Well drilling underground engineering parameter prediction method and device

Abstract

The invention discloses a drilling downhole engineering parameter prediction method and device, and the method comprises the steps: collecting multi-source data, and generating a multi-source data set; the multi-source data comprises any one or more of the following data: well logging data, well logging data and well drilling data; performing cleaning processing on the data in the multi-source data set; performing feature enhancement and data expansion enhancement on the cleaned data to generate a training sample set; training by using the training sample set to obtain an underground parameter prediction model; and predicting underground engineering parameters by using the underground parameter prediction model to obtain an underground engineering parameter prediction result. By means of the scheme, the accuracy of the underground engineering parameter prediction result can be improved, and the requirement for real-time monitoring of the ground-underground engineering parameter curve is met.

Description

Technical Field

[0001] The present invention relates to the technical field of exploration, and particularly to a method and device for predicting downhole engineering parameters of a drilling well. Background Art

[0002] With the continuous deepening of oil and gas exploration and development in China, deep-layer oil and gas will become a major replacement field for oil and gas exploration in China. The accurate prediction and real-time monitoring of downhole engineering parameters are particularly important for the exploration and development process of oil and gas resources. At present, the foreign measurement-while-drilling (MWD) technology has become mature, and relatively accurately measures drilling engineering parameters, including bottom hole pressure, torque, pressure inside the drill string, annulus pressure, vibration, rotational speed, bending moment, temperature, etc. However, the measurement device is expensive and difficult to promote and apply in large-scale downhole operations. Domestic related units have successively developed MWD systems, but mainly focus on formation characteristic parameters and lack measurement technology for drilling engineering parameters. At the same time, due to the influence of mud pulse transmission rate, the MWD has a large and discontinuous measurement point interval, and it is difficult to obtain data with a frequency matching that of surface engineering parameters, and cannot meet the urgent need for real-time monitoring of surface-downhole curve.

[0003] In recent years, the advantages of artificial intelligence technology in the oil and gas field have gradually emerged and have been applied to multiple scenarios such as working condition diagnosis and parameter optimization. For the problems of few samples of key downhole parameters and unbalanced data distribution obtained, how to use the existing measured key parameters combined with intelligent technology to achieve accurate prediction of downhole engineering parameters has become the key issue. Summary of the Invention

[0004] The present invention provides a method and device for predicting downhole engineering parameters of a drilling well to solve the problem that the existing MWD has a large and discontinuous measurement point interval, resulting in low accuracy of the existing intelligent prediction method, improve the accuracy of the prediction result of downhole engineering parameters, and meet the need for real-time monitoring of surface-downhole engineering parameter curves.

[0005] To this end, the present invention provides the following technical solutions:

[0006] A method for predicting downhole engineering parameters of a drilling well, the method comprising:

[0007] Collecting multi-source data to generate a multi-source data set; the multi-source data includes any one or more of the following data: logging data, logging while drilling (LWD) data, drilling data;

[0008] Performing cleaning processing on the data in the multi-source data set;

[0009] Performing feature enhancement and data expansion enhancement on the data after cleaning processing to generate a training sample set;

[0010] Training a downhole parameter prediction model by using the training sample set;

[0011] Predict the downhole engineering parameters using the downhole parameter prediction model to obtain the prediction results of the downhole engineering parameters.

[0012] Optionally, the mud logging data includes any one or more of the following data: well depth, weight on bit, rotary table torque, rotary speed, rate of penetration, pump pressure, total pit volume, inlet density, outlet density, hook load, pump strokes, equivalent density; the logging data includes any one or more of the following data: natural gamma, spontaneous potential, acoustic transit time, deep lateral resistivity curve, shallow lateral resistivity, density or neutron; the drilling data includes any one or more of the following data: natural gamma, spontaneous potential, well diameter, apparent resistivity, deep lateral resistivity curve, shallow lateral resistivity curve, acoustic wave, density, neutron, etc.

