the model can be used to guide staff to determine whether the target well can be exploited.

[0140] In summary, the embodiment of the present application provides a method for predicting well logging information, which obtains time-lapse seismic attribute information of a target well at multiple different periods, and outputs the well logging information of multiple different periods predicted by partial least squares regression method. The well logging information obtained is the well logging information of the target well at different periods. Since rich well logging information is obtained, it provides a basis for establishing an accurate reservoir geological model, thereby providing effective guidance for oil reservoir exploitation.

[0141] Figure 6 : is a schematic diagram of the structure of a well logging information prediction device provided in an embodiment of the present application. The structure of the well logging information prediction device 60 is as follows:

[0142] An acquisition module 601 is used to acquire time-lapse seismic attribute information of a target well in multiple periods, where the multiple periods are different from each other;

[0143] A calculation module 602 is used to calculate the correlation of time-lapse seismic attribute information of every two target wells among multiple target wells in a first period, where the first period is any period among the multiple periods;

[0144] A judgment module 603 is used to determine any two target wells as associated wells when the correlation of the time-lapse seismic attribute information of any two target wells in the first period is greater than a correlation threshold;

[0145] Prediction submodule 604: Based on the time-lapse seismic attribute information of the first period of the associated well, the partial least squares regression method is used to predict the well logging information of the first period.

[0146] The output module 605 is used to output the predicted logging information of multiple periods.

[0147] Generation module 606 is used to generate a reservoir dynamic geological model reflecting the reservoir status of the target well based on logging information of multiple periods, and output the reservoir dynamic geological model; or, based on logging information of multiple periods, generate a time-varying numerical simulation model reflecting the time-varying situation of logging data of the target well, and output the time-varying numerical simulation model.

[0148] In summary, the embodiment of the present application provides a well logging information prediction device, which obtains time-lapse seismic attribute information of a target well at multiple different periods, and outputs the well logging information of multiple different periods predicted by partial least squares regression method. The well logging information obtained is the well logging information of the target well at different periods. Since rich well logging information is obtained, it provides a basis for establishing an accurate reservoir geological model, thereby providing effective guidance for oil reservoir exploitation.

[0149] In order to verify the accuracy of predicting permeability using the method of steps 201-204 above, the present application conducts the following experiments:

[0150] Five wells in the medium and low water cut period were selected, and the well names were Dong 2-13, Dong 6-11, Gang 3-60, Gang 91 and Gang 2-64; five wells in the high water cut period were selected, and the well names were Dong 4-14, Gang 1-62, Gang 2-61-1, Gang 251-2 and Gang 291-4; five wells in the ultra-high water cut period were selected, and the well names were Gang 2-60-1, Gang 2-62-3, Gang 4-64-3, Gang 5-66-2 and Gang 1-63-2. The actual permeability of different wells in the same water cut period was compared with the permeability predicted by the method of steps 201-204. It can be seen from Table 1 below that the numerical error of the permeability predicted by the method of steps 201-204 is within the allowable range. That is, the permeability predicted by the partial least squares regression method has a higher accuracy.

[0151] Table 1 Comparison of original and predicted permeability values ​​at different water content periods

[0152]

[0153]

[0154] At the same time, the original permeability and predicted permeability cross-plots of different water-cut periods were produced using the single-well permeability data of the medium-low water-cut period, the high water-cut period and the extremely high water-cut period. Figure 7 Crossplot of original permeability and predicted permeability during the medium and low water content period. Figure 8 is the cross plot of original permeability and predicted permeability during high water content period, Fig. 9 This is the cross plot of original permeability and predicted permeability during the period of extremely high water content. Figure 7 , 8 In Figure 9, the horizontal axis of the cross-plot is the original permeability of a single well, the vertical axis is the predicted permeability of a single well, and md is the unit of permeability, millidarcy. Figure 7 , 8 As can be seen from Figure 9, the original permeability value and the predicted value are positively correlated, and the error is relatively large only when the permeability is less than 500md. If the permeability value of the mudstone section and the low permeability section is removed, the correlation between the original and predicted permeability of the sand layer can be as high as 9.0 or more. Therefore, the permeability predicted by the partial least squares regression method has a high accuracy. The permeability predicted by the above method is used as the data for making the model, and the obtained model has a high accuracy.

