Ground stress change prediction method and system based on stress relief method

Abstract

The present application relates to the field of data prediction technology, and specifically to a method and system for predicting ground stress changes based on a stress relief method, which specifically includes: constructing the maximum lateral pressure discrete convergence significance and the maximum lateral pressure convergence significance by analyzing the variation law of the ratio of the maximum horizontal principal stress to the vertical stress in the ground stress data with depth, respectively; constructing the comprehensive lateral pressure difference coefficient by analyzing the differences between the ground stress data in different burial depths, and constructing the stress data noise coefficient by combining the comprehensive lateral pressure convergence significance constructed by the maximum lateral pressure comprehensive convergence significance and the minimum lateral pressure comprehensive convergence significance; completing the denoising of the initial ground stress data set according to the stress data noise coefficient, and completing the training of the ground stress change prediction model based on the denoised initial ground stress data set. Thereby, the ground stress change prediction is realized, and the accuracy of the ground stress change prediction model in predicting ground stress data is improved.

Description

Technical Field

[0001] The present application relates to the technical field of data prediction, and in particular to a method and system for predicting ground stress changes based on a stress relief method. Background Art

[0002] Geostress refers to the internal stress of the earth's internal rocks due to crustal movement, plate tectonic activity, geothermal gradient and other factors. By predicting and studying geostress, it can provide important scientific basis for earthquake prediction, large-scale engineering construction of tunnels and mines, and oil and natural gas resource development, reduce engineering risks and improve resource development efficiency.

[0003] The stress relief method is a commonly used geostress measurement method with the longest development time and relatively mature technology. In the process of geostress measurement, due to the influence of factors such as inaccurate calibration of measuring instruments, changes in the environment such as temperature and humidity, and deformation caused by cracks in the rock mass, noise data will be introduced into the collected geostress data. The noise data in the geostress data may cover up or distort the real geostress data, thereby reducing the accuracy of geostress data prediction. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a method and system for predicting ground stress changes based on the stress relief method. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a method for predicting ground stress changes based on a stress relief method, the method comprising the following steps:

[0006] Collect the geostress data at each geostress measuring point at each time to construct an initial geostress data set; construct a first geostress data set based on the normalized values ​​of all geostress data; the geostress data at each location include the maximum horizontal principal stress, the minimum horizontal principal stress, the vertical stress and the burial depth data;

[0007] In the first geostress data set, the maximum horizontal lateral pressure coefficient and the minimum horizontal lateral pressure coefficient of the stress data of each location are obtained based on the maximum horizontal principal stress, the minimum horizontal principal stress and the vertical stress of the stress data of each location; based on the burial depth values ​​of all geostress data, data clusters with similar burial depths are obtained, and all data clusters with similar burial depths are taken as the first total set with similar burial depths;

[0008] Based on the variation correlation between the burial depth data of all the ground stress data in the first burial depth similar total set and the maximum horizontal lateral pressure coefficient, the maximum lateral pressure convergence degree and the maximum lateral pressure discrete convergence degree of the first burial depth similar total set are constructed;

[0009] Based on the maximum lateral pressure convergence degree and the maximum lateral pressure discrete convergence degree, the maximum lateral pressure convergence significance and the maximum lateral pressure discrete convergence significance of each data cluster with similar burial depth are calculated respectively;

[0010] Based on the maximum lateral pressure convergence significance and the maximum lateral pressure discrete convergence significance, the maximum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is obtained; based on the burial depth data of all ground stress data in the first burial depth similar total set and the minimum horizontal lateral pressure coefficient, the minimum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is obtained;

[0011] Based on the maximum lateral pressure comprehensive convergence significance and the minimum lateral pressure comprehensive convergence significance, the comprehensive lateral pressure convergence significance of each data cluster with similar burial depth is obtained;

[0012] Based on the difference of the maximum horizontal lateral pressure coefficient and the minimum horizontal lateral pressure coefficient between different ground stress data in each data cluster with similar burial depth, the comprehensive lateral pressure difference coefficient of stress data in each place is obtained;

[0013] Based on the convergence significance of comprehensive lateral pressure and the comprehensive lateral pressure difference coefficient, the noise coefficient of stress data in various places is obtained;

[0014] The initial geostress data set is denoised according to the noise coefficient, and a geostress change prediction model of the geostress data is obtained according to the denoised geostress data.

[0015] In one embodiment, the maximum horizontal lateral pressure coefficient and the minimum horizontal lateral pressure coefficient of the stress data of each region are obtained by:

[0016] The ratio of the maximum horizontal principal stress to the vertical stress of each geostress data is taken as the maximum horizontal lateral pressure coefficient of each geostress data; the ratio of the minimum horizontal principal stress to the vertical stress of each geostress data is taken as the minimum horizontal lateral pressure coefficient of each geostress data.