[0013] Optionally, the types of the multi-source data include any one or more of the following: numerical type, text type.

[0014] Optionally, the formats of the multi-source data include any one or more of the following: structured data, unstructured data.

[0015] Optionally, the cleaning of the data includes any one or more of the following: outlier rejection, linear regression interpolation, moving average filtering, etc., to screen out downhole anomalies and noise data and improve the density and quality of the data samples.

[0016] Optionally, the feature enhancement of the data after cleaning processing includes: performing feature enhancement on the data after cleaning processing through any one or more of the following methods: wavelet packet decomposition and reconstruction, genetic coding, expert experience.

[0017] Optionally, the data expansion and enhancement of the data after cleaning processing includes:

[0018] Performing data expansion and enhancement on the data after cleaning processing using SMOTE synthetic sampling; or

[0019] Using a generative adversarial network to learn the distribution characteristics of the original data set to generate more data samples.

[0020] Optionally, the method further includes:

[0021] Determining a correction coefficient using the prediction results of the downhole engineering parameters and the corresponding measured results of the downhole engineering parameters;

[0022] Correcting the prediction results using the correction coefficient.

[0023] A device for predicting downhole engineering parameters in drilling, the device includes:

[0024] A data acquisition module for acquiring multi-source data and generating a multi-source data set; the multi-source data includes any one or more of the following data: logging data, logging data, and drilling data;

[0025] A cleaning module for cleaning the data in the multi-source data set;

[0026] An enhancement module for performing feature enhancement and data expansion enhancement on the data after cleaning to generate a training sample set;

[0027] A model training module for training an underground parameter prediction model using the training sample set;

[0028] A prediction module for predicting underground engineering parameters using the underground parameter prediction model to obtain an underground engineering parameter prediction result.

[0029] Optionally, the device further includes:

[0030] A correction coefficient determination module for determining a correction coefficient using the underground engineering parameter prediction result and the corresponding measured result of the underground engineering parameter;

[0031] A correction module for correcting the prediction result using the correction coefficient.

[0032] The drilling underground engineering parameter prediction method and device provided by the present invention can effectively improve the quantity and distribution balance of training samples by performing feature enhancement and data expansion enhancement on the acquired multi-source data, extract effective feature parameters from the training sample set to train an underground parameter prediction model, greatly improve the accuracy of the model prediction result, and further better meet the requirements of real-time monitoring of the surface-underground engineering parameter curve.

[0033] Furthermore, by comparing the measured result of the underground engineering parameter with the intelligent prediction result of the underground engineering parameter, a correction coefficient is determined; the prediction result is corrected using the correction coefficient, thereby further improving the accuracy of the final prediction result. Description of the Drawings

[0034] Figure 1 is a flowchart of a drilling underground engineering parameter prediction method provided by the present invention;

[0035] Figure 2 is a schematic diagram of the result of decomposing the original monitoring curve based on wavelet packet in an embodiment of the present invention;

[0036] Figure 3 is a schematic diagram of the result of feature correlation analysis after feature enhancement based on genetic coding in an embodiment of the present invention;

[0037] Figure 4It is a flowchart of data expansion and enhancement by the generative adversarial network in an embodiment of the present invention;

[0038] Figure 5 It is a schematic diagram of data expansion and enhancement comparison based on the generative adversarial network in an embodiment of the present invention;

[0039] Figure 6 It is another flowchart of the drilling downhole engineering parameter prediction method provided by the present invention;

[0040] Figure 7 It is a schematic structural diagram of a drilling downhole engineering parameter prediction device provided by the present invention;

[0041] Figure 8 It is another schematic structural diagram of a drilling downhole engineering parameter prediction device provided by the present invention. Detailed implementation manners

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the accompanying drawings required for the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0043] The present invention will be described in detail below with reference to the accompanying drawings and specific implementation manners. The implementation manners cannot be enumerated one by one here, but the implementation manners of the present invention are not limited to the following implementation manners.

