Method for preventing well leakage in geothermal drilling

Abstract

The invention provides a geothermal drilling well leakage prevention method which comprises the steps that a drilling historical data set is obtained, the drilling historical data set comprises well leakage data and non-well leakage data, and the well leakage data and the non-well leakage data have the same multiple original parameter characteristics; performing principal component analysis on a plurality of original parameter characteristics in the drilling historical data set to obtain parameter characteristics after dimension reduction; mapping the training set data into a multi-dimensional space according to the parameter features after dimension reduction; constructing a support vector machine model, and segmenting well leakage data and non-well leakage data in the multi-dimensional space through well leakage labels and non-well leakage labels in the training set data; and testing the accuracy of the support vector machine model by using the test set data, and correcting the support vector machine model. By constructing the model, whether different drilling data are in a well leakage state or not can be accurately judged, well leakage prevention is carried out in advance, and the possibility of well leakage is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of geothermal drilling and exploration, and in particular to a method for preventing well leakage in geothermal drilling. Background Art

[0002] Geothermal drilling is a drilling operation aimed at exploring or developing geothermal resources. It aims to guide underground geothermal energy to the surface through drilling technology for people to use. It involves multiple links and technical requirements. Through reasonable drilling design and fine construction operations, geothermal resources can be effectively explored and developed to provide more clean and renewable energy for human production and life. Geothermal resources are a clean and renewable energy, which is of great significance to alleviating energy pressure and reducing environmental pollution.

[0003] Well leakage is mainly caused by the pressure of the drilling fluid column being greater than the pore pressure or fracture pressure of the formation, resulting in a large amount of drilling fluid being lost into the formation. The thermal reservoirs encountered during geothermal drilling are often porous and fractured, and are low-pressure or normal-pressure formations. They are extremely fragile and easily cause mud leakage, making it impossible to continue drilling. In addition, due to improper drilling process measures, it is easy to generate a large excitation pressure in the formation. When the pressure exceeds the formation fracture pressure, it will cause well leakage. Alternatively, the intermediate casing is not lowered deep enough or is not lowered, resulting in high and low pressure formations in the same open hole section. The pressure of the drilling fluid column may exceed the pressure-bearing capacity of the low-pressure formation, causing well leakage.

[0004] At present, in order to prevent blowouts or well leakage caused by excessively high or low pressure, near-balanced drilling is achieved by selecting drilling fluids of appropriate density and type. However, this method only gradually increases the density of drilling fluids in high-pressure formations and gradually reduces the density of drilling fluids in low-pressure formations. It does not find the relevant factors causing well leakage at a deeper level, and cannot completely avoid formation ruptures caused by other environmental factors or parameter changes and other parameter characteristics. There is still a high risk of well leakage during geothermal exploration. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a method for preventing lost circulation in geothermal drilling, by pre-building a support vector machine model of drilling data, so as to prevent lost circulation caused by various parameter characteristics and avoid the occurrence of lost circulation.

[0006] To achieve the above-mentioned object of the invention, the present invention provides a method for preventing lost circulation in geothermal drilling, the method comprising:

[0007] Acquire a drilling history data set, wherein the drilling history data set includes lost circulation data and non-lost circulation data, and the lost circulation data and the non-lost circulation data have the same multiple original parameter features;

[0008] The lost circulation data are labeled as lost circulation, the non-lost circulation data are labeled as non-lost circulation, and the drilling history data set is randomly divided into training set data and test set data;

[0009] Perform principal component analysis on multiple original parameter features in the drilling history data set to obtain parameter features after dimensionality reduction;

[0010] According to the parameter characteristics after dimensionality reduction, the training set data is mapped into a multidimensional space;

[0011] A support vector machine model is constructed to separate the leakage data from the non-leakage data in the multidimensional space through the leakage labels and non-leakage labels in the training set data;

[0012] The test set data is used to test the accuracy of the support vector machine model and to modify the support vector machine model.

[0013] Furthermore, after the model is modified, the current drilling data is collected in real time, and the current drilling data is mapped to the multidimensional space according to the parameter characteristics after dimensionality reduction. By adjusting the data of the original parameter characteristics, the current drilling data is placed on the non-leakage data side of the support vector machine model.