[0017] In one embodiment, the process of obtaining the maximum lateral pressure convergence and the maximum lateral pressure discrete convergence of the first total set of similar burial depths is as follows:

[0018] For each data cluster with similar burial depth, the mean of the burial depth data of all the geostress data in the data cluster with similar burial depth is taken as the stratum burial depth of the data cluster with similar burial depth, the mean of the maximum horizontal lateral pressure coefficient of all the geostress data in the data cluster with similar burial depth is recorded as the maximum lateral pressure mean of the data cluster with similar burial depth, and the coefficient of variation of the maximum horizontal lateral pressure coefficient of all the geostress data in the data cluster with similar burial depth is recorded as the maximum lateral pressure dispersion of the data cluster with similar burial depth;

[0019] A rectangular coordinate system is constructed based on the maximum lateral pressure mean and the burial depth of the formation; a curve fitting is performed on the coordinate points generated by the maximum lateral pressure mean and the burial depth of the formation of all data clusters with similar burial depths in the rectangular coordinate system, and the mean square error between all coordinate points and the fitting curve is calculated; the sum of the mean square error and the adjustment factor is calculated, and the inverse of the sum is used as the maximum lateral pressure convergence degree of the first total set of similar burial depths;

[0020] Based on the maximum lateral pressure discreteness and stratum burial depth of all data clusters with similar burial depths, the maximum lateral pressure discreteness convergence of the first total set with similar burial depths is obtained by adopting the same calculation method as the maximum lateral pressure convergence.

[0021] In one embodiment, the process of obtaining the maximum lateral pressure convergence significance and the maximum lateral pressure discrete convergence significance of each data cluster with similar burial depth is as follows:

[0022] For any data cluster with similar burial depth, a set obtained by removing any data cluster with similar burial depth from the first total set with similar burial depth is recorded as a data cluster missing set of any data cluster with similar burial depth;

[0023] In any of the missing data cluster sets of data clusters with similar burial depths, based on all the data clusters with similar burial depths, the maximum lateral pressure convergence degree of the missing data cluster set is obtained by using the same calculation method as the maximum lateral pressure convergence degree;

[0024] The maximum lateral pressure convergence degree of the first total set of similar burial depths is recorded as the first convergence degree, and the maximum lateral pressure convergence degree of the data cluster missing set of any data cluster of similar burial depths is recorded as the second convergence degree; the ratio of the result of subtracting the first convergence degree from the second convergence degree to the first convergence degree is recorded as the maximum lateral pressure convergence significance of any data cluster of similar burial depths;

[0025] Based on the maximum lateral pressure discrete convergence degree, the maximum lateral pressure discrete convergence significance degree of each data cluster with similar burial depth is obtained by adopting the same calculation method as the maximum lateral pressure convergence significance degree.

[0026] In one embodiment, the process of obtaining the maximum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is as follows:

[0027] The product of the maximum lateral pressure convergence significance of each data cluster with similar burial depth and the maximum lateral pressure discrete convergence significance is taken as the maximum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth.

[0028] In one embodiment, the process of obtaining the comprehensive lateral pressure convergence significance of each data cluster with similar burial depth is as follows:

[0029] Based on the burial depth data and the minimum horizontal lateral pressure coefficient of all geostress data in the first total set of similar burial depths, the minimum lateral pressure comprehensive convergence significance of each data cluster with similar burial depths is obtained by adopting the same calculation method as the comprehensive convergence significance of the maximum lateral pressure; the product of the comprehensive convergence significance of the maximum lateral pressure and the comprehensive convergence significance of the minimum lateral pressure of each data cluster with similar burial depths is recorded as the comprehensive lateral pressure convergence significance of each data cluster with similar burial depths.

[0030] In one embodiment, the process of obtaining the comprehensive lateral pressure difference coefficient of the stress data of each region is as follows:

[0031] In each data cluster with similar burial depth, the difference between the maximum horizontal lateral pressure coefficient of any geostress data and each of the other geostress data is calculated and recorded as the first difference; the average of all the first differences of any geostress data is recorded as the maximum lateral pressure difference coefficient of any geostress data; the difference between the minimum horizontal lateral pressure coefficient of any geostress data and each of the other geostress data is calculated and recorded as the second difference; the average of all the second differences of any geostress data is recorded as the minimum lateral pressure difference coefficient of any geostress data;

[0032] The product of the maximum lateral pressure difference coefficient and the minimum lateral pressure difference coefficient of the stress data in various places is taken as the comprehensive lateral pressure difference coefficient of the stress data in various places.

[0033] In one embodiment, the process of obtaining the noise coefficient of stress data from various locations is as follows:

[0034] For each data cluster with similar burial depth, the product of the comprehensive lateral pressure difference coefficient of each geostress data in the data cluster with similar burial depth and the comprehensive lateral pressure convergence significance of the data cluster with similar burial depth is taken as the noise coefficient of each geostress data.