[0044] Aiming at the problems of large interval and discontinuity of existing MWD measurement points, resulting in few measurement samples and unbalanced data distribution, and thus low accuracy of the prediction results of existing intelligent prediction methods, the present invention provides a drilling downhole engineering parameter prediction method and device. By performing feature enhancement and data expansion and enhancement on the collected multi-source data, the number of samples and the balance of distribution can be effectively improved, making the downhole parameter prediction model obtained by training more accurate.

[0045] As Figure 1 shown, it is a flowchart of a drilling downhole engineering parameter prediction method provided by the present invention, including the following steps:

[0046] Step 101, collect multi-source data to generate a multi-source data set; the multi-source data includes any one or more of the following data: logging data, well logging data, drilling data.

[0047] The multi-source data can come from multiple wells and can include, but is not limited to, data such as drilling data, well logging data, logging data, etc. Among them:

[0048] The logging data mainly includes: well inclination data table, engineering parameter table, drilling fluid property table, bit record table, cuttings description record table, etc. The data therein includes, but is not limited to, any one or more of the following data: well depth, weight on bit, rotary table torque, rotary speed, penetration rate, pump pressure, total pit volume, inlet density, outlet density, hook load, pump strokes, equivalent density, etc.

[0049] The logging data includes, but is not limited to, any one or more of the following data: natural gamma, spontaneous potential, well diameter, apparent resistivity, deep lateral resistivity curve, shallow lateral resistivity curve, acoustic wave, density, neutron, etc.

[0050] The drilling data includes, but is not limited to, any one or more of the following data: well inclination data, drilling fluid property data, bit record data, cuttings description record data, etc. Among them:

[0051] The drilling fluid property data includes, but is not limited to, any one or more of the following data: funnel viscosity (unit: s), plastic viscosity (unit: mPa·s), yield point (unit: Pa), drilling fluid system, etc.;

[0052] The bit record data includes, but is not limited to, any one or more of the following data: bit size, bit model, new degree upon entering the well (unit: %), new degree upon leaving the well (unit: %), weight on bit (unit: kN), rotary speed (unit: r / min), bit nozzle, drill string assembly information, etc.;

[0053] The cuttings description record data includes, but is not limited to, any one or more of the following data: formation, well section, lithology and hydrocarbon-water-bearing description, etc. In the lithology and hydrocarbon-water-bearing description information, it includes numerical and text information such as cuttings shape, maximum particle size, minimum particle size, average particle size, sorting, roundness, etc.

[0054] Among the above multi-source data, there is structured data and there can also be unstructured data; moreover, it can be numerical or text, and the embodiments of the present invention do not make any limitations in this regard.

[0055] Therefore, it is also necessary to preprocess these data, which can specifically include, but is not limited to, processing such as unifying the sampling frequency and format conversion.

[0056] For example, convert the sampling frequencies of well inclination data, drilling fluid property data, bit record data, and cuttings description record data to sample every 1 m. For example, for well inclination data, the wellbore trajectory data can be interpolated using the cubic spline curve interpolation method to convert it to sample 1 point every 1 m.

[0057] After unifying the sampling frequencies of various types of data, due to the inconsistency of each data type, it is necessary to establish corresponding data extraction and processing methods for specific data types.

[0058] For example, well deviation data and engineering parameters are all numerical data. After unifying the sampling frequency to 1m, they can be directly read using the Pandas module in Python, which is convenient for subsequent data splicing operations.

[0059] For another example, the drilling fluid system data is text-based. The drilling fluid system can be roughly classified into three major systems: water-based, oil-based, and gas-based. Among them, the water-based includes non-dispersed, dispersed, polymer, low-solids, calcium-treated, saturated brine drilling fluids, etc.; the gas-based includes air and foam drilling fluids. For different categories, they can be mapped to corresponding numerical values, that is, text-based data is mapped to numerical data.

[0060] For another example, the bit model data is stored in the drill string assembly information in text form, and information is extracted according to the actual bit type, such as PDC bits.