[0014] Furthermore, if the current drilling data indicates that a lost circulation has occurred, the current drilling data is labeled as a lost circulation and added to the support vector machine model to train the model;

[0015] At the same time, the corresponding well leakage points are found according to the current drilling data, and plugging materials are added to the well leakage points to plug the leakage.

[0016] Furthermore, the principal component analysis of multiple parameter features in the drilling history data set includes:

[0017] S1: After data standardization processing of the data in the drilling history dataset, a multi-dimensional random vector matrix is ​​constructed;

[0018] S2: Calculate the covariance matrix of the multidimensional random vector matrix;

[0019] S3: Perform eigendecomposition on the covariance matrix to obtain a set of eigenvalues ​​and corresponding eigenvectors;

[0020] S4: According to the size of the eigenvalue, select the eigenvectors corresponding to the first k eigenvalues ​​from large to small as the principal components.

[0021] Furthermore, the data in the drilling history data set is subjected to data standardization processing, including:

[0022] Use dimensionless criteria to align numerical data;

[0023] Perform structured mapping on text data;

[0024] The aligned digital data and the structured mapped text data are normalized.

[0025] Furthermore, the construction of the multidimensional random vector matrix is ​​expressed as:

[0026]

[0027] Where X represents a multidimensional random vector matrix, x np represents the value corresponding to the nth sample and the pth parameter feature, X p Represents the vector corresponding to the p-th parameter feature.

[0028] Furthermore, the process of constructing the support vector machine model includes:

[0029] According to the distribution of lost circulation data and non-lost circulation data in multidimensional space, a classification hyperplane of lost circulation data and non-lost circulation data is obtained, and a function expression of the classification hyperplane is obtained, one side of the classification hyperplane is all lost circulation data, and the other side is all non-lost circulation data.

[0030] Furthermore, the distance between any lost circulation data and all non-lost circulation data is calculated, the non-lost circulation data closest to the lost circulation data is selected to calculate the center point, and the classification hyperplane is adjusted according to the center point, and the center point is on the classification hyperplane.

[0031] Furthermore, the use of the test set data to test the accuracy of the model and to correct the model includes:

[0032] According to the parameter features after dimensionality reduction, the test set data are mapped into the multidimensional space one by one, and the accuracy of the model is calculated according to the distribution of the leakage labels and non-leakage data in the test set data on both sides of the classification hyperplane;

[0033] After the accuracy test, the test set data is added to the training set data and the model is iteratively trained.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The present invention provides a method for preventing leakage in geothermal drilling. The method collects a historical drilling data set, analyzes multiple parameter features in the historical drilling data set, and finally obtains several deep-level parameter features that cause leakage. The historical drilling data set is mapped to a high-dimensional space according to the deep-level parameter features. A classification hyperplane is generated in the high-dimensional space through a support vector machine model to separate leakage data from non-leakage data. When drilling is carried out on site, it is only necessary to map the deep-level parameter features of the site to the high-dimensional space in advance to determine in advance whether there is a leakage risk at the current drilling point. Then, by adjusting some parameter features, the current drilling point can be adjusted to a non-leakage state to avoid leakage, and valuable reference data can be accumulated for subsequent geological development work. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0037] Figure 1 The present invention is a flowchart of a method for preventing leakage in geothermal drilling provided in an embodiment of the present invention. Figure 2 This is a flow chart of another method for preventing lost circulation in geothermal drilling provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0039] Embodiment 1

[0040] This embodiment provides a method for preventing leakage in geothermal drilling, and this embodiment can be applied to geothermal resource drilling.

[0041] The process of geothermal drilling includes: geological exploration, well site selection, drilling equipment installation, drilling, geothermal well testing, and completion.

[0042] Geological exploration: Through geological surveys, geophysical explorations, geochemical explorations and other methods, detailed investigations and analyses are conducted on the strata, structures, igneous rocks, hydrogeological conditions, etc. of the target area to preliminarily determine the occurrence conditions and distribution range of geothermal resources.

[0043] Well site selection: Based on the geological exploration results, select the most suitable location for drilling to ensure that the geothermal well can accurately penetrate the geothermal reservoir.