[0035] In one embodiment, the initial geostress data set is denoised according to the noise coefficient, and a geostress change prediction model of the geostress data is obtained according to the denoised geostress data, specifically:

[0036] The noise coefficients of all geostress data are segmented using a threshold segmentation algorithm to obtain a segmentation threshold; the geostress data in the initial geostress data set whose noise coefficients are greater than the segmentation threshold are regarded as noisy data, and after the noisy data in the initial geostress data set are removed, the result is recorded as the initial geostress data denoising set;

[0037] Half of the data in the denoised set of initial geostress data is used as the training set, and the other half of the data is used as the test set. The prediction model is trained based on the training set, and the test set is used to verify the model. The obtained prediction model is used as the geostress change prediction model.

[0038] In the second aspect, an embodiment of the present application also provides a ground stress change prediction system based on the stress relief method, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, the steps of any one of the above methods are implemented.

[0039] The embodiments of the present application have at least the following beneficial effects:

[0040] The present application constructs the maximum lateral pressure discrete convergence significance and the maximum lateral pressure comprehensive convergence significance by analyzing the variation law of the ratio of the maximum horizontal principal stress to the vertical stress in the ground stress data with depth, and the beneficial effect is that the distinction between the ground stress data that conforms to the variation law of the ratio of the maximum horizontal principal stress to the vertical stress and the ground stress data that does not conform to the law is improved;

[0041] The comprehensive lateral pressure difference coefficient is constructed by analyzing the differences between the ground stress data at different burial depths, and the stress data noise coefficient is constructed by combining the comprehensive lateral pressure convergence significance constructed by the maximum lateral pressure comprehensive convergence significance and the minimum lateral pressure comprehensive convergence significance. The beneficial effect is that the recognition accuracy of the ground stress data containing noise in the ground stress data set is improved;

[0042] The initial geostress data set is denoised according to the noise coefficient of the stress data, and the geostress change prediction model is trained based on the denoised initial geostress data set. The beneficial effect is that the influence of noise data in the geostress data on the real change of the geostress data is reduced, and the accuracy of the geostress change prediction model in predicting the geostress data is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A flowchart of the steps of a method for predicting ground stress changes based on a stress relief method provided in one embodiment of the present application;

[0045] Figure 2 Schematic diagram of the specific steps for constructing the maximum lateral pressure convergence and the maximum lateral pressure discrete convergence for the first total set of similar burial depths. DETAILED DESCRIPTION

[0046] In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, describes in detail the method and system for predicting ground stress changes based on the stress relief method proposed in the present application, its specific implementation method, structure, features and effects. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0047] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0048] The specific scheme of the method and system for predicting ground stress changes based on the stress relief method provided by the present application is described in detail below with reference to the accompanying drawings.

[0049] See also Figure 1 , which shows a flowchart of a method for predicting ground stress changes based on a stress relief method provided by an embodiment of the present application, the method comprising the following steps:

[0050] Step S1, collecting the geostress data at each geostress measuring point at each time to construct an initial geostress data set; constructing a first geostress data set based on the normalized values ​​of all geostress data; the geostress data of each location include maximum horizontal principal stress, minimum horizontal principal stress, vertical stress and burial depth data.

[0051] This application takes the prediction of geostress changes during coal mining as an example, and uses a geostress measurement device based on the geostress relief method to collect geostress data at all geostress measuring points. The geostress data are the maximum horizontal principal stress, the minimum horizontal principal stress, the vertical stress and the buried depth data, and the buried depth is the stratum depth where the geostress measuring point is located. The geostress data collected within one week is obtained, where the sampling time interval of the geostress data is 0.5h, and the sampling time and sampling time interval of the geostress data can be set by the implementer. The set of geostress data at all geostress measuring points at all times is recorded as the initial geostress data set.

[0052] In order to avoid the influence of dimension, the geostress data collected at each geostress measuring point at each time are normalized by the "max-min" normalization method. It should be noted that there are many existing normalization methods, and the implementer can also use other normalization methods to normalize the geostress data. The set of geostress data at all geostress measuring points at all times after normalization is recorded as the first geostress data set. Each element in the first geostress data set represents each geostress data, and each geostress data includes the maximum horizontal principal stress, the minimum horizontal principal stress, the vertical stress and the burial depth data.

[0053] Step S2, based on the maximum horizontal principal stress, minimum horizontal principal stress and vertical stress of the stress data of each location, obtain the maximum horizontal lateral pressure coefficient and the minimum horizontal lateral pressure coefficient of the stress data of each location; based on the burial depth values ​​of all the ground stress data, divide them to obtain data clusters with similar burial depths, and construct a first total set of data with similar burial depths based on all the data clusters with similar burial depths.