[0061] After preprocessing the above multi-source data, a multi-source data set can be obtained. For example, data such as well deviation, engineering parameters, drilling fluid performance, bit records, and cuttings description records can be spliced according to the well depth as the index to form a comprehensive mud logging parameter data set.

[0062] Step 102: Clean the data in the multi-source data set.

[0063] The methods for cleaning the data may include, but are not limited to, any one or more of the following:

[0064] Outlier rejection, linear regression interpolation, moving average filtering, etc.

[0065] Through data cleaning, downhole abnormal data and noise data can be screened out, the influence of downhole noise on the collected data can be weakened, and the density and quality of the data samples can be improved.

[0066] Step 103: Perform feature enhancement and data expansion enhancement on the data after cleaning to generate training samples.

[0067] The following any one or more methods can be used for feature enhancement of the data after cleaning: wavelet packet decomposition and reconstruction, genetic coding, expert experience.

[0068] For example, refer to Figure 2 , Figure 2 which is a schematic diagram of the original monitoring curve based on wavelet packet decomposition in the embodiment of the present invention.

[0069] Figure 2Among them, the wavelet decomposition method is used to decompose the data obtained from the original monitoring curve 20 into a group of sub-bands IMF1, IMF2, and IMF3 with different frequency bands, so as to better capture the local characteristics of the ground monitoring data, and then the data curve is reconstructed according to the selected wavelet packet. By analyzing the distribution law of the extracted features, new characterization features such as the change rate of parameters with different physical meanings are constructed, and based on the ensemble learning model in the machine learning algorithm, such as the random forest and XGBOOST models, the importance of the newly constructed features is calculated.

[0070] Reference Figure 3 , Figure 3 It is a schematic diagram of the feature correlation analysis result after feature enhancement based on genetic coding in the embodiment of the present invention.

[0071] According to the input ground monitoring parameters, genetic coding encodes each parameter into a gene individual, performs replication, crossover, and mutation operations on it, combines and transforms the existing features, and finally generates 6 new high-dimensional features that are more relevant to the target engineering parameters. At the same time, 6 expressions of the new features are obtained as shown in Table 1 below.

[0072] Table 1

[0073] Feature Serial Number New Feature Expression New Feature 1 <![CDATA[Export density 2 / Export flow rate]]> New Feature 2 Outlet Density / (0.287 * Torque) New Feature 3 Outlet Flow + Total Tank Volume New Feature 4 (Outlet Flow - Inlet Flow) / Standpipe Pressure 2 * 0.514 New Feature 5 Outlet Flow * Total Hydrocarbons / Drilling Time New Feature 6 Outlet Density * Total Hydrocarbons / Torque

[0074] In addition, feature construction can also be based on user-defined, for example, functions such as addition, subtraction, multiplication, division, trigonometric functions, and exponents can be used to combine the original features to generate more physically meaningful feature parameters, such as the specific weight on bit can be constructed by dividing the weight on bit by the bit diameter.

[0075] Of course, other feature enhancement methods can also be adopted, and the embodiments of the present invention do not limit this.

[0076] In the embodiment of the present invention, after feature enhancement of the data after cleaning processing, data expansion enhancement is performed on the data. For the data expansion enhancement, multiple methods can also be adopted, such as: using SMOTE (Synthetic Minority Oversampling Technique) comprehensive sampling to perform sampling interpolation synthesis on the data; or using a generative adversarial network to learn the distribution characteristics of the original data set to generate more data samples, etc.

[0077] SMOTE is a comprehensive sampling artificial synthetic data algorithm used to solve the problem of data class imbalance, and synthesizes data in a way that combines over-sampling the minority class and under-sampling the majority class.

[0078] Through SMOTE comprehensive sampling, new synthetic samples can be generated to fill the shortage of samples, thus solving the problem of few samples for measuring key underground parameters. By this method, new underground engineering parameter data samples are synthesized, maintaining the distribution characteristics of the original features, expanding the sample space, and enhancing the generalization ability of the model.