[0044] Drilling equipment installation: Install drilling equipment, including drilling rigs, drill pipes, drill bits, etc., to prepare for subsequent drilling work.

[0045] Drilling: Drilling equipment is used to drill into the ground until it reaches the geothermal reservoir. During the drilling process, appropriate drilling methods, flushing media and drilling tools are required to overcome special difficulties in drilling.

[0046] Geothermal well testing: During or after the drilling process, geothermal wells are tested to determine the temperature, flow rate and other parameters of geothermal resources to provide a basis for subsequent exploitation and utilization.

[0047] Completion: After drilling, the geothermal well is completed, including installation of wellhead equipment, well sealing, etc., to ensure that the geothermal well can operate stably and for a long time.

[0048] It should be noted that in the prior art, a variety of major preventive measures are taken to prevent lost circulation, including geological surveys, reasonable wellbore structure design, drilling fluid adjustment, drilling operation specifications, parameter abnormality records, etc. However, these preventive measures can only reduce the risk of lost circulation to a certain extent. Not only can they not avoid lost circulation caused by the combined effect of multiple characteristic parameters, but they also require repeated drilling and drilling, adjustment of drilling fluid density and drilling speed, and even continuous replacement of casing, which seriously affects the progress and efficiency of drilling.

[0049] In order to solve the above problems, an embodiment of the present invention provides a method for preventing leakage in geothermal drilling. The main parameter features are obtained by performing principal component analysis on historical drilling data, and then a support vector machine model is constructed to segment the leakage data and non-leakage data in space, thereby preventing leakage caused by multiple parameter features and avoiding the occurrence of leakage.

[0050] like Figure 1 As shown, the method specifically comprises the following steps:

[0051] S110: Acquire a drilling history data set, where the drilling history data set includes lost circulation data and non-lost circulation data, where the lost circulation data and the non-lost circulation data have the same multiple original parameter features.

[0052] Among them, multiple original parameter features include but are not limited to the following aspects:

[0053] Drilling time: The time the drill bit stays at a certain depth during drilling. It reflects the nature of the rock formation. Harder or more brittle rock formations will extend the drilling time and may even cause the drill to get stuck during the drilling process.

[0054] Drilling depth: Records the maximum depth reached during the drilling process, which is an important basis for evaluating drilling capabilities and formation structure.

[0055] Drilling speed: reflects the drilling efficiency of the drill bit in a specific formation and is an important parameter for evaluating drilling technology and formation characteristics.

[0056] Drill bit wear: Recording the wear of the drill bit during the drilling process is of great significance for optimizing drill bit selection and improving drilling efficiency.

[0057] Lithological data: including rock type, mineral composition, structural structure, etc., is the basis for evaluating formation properties and oil and gas reservoir conditions.

[0058] Stratigraphic data: Recording the stratigraphic sequence, thickness, lithology changes, etc. revealed during the drilling process is crucial for building geological models and predicting the distribution of oil and gas reservoirs.

[0059] Porosity, permeability, and saturation: These parameters reflect the reservoir properties of rocks and the flow capacity of fluids, and are important bases for evaluating the reservoir characteristics of oil and gas reservoirs.

[0060] Downhole temperature and pressure: reflect the formation temperature and pressure conditions, which are of great significance for understanding the oil and gas generation, migration and accumulation processes.

[0061] Fluid resistivity and natural gamma rays: These parameters reflect the properties of formation fluids and the radioactive characteristics of rocks, and are important means of evaluating the properties of oil and gas reservoirs and reservoir characteristics.

[0062] Electrical logging, acoustic logging, and nuclear logging data provide a wealth of information on the physical properties of formations, such as resistivity, acoustic wave velocity, density, etc., which play a key role in formation division and reservoir evaluation.

[0063] Drilling fluid parameters: including the density, viscosity, water loss, etc. of the drilling fluid. These parameters reflect the performance of the drilling fluid and the downhole conditions, and are of great significance for maintaining the stability of the well wall and preventing formation pollution.

[0064] Drilling accident records: Recording accidents such as stuck drill, blowout, and lost circulation that occur during the drilling process and their treatment measures are of guiding significance for improving drilling safety and reducing accident risks.