[0054] As an example, this step may include the following steps:

[0055] In general, when the burial depth is shallow, the stress state of the crust will be affected by various factors such as surface processes such as weathering and erosion, the ups and downs of different terrains, and the tectonic stress caused by crustal movement and fault activity, resulting in the maximum horizontal principal stress being greater than the vertical stress, the minimum horizontal principal stress being greater than the vertical stress, and the ratio of the maximum horizontal principal stress to the vertical stress, hereinafter referred to as the maximum horizontal lateral pressure coefficient, and the ratio of the minimum horizontal principal stress to the vertical stress, hereinafter referred to as the minimum horizontal lateral pressure coefficient, both have strong discreteness; with the increase of burial depth, the values ​​of the maximum horizontal principal stress and the vertical stress will gradually approach, making the value and discreteness of the maximum horizontal lateral pressure coefficient have a convergence trend, and with the increase of burial depth, the vertical stress will gradually exceed the minimum horizontal principal stress, and the gap between the two will become larger and larger, which will also make the value and discreteness of the minimum horizontal lateral pressure coefficient have a convergence trend. However, the values ​​of the in-situ stress data in the presence of noise are usually random, and this randomness will destroy this convergence trend in the in-situ stress data.

[0056] (1) The ratio of the maximum horizontal principal stress to the vertical stress of each geostress data is taken as the maximum horizontal lateral pressure coefficient of each geostress data; the ratio of the minimum horizontal principal stress to the vertical stress of each geostress data is taken as the minimum horizontal lateral pressure coefficient of each geostress data;

[0057] (2) Take the burial depth values ​​of all geostress data as input, use the natural breakpoint algorithm to divide all geostress data, and record the output N sets as data clusters with similar burial depths to characterize the set of geostress data with similar burial depth values ​​in the geostress data set; record the set consisting of all data clusters with similar burial depths as the first total set with similar burial depths. The natural breakpoint algorithm is a well-known technology, and the specific process will not be repeated; it should be noted that the value of N can be set by the implementer, and the value of N is set to 20 in this embodiment; this embodiment only provides a method for dividing based on the burial depth value of geostress data. There are many existing data division methods. The implementer can also use other data division algorithms to divide the geostress data, and this embodiment does not make specific restrictions.

[0058] Step S3, constructing the maximum lateral pressure convergence degree and the maximum lateral pressure discrete convergence degree of the first burial depth similar total set based on the change correlation between the burial depth data of all the geostress data in the first burial depth similar total set and the maximum horizontal lateral pressure coefficient.

[0059] As an example, this step may include the following steps:

[0060] (1) Based on the burial depth data, the stratum burial depth of each data cluster with similar burial depth is calculated; based on the maximum horizontal lateral pressure coefficient, the maximum lateral pressure mean of each data cluster with similar burial depth is calculated; based on the dispersion degree of the maximum horizontal lateral pressure coefficient, the maximum lateral pressure dispersion of each data cluster with similar burial depth is calculated:

[0061] For each data cluster with similar burial depth, the mean of the burial depth data of all the geostress data in the data cluster with similar burial depth is taken as the stratum burial depth of the data cluster with similar burial depth, which is used to characterize the stratum depth of the overall data collection position of the geostress data when collecting the geostress data in the data cluster with similar burial depth;

[0062] The mean value of the maximum horizontal lateral pressure coefficient of all the ground stress data in the data cluster with similar burial depth is recorded as the mean value of the maximum lateral pressure of the data cluster with similar burial depth, so as to characterize the overall distribution of the maximum horizontal lateral pressure coefficient of all the ground stress data in the data cluster with similar burial depth;

[0063] The coefficient of variation of the maximum horizontal lateral pressure coefficient of all geostress data in a data cluster with similar burial depth is recorded as the maximum lateral pressure dispersion of the data cluster with similar burial depth, so as to characterize the degree of dispersion of the maximum horizontal lateral pressure coefficient of all geostress data in the data cluster with similar burial depth.

[0064] (2) A rectangular coordinate system is constructed based on the maximum lateral pressure mean and the burial depth of the formation. A curve fitting algorithm is used to fit the coordinate points generated by the maximum lateral pressure mean and the burial depth of the formation in the rectangular coordinate system for all data clusters with similar burial depths, and the mean square error between all coordinate points and the fitting curve is calculated:

[0065] As the burial depth increases, the value and discreteness of the maximum horizontal lateral pressure coefficient tend to converge, and the function The ordinate y of a point on the function curve will approach the constant value b as the abscissa x increases. Therefore, first, a rectangular coordinate system is constructed with the maximum lateral pressure mean and the burial depth of the formation as the abscissa and ordinate, respectively. Then, each coordinate point in the rectangular coordinate system is obtained by the maximum lateral pressure mean and the burial depth of each data cluster with similar burial depth. After that, all the coordinate points in the rectangular coordinate system are fitted by the least squares curve fitting algorithm to obtain a fitting curve, in which the function is used. As the fitting function, finally, the mean square error between all coordinate points and the fitting curve is calculated. The mean square error and the least square curve fitting algorithm are well-known technologies, and the specific calculation process will not be repeated here.

[0066] Each coordinate point in the above coordinate system represents the data distribution of the maximum horizontal lateral pressure coefficient of the ground stress data in the data clusters with similar burial depths at different stratum depths as the burial depth changes; the above mean square error can represent the curve fitting effect of the coordinate point in the coordinate system. The smaller the value of the mean square error, the smaller the difference between the data point in the coordinate system and the fitting result, and the better the curve fitting effect of the data point in the coordinate system.