[0079] GAN (Generative Adversarial Network) is a generative model that learns through the mutual game of two neural networks. The generative adversarial network can learn to perform generation tasks without using labeled data. The generative adversarial network consists of a generator and a discriminator. The generator randomly samples from the latent space as input, and its output results need to mimic the real samples in the training set as much as possible. The input of the discriminator is either the real sample or the output of the generator, and its purpose is to distinguish the output of the generator from the real samples as much as possible. The generator and the discriminator confront and learn from each other, and the ultimate goal is to make the discriminator unable to judge whether the output result of the generator is real.

[0080] Therefore, in a non-limiting embodiment of the present invention, GAN is used to expand and enhance the data.

[0081] Refer to Figure 4 , Figure 4 is a flowchart of using the generative adversarial network to expand and enhance data in an embodiment of the present invention.

[0082] In step 501, obtain the real data set collected underground.

[0083] In step 502, by inputting a random noise vector, train the generative network G to generate a series of data with the same shape as the real data.

[0084] In step 503, input the generated data and the real data into the discriminant network D to distinguish the distribution differences of the data. An output of 1 represents that the two types of data are similar and is judged as real data, and an output of 0 represents that the two types of data have different distributions and is judged as false data.

[0085] In step 504, by judging the true and false data, the generative network G and the discriminant network D form a dynamic confrontation. The generative network G gradually optimizes the quality of the generated samples, thereby generating data closer to the underground engineering parameters, making it difficult for the discriminant network D to distinguish between true and false.

[0086] Through the training of GAN, the diversity and quantity of samples can be increased, further improving the robustness and accuracy of the prediction model. The training method of GAN can adopt the existing technology, and the embodiments of the present invention do not make limitations in this regard.

[0087] Figure 5The figure shows a schematic diagram of data expansion and enhancement comparison based on a generative adversarial network in an embodiment of the present invention. Figure 5 The upper curve corresponds to the time series of the total pit volume data of multiple wells, and the lower curve corresponds to the time series after data expansion and enhancement of the total pit volume data of multiple wells.

[0088] Step 104: Train an underground parameter prediction model using the training sample set.

[0089] In an embodiment of the present invention, the underground parameter prediction model may be, but is not limited to, any one of the following: a machine learning model, a deep learning model, a variety of deep learning integrated models, etc.

[0090] The underground parameter prediction model reflects the mapping relationship between surface engineering parameters and underground measured data. Its input is multi-source data after feature enhancement and data expansion, and its output is one or more parameters such as bit weight on bit, torque, total pit volume, standpipe pressure, and bottom hole pressure.

[0091] Furthermore, multiple different candidate underground parameter prediction models can be respectively established based on the above training sample set, and then these candidate models are tested using the corresponding test set. The optimal candidate model is selected as the underground parameter prediction model according to the test results.

[0092] In specific applications, the root mean square error RMSE, average relative error δ, maximum relative error δmax, training time, etc. can be selected as evaluation indicators for the model prediction effect, so as to determine the underground parameter prediction model according to the corresponding evaluation indicators.

[0093] The calculation formulas for the root mean square error RMSE, average relative error δ, and maximum relative error δmax are as follows:

[0094]

[0095]

[0096] δ max =max{δ1,δ2,δ3,……δ N}

[0097] where, y itrue is the target true value of the i-th data; y ipre is the target predicted value of the i-th data; N is the total number of samples; δ i is the relative error of the i-th value.

[0098] Further, when constructing the downhole parameter prediction model, physical knowledge constraints such as hydraulics and string mechanics can also be embedded to ensure that the prediction results conform to physical laws, avoid prediction results violating actual engineering conditions, and approach real downhole engineering parameters.