[0065] Core and rock chip samples: Core and rock chip samples obtained through drilling are important physical data for geological research and can be used for analysis in many aspects such as petrology, mineralogy, and geochemistry.

[0066] In addition, the well leakage data in the drilling history data set refers to the above-mentioned multiple original parameter features collected when the well leakage occurs. This information can be collected through sensors, sonic logging and other devices or methods. This information includes digital data and text data. For example, drilling depth, drilling speed, etc. are reflected in the form of numerical values ​​and belong to digital data, while drill bit wear, lithology data, etc. are reflected in specific text descriptions and belong to text data.

[0067] Text data needs to be converted into digital data before principal component analysis can be performed. The conversion of text data includes: structured mapping, etc.

[0068] Non-lost-well data refers to various parameter data collected under normal drilling conditions.

[0069] S120: label the lost circulation data as lost circulation, label the non-lost circulation data as non-lost circulation, and randomly divide the drilling history data set into training set data and test set data.

[0070] The purpose of labeling the lost well data and the non-lost well data here is to distinguish the two types of data in space later. The division of the data set is to build the model later and train and test the accuracy of the model.

[0071] Usually the ratio of training set data to test set data is 4:1, that is, in the entire drilling history data set, 80% of the data belongs to the training set data, and the remaining 20% ​​of the data belongs to the test set data.

[0072] S130: performing principal component analysis on multiple original parameter features in the drilling history data set to obtain parameter features after dimension reduction.

[0073] Principal component analysis can reveal the internal connection between multivariate samples. Its principle is to retain the information contained in the original variables as much as possible while reducing the variable dimension. There are two types of principal component analysis: one is to discard some features, select K features from the original P features as the main features, and discard the rest of the features; the other is to combine several features into one feature, and finally transform P features into K features.

[0074] Principal component analysis, including:

[0075] Construct a multi-dimensional random vector (P-dimensional random vector):

[0076] X=(x1,x2,L,x p ) T , then the principal component is: Y i (i=1, 2,...,k; k≤p).

[0077] The principal component meets the following conditions:

[0078] Y i =a i T X,

[0079] Among them, a i is a p×1-dimensional numeric vector, a i Satisfy a i T a i =1, and its value is selected so that Y i The variance of each principal component Y1,Y2,...Y k They are independent of each other.

[0080] The original P-dimensional features can be reduced to K-dimensional features through principal component analysis, which is not only for screening the main parameter features in the drilling history data set, but also for facilitating the subsequent mapping of lost circulation data and non-lost circulation data in multidimensional space.

[0081] S140: Mapping the training set data into a multidimensional space according to the parameter features after dimensionality reduction.

[0082] The drilling history data set after principal component analysis contains only K parameter features, which are sorted in a certain order, and each lost circulation data and non-lost circulation data in the training set data are projected into multidimensional space.

[0083] S150: construct a support vector machine model, and separate the leakage data from the non-leakage data in the multidimensional space through the leakage labels and non-leakage labels in the training set data.

[0084] After the training set data is mapped to the multidimensional space, since the leaky data is labeled as leaky and the non-leaky data is labeled as non-leaky, the two types of data can be directly distinguished and their distribution in space can be seen. What the support vector machine model needs to do is to separate the two types of data through a function or a hyperplane expression.

[0085] S160: Use the test set data to test the accuracy of the support vector machine model and modify the support vector machine model.

[0086] After the support vector machine model finds such a function or hyperplane expression, the test set data is input into the support vector machine model, and the classification results of the test set data are statistically analyzed. The proportion of correct classification results is used as the accuracy of the model. After the accuracy test is completed, all test set data are added to the training set data to update the model.

[0087] This embodiment provides a method for preventing lost circulation in geothermal drilling. By analyzing the main parameter characteristics and establishing a support vector machine model, the model can be used to separate lost circulation data from non-lost circulation data, thereby preventing lost circulation accidents. At the same time, since there are a large number of adjustable features in these parameter characteristics, such as drilling speed, replacement of drill bits and drill pipes, drilling fluid parameters, etc., the drilling process can be prevented from entering a lost circulation state by adjusting them in advance.