[0067] (3) Obtaining the maximum lateral pressure convergence of the first total set of similar burial depths based on the mean square error:

[0068] The sum of the mean square error and the adjustment factor is calculated, and the reciprocal of the sum is used as the maximum lateral pressure convergence degree of the first total set of similar burial depths.

[0069] The adjustment factor is a positive number preset by humans to prevent the denominator from being zero due to the mean square error being zero. The value of the adjustment factor can be set by the implementer. In this embodiment, the adjustment factor is set to 0.01. The smaller the mean square error, the better the curve fitting effect of the data points in the coordinate system, and the closer the maximum horizontal lateral pressure coefficient changes with the depth to the function. The curve shape shows that the value of the maximum horizontal lateral pressure coefficient of the ground stress data in the ground stress data set has a more convergent trend with the increase of burial depth, and the value of the maximum lateral pressure convergence degree of the total set with similar first burial depth is greater.

[0070] (4) Based on the maximum lateral pressure discreteness and the burial depth of all data clusters with similar burial depths, the maximum lateral pressure discreteness convergence of the first total set with similar burial depths is obtained by adopting the same calculation method as the maximum lateral pressure convergence.

[0071] Step S4, based on the maximum lateral pressure convergence degree and the maximum lateral pressure discrete convergence degree, respectively calculate the maximum lateral pressure convergence significance and the maximum lateral pressure discrete convergence significance of each data cluster with similar burial depth.

[0072] As an example, this step may include the following steps:

[0073] (1) Based on the maximum lateral pressure convergence degree, the maximum lateral pressure convergence significance of each data cluster with similar burial depth is calculated:

[0074] For any data cluster with similar burial depth, a set obtained by removing any data cluster with similar burial depth from the first total set with similar burial depth is recorded as a data cluster missing set of any data cluster with similar burial depth;

[0075] In any of the missing data cluster sets of data clusters with similar burial depths, based on all the data clusters with similar burial depths, the maximum lateral pressure convergence degree of the missing data cluster set is obtained by using the same calculation method as the maximum lateral pressure convergence degree;

[0076] The maximum lateral pressure convergence degree of the total set with similar burial depths is recorded as the first convergence degree, and the maximum lateral pressure convergence degree of the data cluster missing set of any data cluster with similar burial depths is recorded as the second convergence degree; the ratio of the result of subtracting the first convergence degree from the second convergence degree to the first convergence degree is recorded as the maximum lateral pressure convergence significance of any data cluster with similar burial depths.

[0077] The greater the degree to which the value of the second convergence degree exceeds the value of the first convergence degree, the more obvious the convergence trend of the value of the maximum horizontal lateral pressure coefficient of the geostress data in the removed data cluster with similar burial depth after the first geostress data set removes the geostress data in a data cluster with similar burial depth as the burial depth increases, compared with the change in the value of the maximum horizontal lateral pressure coefficient of the geostress data in the removed data cluster with similar burial depth. The greater the convergence significance of the maximum lateral pressure of the removed data cluster with similar burial depth.

[0078] (2) Based on the maximum lateral pressure discrete convergence degree, the maximum lateral pressure discrete convergence significance degree of each data cluster with similar burial depth is obtained by adopting the same calculation method as the maximum lateral pressure convergence significance degree.

[0079] After removing the geostress data in any data cluster with similar burial depth from the first geostress data set, the remaining geostress data has a more obvious convergence trend than the maximum lateral pressure discreteness of the geostress data in the geostress data set as the burial depth increases. The larger the value of the maximum lateral pressure discrete convergence significance, the greater the convergence significance.

[0080] Step S5, based on the maximum lateral pressure convergence significance and the maximum lateral pressure discrete convergence significance, the maximum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is obtained; based on the burial depth data of all ground stress data in the first burial depth similar total set and the minimum horizontal lateral pressure coefficient, the minimum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is obtained.

[0081] As an example, this step may include the following steps:

[0082] The product of the maximum lateral pressure convergence significance of each data cluster with similar burial depth and the maximum lateral pressure discrete convergence significance is taken as the maximum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth;

[0083] Based on the burial depth data and the minimum horizontal lateral pressure coefficient of all the geostress data in the first burial depth similar total set, the minimum lateral pressure comprehensive convergence significance of each burial depth similar data cluster is obtained by adopting the same calculation method as the maximum lateral pressure comprehensive convergence significance.

[0084] After removing the geostress data in the geostress data set, the remaining geostress data has a more obvious convergence trend than the maximum and minimum horizontal lateral pressure coefficients and the discreteness of the geostress data in the geostress data set with increasing burial depth, and the greater the significance value.

[0085] Step S6, obtaining the comprehensive lateral pressure convergence significance of each data cluster with similar burial depth based on the maximum lateral pressure comprehensive convergence significance and the minimum lateral pressure comprehensive convergence significance.

[0086] As an example, this step may include the following steps:

[0087] The product of the maximum lateral pressure comprehensive convergence significance and the minimum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is recorded as the comprehensive lateral pressure convergence significance of each data cluster with similar burial depth.