[0099] The above physical constraints include physical laws and engineering constraint conditions of downhole engineering parameters, such as direct and inverse proportional response relationships, multiphase flow equations, string force balance, etc., so as to obtain inequality constraints and their neural network expressions:

[0100]

[0101] Among them, w [n] represents the model weight matrix of the nth layer; b represents the model bias value; X represents the function variable for constraint, such as well depth, drilling fluid density, bottom hole torque, etc.; T represents the downhole parameter to be predicted, such as bottom hole drill pressure, torque, etc.; x is the network input parameter.

[0102] The inequality relationship can be embedded into the final objective function through a penalty function for network training. Since the objective function of the constrained neural network is relatively complex, the penalty function method can be used to construct the objective function and constraint conditions into an auxiliary function, thereby converting the constrained nonlinear programming problem into an unconstrained nonlinear programming problem. At the same time, in order to balance the prediction accuracy and the fitting of physical constraints, the weight of physical constraints in the loss function needs to be reasonably set. Therefore, different penalty factors λ can be set and solved to make the model have different responsiveness when facing different degrees of constraints.

[0103]

[0104] Among them, F is the objective function to be optimized, T pre is the model prediction value, T true is the true value of the parameter, and variables such as x1, x2 represent network input parameters, such as well depth, drilling fluid density, bottom hole torque, etc.

[0105] Step 105, use the downhole parameter prediction model to predict downhole engineering parameters to obtain downhole engineering parameter prediction results.

[0106] The drilling downhole engineering parameter prediction method provided by the present invention can effectively improve the quantity and distribution balance of training samples through feature enhancement and data expansion enhancement of the collected multi-source data, extract effective feature parameters from the training sample set to train the downhole parameter prediction model, greatly improve the accuracy of the model prediction results, and further better meet the requirements of real-time monitoring of ground-downhole engineering parameter curves.

[0107] Such as Figure 6This is another flowchart of the method for predicting downhole engineering parameters provided by the present invention, including the following steps:

[0108] Step 701, collect multi-source data and generate a multi-source data set; the multi-source data includes any one or more of the following data: logging data, well logging data, drilling data;

[0109] Step 702, perform cleaning processing on the data in the multi-source data set;

[0110] Step 703, perform feature enhancement and data expansion enhancement on the cleaned data to generate a training sample set;

[0111] Step 704, train a downhole parameter prediction model using the training sample set;

[0112] Step 705, use the downhole parameter prediction model to predict downhole engineering parameters and obtain a downhole engineering parameter prediction result.

[0113] The above steps 701 to 705 are the same as the corresponding steps in Figure 1 the embodiment shown, and will not be described in detail here.

[0114] Step 706, use the downhole engineering parameter prediction result and the corresponding measured result of the downhole engineering parameter to determine a correction coefficient.

[0115] Specifically, compare the predicted downhole engineering parameter with the corresponding measured downhole parameter, analyze the prediction error, and determine the correction coefficient according to the error magnitude.

[0116] Step 707, correct the prediction result using the correction coefficient.

[0117] In the embodiment of the present invention, the Adam gradient descent optimization algorithm is used to perform iterative correction of the model parameters, and the adaptive learning rate of each parameter is calculated based on the gradient and historical information of the model parameters, so as to better adjust the model parameters and make the error between the predicted value and the true value as small as possible.

[0118] First, calculate the gradient g t at the t-th time step, which contains the partial derivatives corresponding to each parameter:

[0119] g t =▽ θ F(θ t-1 )

[0120] where θ t is the model parameter at the t-th iteration; ▽ θ represents the first-order partial derivative of the objective function with respect to the parameter θ, and F(θ t-1 ) represents with respect to θt-1 Objective function

[0121] Secondly, calculate the exponentially weighted average m of the gradients t and the exponentially weighted average v of the squared gradients t :

[0122] m t = β1m t-1 + (1 - β1)g t

[0123] v t = β2v t-1 + (1 - β2)g t 2

[0124] where β1 is the exponential decay rate that controls the weight distribution, usually taking a value close to 1, with a default value of 0.9; β2 is the exponential decay rate that controls the influence of the previous squared gradients