[0088] Embodiment 2

[0089] Figure 2 This is a flow chart of a method for preventing lost circulation in geothermal drilling provided by the second embodiment of the present invention. This embodiment is further optimized based on the first embodiment. Figure 2 As shown, the method for preventing leakage in geothermal drilling provided in this embodiment specifically includes the following steps:

[0090] S210: Acquire a drilling history data set, where the drilling history data set includes lost circulation data and non-lost circulation data, where the lost circulation data and the non-lost circulation data have the same multiple original parameter features.

[0091] As described in Example 1, the original parameter characteristics include but are not limited to: drilling time, drilling depth, drilling speed, drill bit wear, lithology data, formation data, porosity, permeability, saturation, downhole temperature, pressure, fluid resistivity, natural gamma rays, electrical logging, acoustic logging, nuclear logging data, drilling fluid parameters, drilling accident records, cores, cuttings samples, etc.

[0092] In addition, the original parameter characteristics may also include: water level, water quality, water temperature, recharge volume, etc. of the recharge well.

[0093] S220: Label the lost circulation data as lost circulation, label the non-lost circulation data as non-lost circulation, and randomly divide the drilling history data set into training set data and test set data.

[0094] The purpose of labeling the lost well data and the non-lost well data is to distinguish the two types of data in space later. The division of the data set is to build the model later and train and test the accuracy of the model.

[0095] S230: After performing data standardization processing on the data in the drilling history data set, a multi-dimensional random vector matrix is ​​constructed.

[0096] Since principal component analysis is sensitive to data scale, the data needs to be standardized so that the mean of each feature is 0 and the standard deviation is 1. Standardization helps to eliminate the impact of different dimensions on the results.

[0097] The data standardization process includes:

[0098] The digital data is aligned using dimensionless criteria. Dimensionlessness, also known as data standardization, is a commonly used data preprocessing method. Its main purpose is to eliminate the dimensional effects between different features or indicators through data transformation, making the data comparable for subsequent analysis.

[0099] Perform structured mapping on text data. For example, loose rock, clastic rock, and carbonate rock are mapped to different numbers 2, 4, and 6, respectively, according to different rock types.

[0100] The aligned digital data and the structured mapped text data are normalized and mapped to (0, 1). For example, 2, 4, and 6 become 0, 0.5, and 1 after mapping.

[0101] The multidimensional random vector matrix is ​​represented as:

[0102]

[0103] Where X represents a multidimensional random vector matrix, x np represents the value corresponding to the nth sample and the pth parameter feature, X p Represents the vector corresponding to the pth parameter feature. Each element in the matrix is ​​between (0, 1).

[0104] S240: Calculate the covariance matrix of the multidimensional random vector matrix.

[0105] For the standardized data, calculate its covariance matrix. The covariance matrix describes the correlation between the features, where the diagonal elements are the variance of each feature and the off-diagonal elements are the covariance between the two features.

[0106] The covariance matrix is ​​expressed as:

[0107]

[0108] Where R represents the covariance matrix, i represents the row of the matrix, and j represents the column of the matrix. The calculation for any element in the covariance matrix is:

[0109]

[0110] Among them, r ij represents the element in the i-th row and j-th column of the covariance matrix, X ki represents the element in the i-th row and k-th column of the multidimensional random vector matrix, represents the average value of the i-th row in the multidimensional random vector matrix, Represents the mean of the j-th column in the multidimensional random vector matrix.

[0111] S250: Perform eigendecomposition on the covariance matrix to obtain a set of eigenvalues ​​and corresponding eigenvectors.

[0112] R=λ1a1+λ2a2+...+λ p a p

[0113] Each column in the covariance matrix constitutes an eigenvector, a1, a2, ..., a p are all eigenvectors, corresponding to a parameter feature. Eigenvalues ​​λ1, λ2, ..., λ p It represents the importance of each principal component (i.e., the size of the variance), while the eigenvector describes the distribution of the data in the direction of the principal component.

[0114] S260: According to the size of the eigenvalues, select the eigenvectors corresponding to the first k eigenvalues ​​from large to small as the principal components.

[0115] The selection criteria usually include: the eigenvalue is greater than 1, and the cumulative variance contribution rate reaches a certain value (for example, more than 80%).