[0088] After removing the geostress data in a data cluster with similar burial depth from the first geostress data set, the remaining geostress data have a more obvious convergence trend than the maximum and minimum horizontal lateral pressure coefficient values ​​and the discreteness of the geostress data in the complete first geostress data set as the burial depth increases. The greater the value of the comprehensive lateral pressure convergence significance, the greater the degree to which the geostress data in the data cluster with similar burial depth destroys the convergence trend of the maximum and minimum horizontal lateral pressure coefficient values ​​and the discreteness of the geostress data in the complete geostress data set, and the more likely it is that there is geostress data containing noise in the data cluster with similar burial depth.

[0089] Step S7, obtaining the comprehensive lateral pressure difference coefficient of the stress data in each location based on the difference in the maximum horizontal lateral pressure coefficient and the difference in the minimum horizontal lateral pressure coefficient between different ground stress data in each data cluster with similar burial depth.

[0090] As an example, this step may include the following steps:

[0091] Since rocks at the same stratum depth usually have similar geological characteristics, such as rock type, density, elastic modulus, etc., the geostress data collected at similar burial depths are often close, so that the maximum horizontal lateral pressure coefficient and the minimum horizontal lateral pressure coefficient of the geostress data at the same stratum depth in the complete geostress data set are also close. However, the randomness of geostress data in geostress data with noise will destroy this data distribution feature.

[0092] Based on the above analysis, in each data cluster with similar burial depth, the difference between the maximum horizontal lateral pressure coefficient of any geostress data and each of the remaining geostress data is calculated and recorded as the first difference; the average of all the first differences of any geostress data is recorded as the maximum lateral pressure difference coefficient of any geostress data; the difference between the minimum horizontal lateral pressure coefficient of any geostress data and each of the remaining geostress data is calculated and recorded as the second difference; the average of all the second differences of any geostress data is recorded as the minimum lateral pressure difference coefficient of any geostress data;

[0093] The product of the maximum lateral pressure difference coefficient and the minimum lateral pressure difference coefficient of the stress data in various places is taken as the comprehensive lateral pressure difference coefficient of the stress data in various places.

[0094] The greater the difference between the maximum and minimum horizontal lateral pressure coefficients between the local stress data and the other stress data in the data cluster with similar burial depth, the greater the value of the comprehensive lateral pressure difference coefficient, and the greater the possibility that the stress data is the stress data containing noise in the data cluster with similar burial depth.

[0095] Step S8, obtaining the noise coefficient of stress data in various places based on the comprehensive lateral pressure convergence significance and the comprehensive lateral pressure difference coefficient.

[0096] As an example, this step may include the following steps:

[0097] For each data cluster with similar burial depth, the product of the comprehensive lateral pressure difference coefficient of each geostress data in the data cluster with similar burial depth and the comprehensive lateral pressure convergence significance of the data cluster with similar burial depth is taken as the noise coefficient of each geostress data.

[0098] The greater the noise coefficient of the geostress data is, the greater the possibility that the geostress data is geostress data containing noise in the complete geostress data set is.

[0099] Step S9, denoising the initial geostress data set according to the noise coefficient, and obtaining a geostress change prediction model of the geostress data according to the denoised geostress data.

[0100] As an example, this step may include the following steps:

[0101] The noise coefficient of all geostress data is used as the input of the maximum inter-class variance algorithm, and the segmentation threshold is output, wherein the maximum inter-class variance algorithm is a well-known technology, and the specific process is not repeated. The geostress data with a noise coefficient greater than the segmentation threshold in the initial geostress data set is taken as noisy data, and the noisy data is geostress data that may have noise. The noisy data in the initial geostress data set is removed, and the removal result of the initial geostress data set is recorded as the initial geostress data denoising set.

[0102] Half of the data in the denoised set of initial geostress data is used as a training set, and the other half of the data is used as a test set. The training set is used as the training set of the support vector machine model to train the prediction model of geostress change, wherein the radial basis kernel function and the gradient descent method are used as the kernel function and optimization algorithm of the support vector machine model respectively, and the support vector machine model is verified using the test set, wherein the accuracy rate is used as the model performance verification index, wherein the training of the support vector machine model is a well-known technology, and the specific process is not repeated here. The prediction model obtained by training is used as the prediction model of geostress change to complete the prediction of geostress data.

[0103] The schematic diagram of the specific steps for constructing the maximum lateral pressure convergence and the maximum lateral pressure discrete convergence of the first total set of similar burial depths is as follows: Figure 2 shown.

[0104] Based on the same inventive concept as the above-mentioned method, an embodiment of the present application also provides a ground stress change prediction system based on the stress relief method, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned ground stress change prediction methods based on the stress relief method are implemented.