[0125] Since m0 is initialized to 0, m t will tend to be 0, so it is necessary to correct the bias of the gradient mean m t ; similar to m0, because v0 is initialized to 0, v t tends to be 0 in the initial stage of training, and correct it:

[0126]

[0127]

[0128]

[0129] where η is the learning rate, initially set to 0.001 for example, and ε is the smoothing term, which can be set to 10 -8 for example, to avoid division by zero

[0130] Through multiple iterations and corrections, the prediction results are continuously optimized. When the set error threshold is met, the optimization stops, and a downhole parameter prediction model that meets certain accuracy requirements is obtained

[0131] Using this downhole parameter prediction model, the downhole engineering parameters during drilling can be monitored in real time to guide the drilling operation

[0132] The downhole engineering parameter prediction method provided by the present invention determines a correction coefficient by comparing the measured results of the downhole engineering parameters with the intelligent prediction results of the downhole engineering parameters; and corrects the prediction results using the correction coefficient, thereby further improving the accuracy of the final prediction results

[0133] Correspondingly, the present invention also provides a downhole engineering parameter prediction device, such asFigure 7 As shown, it is a schematic structural diagram of the device.

[0134] In this embodiment, the downhole engineering parameter prediction device 800 includes the following modules:

[0135] A data acquisition module 801, configured to acquire multi-source data and generate a multi-source data set; the multi-source data includes any one or more of the following data: logging data, well logging data, and drilling data;

[0136] A cleaning module 802, configured to perform cleaning processing on the data in the multi-source data set;

[0137] An enhancement module 803, configured to perform feature enhancement and data expansion enhancement on the data after cleaning processing to generate a training sample set;

[0138] A model training module 804, configured to train a downhole parameter prediction model using the training sample set;

[0139] A prediction module 805, configured to predict downhole engineering parameters using the downhole parameter prediction model to obtain a downhole engineering parameter prediction result.

[0140] The above cleaning module 802 can use various methods for data cleaning, such as outlier removal, linear regression interpolation, moving average filtering, etc., to screen out downhole anomalies and noise data and improve the density and quality of data samples.

[0141] The above enhancement module 803 can use various methods for feature enhancement and data enhancement. For example, the following any one or more methods can be used to perform feature enhancement on the data after cleaning processing: wavelet packet decomposition and reconstruction, genetic coding, expert experience, etc.; for data enhancement, the SMOTE synthetic sampling method can be used, or the generative adversarial network can be used to learn the distribution characteristics of the original data set to generate more data samples. Of course, other data enhancement methods can also be used, and the embodiments of the present invention do not make limitations in this regard.

[0142] As Figure 8 shown, it is another schematic structural diagram of the downhole engineering parameter prediction device provided by the present invention.

[0143] Different from the Figure 7 embodiment shown, in this embodiment, the downhole engineering parameter prediction device 800 further includes the following modules:

[0144] A correction coefficient determination module 806, configured to determine a correction coefficient using the downhole engineering parameter prediction result and the corresponding measured result of the downhole engineering parameter;

[0145] A correction module 807 for correcting the prediction result by using the correction coefficient.

[0146] For more descriptions of each module in the above-mentioned downhole engineering parameter prediction device of the present invention, reference can be made to the descriptions in the method embodiment of the present invention before, which will not be repeated here.

[0147] The downhole engineering parameter prediction device provided by the present invention can effectively improve the quantity and distribution balance of training samples by performing feature enhancement and data expansion enhancement on the collected multi-source data, extract effective feature parameters from the training sample set to train the downhole parameter prediction model, greatly improve the accuracy of the model prediction result, and further better meet the requirements of real-time monitoring of the ground-downhole engineering parameter curve. Further, by comparing the measured result of the downhole engineering parameter with the intelligent prediction result of the downhole engineering parameter, the correction coefficient is determined; the prediction result is corrected by using the correction coefficient, thereby further improving the accuracy of the final prediction result.