[0116] After the principal components are selected, the principal component scores can also be calculated: the original data is projected onto the selected principal components to obtain the scores of each sample on each principal component. These scores can be used for subsequent analysis and processing, such as clustering, classification, etc. Finally, the selected principal components can also be analyzed and interpreted to understand the characteristics and information of the original data they represent.

[0117] S270: Mapping the training set data into a multidimensional space according to the parameter features after dimensionality reduction.

[0118] Each lost circulation data and non-lost circulation data in the training set data are projected into the multidimensional space.

[0119] S280: Construct a support vector machine model, and separate the leakage data from the non-leakage data in the multidimensional space through the leakage labels and non-leakage labels in the training set data.

[0120] According to the distribution of lost circulation data and non-lost circulation data in multidimensional space, a classification hyperplane of lost circulation data and non-lost circulation data is obtained, and a function expression of the classification hyperplane is obtained, one side of the classification hyperplane is all lost circulation data, and the other side is all non-lost circulation data.

[0121] The distance between any lost-well data and all non-lost-well data is calculated, and the non-lost-well data closest to the lost-well data is selected, and the classification hyperplane is adjusted according to the center point, wherein the center point is on the classification hyperplane.

[0122] Taking two-dimensional space as an example, there are leaking data (x1, y1) and non-leaking data (x2, y2). The distance between the leaking data and the non-leaking data is

[0123] For three-dimensional space, there are leaking data (x1, y1, z1) and non-leaking data (x2, y2, z1). The distance between the leaking data and the non-leaking data is The distance between points in high-dimensional space can be calculated by analogy.

[0124] The classification hyperplane can separate the lost-well data from the non-lost-well data.

[0125] S290: Use the test set data to test the accuracy of the model and modify the model.

[0126] Use the test set data to test the accuracy of the model and modify the model, including:

[0127] According to the parameter features after dimensionality reduction, the test set data are mapped into the multidimensional space one by one, and the accuracy of the model is calculated according to the distribution of the leakage labels and non-leakage data in the test set data on both sides of the classification hyperplane;

[0128] After the accuracy test, the test set data is added to the training set data and the model is iteratively trained.

[0129] After the model is modified, the current drilling data is collected in real time, and mapped to the multidimensional space according to the parameter characteristics after dimensionality reduction. By adjusting the data of the original parameter characteristics, the current drilling data is placed on the non-loosening data side in the support vector machine model.

[0130] Since there are some adjustable parameters in these parameter characteristics, such as drilling speed and other factors that can be controlled by humans, these parameters can be extracted and adjusted before the drill bit reaches a certain depth, so that the current drilling point is more likely to be in a normal drilling state and avoid entering a leak state.

[0131] If the current drilling data indicates that a lost circulation has occurred, the current drilling data is labeled as a lost circulation and added to the support vector machine model for model training;

[0132] At the same time, the corresponding well leakage points are found according to the current drilling data, and plugging materials are added to the well leakage points to plug the leakage.

[0133] For the first drilling, the density before drilling is 1.05g / cm 3 Prehydrated bentonite slurry 80m 3Use, stir evenly; for the second drilling, polyacrylamide potassium salt needs to be configured into a dilute solution, added to the bentonite slurry, and stirred evenly before the second drilling. To keep the polymer content above 0.5%, 50 to 80 kg of polymer needs to be added for every 120m or 24h of drilling. Strengthen solid control measures, use vibrating screens 100% during drilling, and the use rate of desanders and desilters reaches 85%. Wash and fill the pool in time to minimize the sand content and drill cuttings content of the mud. For the third and fourth drilling, polyacrylamide potassium salt needs to be configured into a dilute solution, added to the bentonite slurry, and stirred evenly before drilling. Add anti-collapse lubricants to ensure the stability of the wellbore and prevent collapse. Strictly controlling the appropriate bentonite content and strengthening solid control treatment are the key to controlling the performance of the well slurry. Use vibrating screens, desanders, and desilters during drilling, wash and fill the pool in time to minimize the sand content and drill cuttings content of the well slurry.