[0105] In summary, the embodiment of the present application provides a method for predicting ground stress changes based on the stress relief method, by analyzing the variation law of the ratio of the maximum horizontal principal stress to the vertical stress in the ground stress data with depth, respectively constructing the maximum lateral pressure discrete convergence significance and the maximum lateral pressure convergence significance to construct the maximum lateral pressure comprehensive convergence significance, which has the beneficial effect of improving the distinction between the ground stress data that conforms to the variation law of the ratio of the maximum horizontal principal stress to the vertical stress in the ground stress data set and the ground stress data that does not conform to the law;

[0106] The comprehensive lateral pressure difference coefficient is constructed by analyzing the differences between the ground stress data at different burial depths, and the stress data noise coefficient is constructed by combining the comprehensive lateral pressure convergence significance constructed by the maximum lateral pressure comprehensive convergence significance and the minimum lateral pressure comprehensive convergence significance. The beneficial effect is that the recognition accuracy of the ground stress data containing noise in the ground stress data set is improved;

[0107] The initial geostress data set is denoised according to the noise coefficient of the stress data, and the geostress change prediction model is trained based on the denoised initial geostress data set. The beneficial effect is that the influence of noise data in the geostress data on the real change of the geostress data is reduced, and the accuracy of the geostress change prediction model in predicting the geostress data is improved.

[0108] It should be noted that the above sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0109] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0110] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for predicting ground stress changes based on stress relief method, characterized in that: The method comprises the following steps: Collect the geostress data at each geostress measuring point at each time to construct an initial geostress data set; construct a first geostress data set based on the normalized values ​​of all geostress data; the geostress data at each location include the maximum horizontal principal stress, the minimum horizontal principal stress, the vertical stress and the burial depth data; In the first geostress data set, the maximum horizontal lateral pressure coefficient and the minimum horizontal lateral pressure coefficient of the stress data of each location are obtained based on the maximum horizontal principal stress, the minimum horizontal principal stress and the vertical stress of the stress data of each location; based on the burial depth values ​​of all geostress data, data clusters with similar burial depths are obtained, and all data clusters with similar burial depths are taken as the first total set with similar burial depths; Based on the variation correlation between the burial depth data of all the ground stress data in the first burial depth similar total set and the maximum horizontal lateral pressure coefficient, the maximum lateral pressure convergence degree and the maximum lateral pressure discrete convergence degree of the first burial depth similar total set are constructed; Based on the maximum lateral pressure convergence degree and the maximum lateral pressure discrete convergence degree, the maximum lateral pressure convergence significance and the maximum lateral pressure discrete convergence significance of each data cluster with similar burial depth are calculated respectively; Based on the maximum lateral pressure convergence significance and the maximum lateral pressure discrete convergence significance, the maximum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is obtained; based on the burial depth data of all ground stress data in the first burial depth similar total set and the minimum horizontal lateral pressure coefficient, the minimum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is obtained; Based on the maximum lateral pressure comprehensive convergence significance and the minimum lateral pressure comprehensive convergence significance, the comprehensive lateral pressure convergence significance of each data cluster with similar burial depth is obtained; Based on the difference of the maximum horizontal lateral pressure coefficient and the minimum horizontal lateral pressure coefficient between different ground stress data in each data cluster with similar burial depth, the comprehensive lateral pressure difference coefficient of stress data in each place is obtained; Based on the convergence significance of comprehensive lateral pressure and the comprehensive lateral pressure difference coefficient, the noise coefficient of stress data in various places is obtained; The initial geostress data set is denoised according to the noise coefficient, and a geostress change prediction model of the geostress data is obtained according to the denoised geostress data.

2. The method for predicting ground stress changes based on the stress relief method according to claim 1, characterized in that: The process of obtaining the maximum horizontal lateral pressure coefficient and the minimum horizontal lateral pressure coefficient of the stress data of each place is as follows: The ratio of the maximum horizontal principal stress to the vertical stress of each geostress data is taken as the maximum horizontal lateral pressure coefficient of each geostress data; the ratio of the minimum horizontal principal stress to the vertical stress of each geostress data is taken as the minimum horizontal lateral pressure coefficient of each geostress data.

3. The method for predicting ground stress changes based on the stress relief method according to claim 1, characterized in that: The process of obtaining the maximum lateral pressure convergence and the maximum lateral pressure discrete convergence of the first total set of similar burial depths is as follows: For each data cluster with similar burial depth, the mean of the burial depth data of all the geostress data in the data cluster with similar burial depth is taken as the stratum burial depth of the data cluster with similar burial depth, the mean of the maximum horizontal lateral pressure coefficient of all the geostress data in the data cluster with similar burial depth is recorded as the maximum lateral pressure mean of the data cluster with similar burial depth, and the coefficient of variation of the maximum horizontal lateral pressure coefficient of all the geostress data in the data cluster with similar burial depth is recorded as the maximum lateral pressure dispersion of the data cluster with similar burial depth; A rectangular coordinate system is constructed based on the maximum lateral pressure mean and the burial depth of the formation; a curve fitting algorithm is used to fit the coordinate points generated by the maximum lateral pressure mean and the burial depth of the formation in the rectangular coordinate system of all data clusters with similar burial depths, and the mean square error between all coordinate points and the fitting curve is calculated; Calculating the sum of the mean square error and the adjustment factor, and taking the reciprocal of the sum as the maximum lateral pressure convergence of the first total set of similar burial depths; Based on the maximum lateral pressure discreteness and stratum burial depth of all data clusters with similar burial depths, the maximum lateral pressure discreteness convergence of the first total set with similar burial depths is obtained by adopting the same calculation method as the maximum lateral pressure convergence.