[0148] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0149] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between each embodiment, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. Moreover, the system embodiments described above are only illustrative. The modules and units described as separate components may or may not be physically separated, that is, they may be located on one network unit, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative work.

[0150] The above has introduced the embodiments of the present invention in detail. Specific implementation manners are used herein to elaborate on the present invention. The description of the above embodiments is only used to help understand the method and system of the present invention. They are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. The content of this specification should not be construed as a limitation to the present invention. Therefore, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting downhole engineering parameters during drilling, characterized in that, The method includes: Collecting multi-source data to generate a multi-source data set; the multi-source data includes any one or more of the following data: logging data, logging data, drilling data; Cleaning the data in the multi-source data set; Performing feature enhancement and data expansion enhancement on the cleaned data to generate a training sample set; Training an underground parameter prediction model using the training sample set; Predicting underground engineering parameters using the underground parameter prediction model to obtain an underground engineering parameter prediction result.

2. The method for predicting underground engineering parameters in drilling according to claim 1, wherein The logging data includes any one or more of the following data: well depth, hook load, rotary table torque, rotation speed, rate of penetration, pump pressure, total pit volume, inlet density, outlet density, hook load, pump strokes, equivalent density; The logging data includes any one or more of the following data: natural gamma, spontaneous potential, acoustic transit time, deep lateral resistivity curve, shallow lateral resistivity, density or neutron; The drilling data includes any one or more of the following data: natural gamma, spontaneous potential, hole diameter, apparent resistivity, deep lateral resistivity curve, shallow lateral resistivity curve, acoustic wave, density, neutron, etc.

3. The prediction method of downhole engineering parameters for drilling according to claim 1, characterized in that The types of the multi-source data include any one or more of the following: numerical type, text type.

4. The prediction method of downhole engineering parameters for drilling according to claim 1, characterized in that The formats of the multi-source data include any one or more of the following: structured data, unstructured data.

5. The method for predicting downhole engineering parameters of a well drilling according to claim 1, characterized in that, The cleaning of the data includes any one or more of the following: outlier removal, linear regression interpolation, moving average filtering, etc., to screen out underground anomalies and noise data and improve the density and quality of data samples.

6. The method for predicting downhole engineering parameters of a well drilling according to claim 1, characterized in that, The feature enhancement of the cleaned data includes: Performing feature enhancement on the cleaned data by any one or more of the following methods: wavelet packet decomposition and reconstruction, genetic coding, expert experience.

7. The prediction method of downhole engineering parameters for drilling according to claim 1, wherein The data expansion enhancement of the cleaned data includes: Performing data expansion enhancement on the cleaned data using SMOTE synthetic sampling; or Using a generative adversarial network to learn the distribution characteristics of the original data set to generate more data samples.

8. The method for predicting downhole engineering parameters of a well drilling according to any one of claims 1 to 7, characterized in that, The method further includes: Determining a correction coefficient using the underground engineering parameter prediction result and the corresponding measured result of the underground engineering parameter; Correcting the prediction result using the correction coefficient.

9. A device for predicting downhole engineering parameters of a drilling well, characterized in that, The device includes: A data acquisition module for collecting multi-source data to generate a multi-source data set; the multi-source data includes any one or more of the following data: logging data, logging data, drilling data; A cleaning module for cleaning the data in the multi-source data set; An enhancement module for performing feature enhancement and data expansion enhancement on the cleaned data to generate a training sample set; A model training module for training an underground parameter prediction model using the training sample set; A prediction module for predicting underground engineering parameters using the underground parameter prediction model to obtain an underground engineering parameter prediction result.

10. The drilling downhole engineering parameter prediction device according to claim 9, wherein The device further includes: A correction coefficient determination module for determining a correction coefficient using the underground engineering parameter prediction result and the corresponding measured result of the underground engineering parameter; A correction module for correcting the prediction result by using the correction coefficient.

Images