[0155] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting well logging information, characterized in that: The method comprises: Acquiring time-lapse seismic attribute information of a target well in multiple periods, wherein the multiple periods are different from each other; Based on the time-lapse seismic attribute information of the target well in multiple periods, using partial least squares regression method to predict the logging information of the multiple periods, the logging information including permeability; Outputting the predicted logging information of the multiple periods; The method of using partial least squares regression method based on the time-lapse seismic attribute information of the target well in multiple periods to predict the logging information of the multiple periods includes: using partial least squares regression method based on the time-lapse seismic attribute information of the target well in a first period to predict the logging information of the first period, wherein the first period is any one of the multiple periods; The method of using partial least squares regression to predict the well logging information in the first period includes: Set the independent variable X and the dependent variable Y. The independent variable X includes p elements, and each element has n sample data. The dependent variable Y includes n elements. Both p and n are positive integers. Then the independent variable X and the dependent variable Y satisfy the following formula: The p elements included in the independent variable X are determined based on the known time-lapse seismic attribute information of the target well in the target period, the p elements include p of the root mean square amplitude, instantaneous frequency, main frequency, apparent polarity, instantaneous phase, amplitude envelope or instantaneous phase cosine, and the n sample data of each element are obtained by sampling the element in the first period n times; After obtaining the independent variable X and the dependent variable Y, the independent variable X and the dependent variable Y are standardized to obtain the standardized matrix X and the standardized vector Y; the standardized matrix X is recorded as E0, E0 is an n×p prediction matrix, and the standardized vector Y is recorded as F0, F0 is an n×1 response matrix; the element x* in the i-th row and j-th column of the standardized matrix ij Satisfies the following formula: in, (j=1,2,...,p),n>1;x ij represents the element in row i and column j in E0, x j represents the element in the jth column of E0, represents the average value of the elements in the jth column, and var represents the integer part; The principal components are extracted from the independent variable matrix E0 and the dependent variable F0. The extracted principal components are the first principal components t1 and u1, where the covariance of t1 and u1 reaches the maximum. The first principal components t1 and u1 satisfy the following formulas respectively: t1=Xw1+w1x1+w2x2+…w p x p ; Among them, cov means covariance, E 01 To E 0p They are x1 to x p Normalized value; A regression relationship equation is established between the principal component and the independent variable and the dependent variable, and the regression relationship equation satisfies the following first regression formula: E0=μ1p1+E1;F0=u1r1+F1;where p1 and r1 represent regression coefficients, E′0 is the residual matrix of E0, F′0 is the residual matrix of F0, E1 is the new independent variable, and F1 is the new dependent variable; Based on the determined regression relationship equation, a regression equation is established; if the regression equation reaches the expected accuracy, the regression process is stopped; if the regression equation does not reach the expected accuracy, the residual matrix corresponding to the independent variable is replaced by the independent variable to obtain a new independent variable; the dependent variable is replaced by the residual matrix corresponding to the dependent variable to obtain a new dependent variable, and the step of extracting the principal component from the independent variable matrix E0 and the dependent variable F0 is started until the accuracy of the obtained regression equation is greater than the expected accuracy, and the regression equation with an accuracy greater than the expected accuracy is determined as the target regression equation, and the target regression equation is used to represent the relationship between the seismic attribute information and the well logging information in the same period; For the time-lapse seismic attribute information of each period in the multiple periods of the target well, the time-lapse seismic attribute information of the period is substituted into the target regression equation as an independent variable, and the obtained dependent variable is used as the well logging information of the period.

2. The method according to claim 1, characterized in that There are multiple target wells, and based on the time-lapse seismic attribute information of the target wells in multiple periods, the partial least squares regression method is used to predict the logging information of the multiple periods. The method further includes: Calculating the correlation of time-lapse seismic attribute information of any target well among the plurality of target wells in two periods among the plurality of periods; When the correlation of the time-lapse seismic attribute information of any target well in the two periods is greater than the correlation threshold, the target well is eliminated; Based on the time-lapse seismic attribute information of the target well in the first period, the partial least squares regression method is used to predict the logging information of the first period.

3. The method according to claim 1, characterized in that The method further comprises: Establishing a seismic attribute logging curve of the target well based on the time-lapse seismic attribute information of the target well in multiple periods; Output the seismic attribute logging curve of each target well.

4. The method according to claim 1, characterized in that: The output prediction of the logging information of the multiple periods includes: Based on the predicted logging information of the multiple periods, generating a logging curve of the logging information of each period; Output the well log curves for each of the stated periods.

5. The method according to any one of claims 1 to 4, characterized in that: The time-lapse seismic attribute information includes at least one of the following: root mean square amplitude, instantaneous frequency, dominant frequency, apparent polarity, instantaneous phase, amplitude envelope or instantaneous phase cosine.

6. The method according to claim 1, characterized in that The method further comprises: Based on the logging information of the multiple periods, a reservoir dynamic geological model reflecting the reservoir state of the target well is generated, and the reservoir dynamic geological model is output; Alternatively, based on the logging information of the multiple periods, a time-varying numerical simulation model reflecting the time-varying state of the logging data of the target well is generated, and the time-varying numerical simulation model is output.