4. The method for predicting ground stress changes based on the stress relief method according to claim 1, characterized in that: The process of obtaining the maximum lateral pressure convergence significance and the maximum lateral pressure discrete convergence significance of each data cluster with similar burial depth is as follows: For any data cluster with similar burial depth, a set obtained by removing any data cluster with similar burial depth from the first total set with similar burial depth is recorded as a data cluster missing set of any data cluster with similar burial depth; In any of the missing data cluster sets of data clusters with similar burial depths, based on all the data clusters with similar burial depths, the maximum lateral pressure convergence degree of the missing data cluster set is obtained by using the same calculation method as the maximum lateral pressure convergence degree; The maximum lateral pressure convergence degree of the first total set of similar burial depths is recorded as the first convergence degree, and the maximum lateral pressure convergence degree of the data cluster missing set of any data cluster of similar burial depths is recorded as the second convergence degree; the ratio of the result of subtracting the first convergence degree from the second convergence degree to the first convergence degree is recorded as the maximum lateral pressure convergence significance of any data cluster of similar burial depths; Based on the maximum lateral pressure discrete convergence degree, the maximum lateral pressure discrete convergence significance degree of each data cluster with similar burial depth is obtained by adopting the same calculation method as the maximum lateral pressure convergence significance degree.

5. The method for predicting ground stress changes based on the stress relief method according to claim 1, characterized in that: The process of obtaining the maximum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth is as follows: The product of the maximum lateral pressure convergence significance of each data cluster with similar burial depth and the maximum lateral pressure discrete convergence significance is taken as the maximum lateral pressure comprehensive convergence significance of each data cluster with similar burial depth.

6. The method for predicting ground stress changes based on the stress relief method according to claim 1, characterized in that: The process of obtaining the comprehensive lateral pressure convergence significance of the data clusters with similar burial depths is as follows: Based on the burial depth data and the minimum horizontal lateral pressure coefficient of all geostress data in the first total set of similar burial depths, the minimum lateral pressure comprehensive convergence significance of each data cluster with similar burial depths is obtained by adopting the same calculation method as the comprehensive convergence significance of the maximum lateral pressure; the product of the comprehensive convergence significance of the maximum lateral pressure and the comprehensive convergence significance of the minimum lateral pressure of each data cluster with similar burial depths is recorded as the comprehensive lateral pressure convergence significance of each data cluster with similar burial depths.

7. The method for predicting ground stress changes based on the stress relief method according to claim 1, characterized in that: The process of obtaining the comprehensive lateral pressure difference coefficient of the stress data of various places is as follows: In each data cluster with similar burial depth, the difference between the maximum horizontal lateral pressure coefficient of any geostress data and each of the other geostress data is calculated and recorded as the first difference; the average of all the first differences of any geostress data is recorded as the maximum lateral pressure difference coefficient of any geostress data; Calculate the difference between the minimum horizontal lateral pressure coefficient of any one of the in-situ stress data and each of the remaining in-situ stress data, and record it as the second difference; Recording the average of all second differences of any one of the in-situ stress data as the minimum lateral pressure difference coefficient of any one of the in-situ stress data; The product of the maximum lateral pressure difference coefficient and the minimum lateral pressure difference coefficient of the stress data in various places is taken as the comprehensive lateral pressure difference coefficient of the stress data in various places.

8. The method for predicting ground stress changes based on the stress relief method according to claim 1, characterized in that: The process of obtaining the noise coefficient of stress data from various locations is as follows: For each data cluster with similar burial depth, the product of the comprehensive lateral pressure difference coefficient of each geostress data in the data cluster with similar burial depth and the comprehensive lateral pressure convergence significance of the data cluster with similar burial depth is taken as the noise coefficient of each geostress data.

9. The method for predicting ground stress changes based on the stress relief method according to claim 1, characterized in that: The method of denoising the initial geostress data set according to the noise coefficient and obtaining a geostress change prediction model of the geostress data according to the denoised geostress data is specifically as follows: The noise coefficients of all geostress data are segmented using a threshold segmentation algorithm to obtain a segmentation threshold; the geostress data in the initial geostress data set whose noise coefficients are greater than the segmentation threshold are regarded as noisy data, and after the noisy data in the initial geostress data set are removed, the result is recorded as the initial geostress data denoising set; Half of the data in the denoised set of initial geostress data is used as the training set, and the other half of the data is used as the test set. The prediction model is trained based on the training set, and the test set is used to verify the model. The obtained prediction model is used as the geostress change prediction model.

10. A system for predicting ground stress changes based on a stress relief method, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images