[0134] The present embodiment provides a method for preventing leakage in geothermal drilling. On the basis of the first embodiment, by constructing a covariance matrix to analyze the main parameter characteristics and establishing a support vector machine model, the model can be used to separate leakage data from non-leakage data, thereby preventing leakage accidents. At the same time, since there are a large number of adjustable features in these parameter characteristics, leakage can be avoided through pre-adjustment. If leakage occurs, the corresponding leakage point can be found according to the current drilling data to plug the leakage, and the model can be optimized to accumulate valuable reference data for subsequent geological development work.

[0135] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for preventing lost circulation in geothermal drilling, characterized in that: The method comprises: Acquire a drilling history data set, wherein the drilling history data set includes lost circulation data and non-lost circulation data, and the lost circulation data and the non-lost circulation data have the same multiple original parameter features; The lost circulation data are labeled as lost circulation, the non-lost circulation data are labeled as non-lost circulation, and the drilling history data set is randomly divided into training set data and test set data; Perform principal component analysis on multiple original parameter features in the drilling history data set to obtain parameter features after dimensionality reduction; According to the parameter characteristics after dimensionality reduction, the training set data is mapped into a multidimensional space; A support vector machine model is constructed to separate the leakage data from the non-leakage data in the multidimensional space through the leakage labels and non-leakage labels in the training set data; The test set data is used to test the accuracy of the support vector machine model and to modify the support vector machine model.

2. A method for preventing lost circulation in geothermal drilling according to claim 1, characterized in that: After the model is modified, the current drilling data is collected in real time, and mapped to the multidimensional space according to the parameter characteristics after dimensionality reduction. By adjusting the data of the original parameter characteristics, the current drilling data is placed on the non-loosening data side in the support vector machine model.

3. A method for preventing lost circulation in geothermal drilling according to claim 2, characterized in that: If the current drilling data indicates that a lost circulation has occurred, the current drilling data is labeled as a lost circulation and added to the support vector machine model for model training; At the same time, the corresponding well leakage points are found according to the current drilling data, and plugging materials are added to the well leakage points to plug the leakage.

4. A method for preventing lost circulation in geothermal drilling according to claim 1, characterized in that: The principal component analysis of multiple parameter features in the drilling history data set includes: S1: After data standardization processing of the data in the drilling history dataset, a multi-dimensional random vector matrix is ​​constructed; S2: Calculate the covariance matrix of the multidimensional random vector matrix; S3: Perform eigendecomposition on the covariance matrix to obtain a set of eigenvalues ​​and corresponding eigenvectors; S4: According to the size of the eigenvalue, select the eigenvectors corresponding to the first k eigenvalues ​​from large to small as the principal components.

5. A method for preventing lost circulation in geothermal drilling according to claim 4, characterized in that: The data standardization process for the data in the drilling history data set includes: Use dimensionless criteria to align numerical data; Perform structured mapping processing on text data; The aligned digital data and the structured mapped text data are normalized.

6. A method for preventing lost circulation in geothermal drilling according to claim 5, characterized in that: The multidimensional random vector matrix is ​​expressed as: Where X represents a multidimensional random vector matrix, x np represents the value corresponding to the nth sample and the pth parameter feature, X p Represents the vector corresponding to the p-th parameter feature.

7. A method for preventing lost circulation in geothermal drilling according to claim 1, characterized in that: The process of constructing the support vector machine model includes: According to the distribution of lost circulation data and non-lost circulation data in multidimensional space, a classification hyperplane of lost circulation data and non-lost circulation data is obtained, and a function expression of the classification hyperplane is obtained, one side of the classification hyperplane is all lost circulation data, and the other side is all non-lost circulation data.

8. A method for preventing lost circulation in geothermal drilling according to claim 7, characterized in that: The distance between any lost circulation data and all non-lost circulation data is calculated, and the non-lost circulation data closest to the lost circulation data is selected to calculate the center point. The classification hyperplane is adjusted according to the center point, and the center point is on the classification hyperplane.

9. A method for preventing lost circulation in geothermal drilling according to claim 8, characterized in that: The method of using the test set data to test the accuracy of the model and correcting the model includes: According to the parameter features after dimensionality reduction, the test set data are mapped into the multidimensional space one by one, and the accuracy of the model is calculated according to the distribution of the leakage labels and non-leakage data in the test set data on both sides of the classification hyperplane; After the accuracy test, the test set data is added to the training set data and the model is iteratively trained.

Images