7. A well logging information prediction device, characterized in that: The device comprises: An acquisition module, used to acquire time-lapse seismic attribute information of a target well in multiple periods, wherein the multiple periods are different from each other; A prediction module, used to predict the logging information of the target well in multiple periods by using a partial least squares regression method based on the time-lapse seismic attribute information of the target well in multiple periods; An output module, used for outputting the predicted logging information of the multiple periods; The prediction module is used to predict the logging information of the first period based on the time-lapse seismic attribute information of the target well by using a partial least squares regression method, where the first period is any one of the multiple periods; The prediction module is used to: Set the independent variable X and the dependent variable Y. The independent variable X includes p elements, and each element has n sample data. The dependent variable Y includes n elements. Both p and n are positive integers. Then the independent variable X and the dependent variable Y satisfy the following formula: The p elements included in the independent variable X are determined based on the known time-lapse seismic attribute information of the target well in the target period, the p elements include p of the root mean square amplitude, instantaneous frequency, main frequency, apparent polarity, instantaneous phase, amplitude envelope or instantaneous phase cosine, and the n sample data of each element are obtained by sampling the element in the first period n times; After obtaining the independent variable X and the dependent variable Y, the independent variable X and the dependent variable Y are standardized to obtain the standardized matrix X and the standardized vector Y; the standardized matrix X is recorded as E0, E0 is an n×p prediction matrix, and the standardized vector Y is recorded as F0, F0 is an n×1 response matrix; the element x* in the i-th row and j-th column of the standardized matrix ij Satisfies the following formula: in, (j=1,2,...,p),n>1;x ij represents the element in row i and column j in E0, x j represents the element in the jth column of E0, represents the average value of the elements in the jth column, and var represents the integer part; The principal components are extracted from the independent variable matrix E0 and the dependent variable F0. The extracted principal components are the first principal components t1 and u1, where the covariance of t1 and u1 reaches the maximum. The first principal components t1 and u1 satisfy the following formulas respectively: t1=Xw1+w1x1+w2x2+…w p x p ; Among them, cov means covariance, E 01 To E 0p They are x1 to x p Normalized value; A regression relationship equation is established between the principal component and the independent variable and the dependent variable, and the regression relationship equation satisfies the following first regression formula: E0=t1p1+E1;F0=u1r1+F1;where p1 and r1 represent regression coefficients, E′0 is the residual matrix of E0, F′0 is the residual matrix of F0, E1 is the new independent variable, and F1 is the new dependent variable; Based on the determined regression relationship equation, a regression equation is established; if the regression equation reaches the expected accuracy, the regression process is stopped; if the regression equation does not reach the expected accuracy, the residual matrix corresponding to the independent variable is replaced by the independent variable to obtain a new independent variable; the dependent variable is replaced by the residual matrix corresponding to the dependent variable to obtain a new dependent variable, and the step of extracting the principal component from the independent variable matrix E0 and the dependent variable F0 is started until the accuracy of the obtained regression equation is greater than the expected accuracy, and the regression equation with an accuracy greater than the expected accuracy is determined as the target regression equation, and the target regression equation is used to represent the relationship between the seismic attribute information and the well logging information in the same period; For the time-lapse seismic attribute information of each period in the multiple periods of the target well, the time-lapse seismic attribute information of the period is substituted into the target regression equation as an independent variable, and the obtained dependent variable is used as the well logging information of the period.

8. The device according to claim 7, characterized in that There are multiple target wells, and the partial least squares regression method is used based on the time-lapse seismic attribute information of multiple periods of the target wells to predict the logging information of the multiple periods. The device also includes: A calculation module, used for calculating the correlation of the time-lapse seismic attribute information of every two target wells among the plurality of target wells in a first period, wherein the first period is any period among the plurality of periods; A judgment module, configured to determine any two target wells as associated wells when the correlation of the time-lapse seismic attribute information of any two target wells in the first period is greater than a correlation threshold; The prediction submodule predicts the logging information of the first period by using a partial least squares regression method based on the time-lapse seismic attribute information of the associated wells in the first period.

9. The device according to claim 7, characterized in that The well logging information prediction device also includes: A generating module, configured to generate a reservoir dynamic geological model reflecting the reservoir state of the target well based on the well logging information of the plurality of periods, and output the reservoir dynamic geological model; Alternatively, based on the logging information of the multiple periods, a time-varying numerical simulation model reflecting the time-varying state of the logging data of the target well is generated, and the time-varying numerical simulation model is output.

Images