Methods, systems, media, and apparatuses for predicting formation pore pressure of a target well based on well log data of drilled wells and real-time while-drilling data of the target well
By using logging data from drilled wells and real-time drilling data from target wells, and employing sequence neural networks and related alignment algorithms for data processing, the accuracy and applicability issues of formation pore pressure prediction were resolved. Real-time optimization and cross-block model updates were achieved, thereby improving drilling safety and intelligence.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA NATIONAL OFFSHORE OIL (CHINA) CO LTD
- Filing Date
- 2026-04-14
- Publication Date
- 2026-06-09
AI Technical Summary
Existing technologies lack effective methods for predicting formation pore pressure in target wells based on logging data from drilled wells and real-time drilling data from the target well. This results in poor model applicability in new areas, inability to dynamically update and optimize in real time, and failure to effectively utilize real-time drilling data for correction and optimization.
By acquiring the logging curve data of the target well and the relationship between formation pore pressure labels, a sequence neural network model is used for preprocessing and outlier removal. A data distribution alignment is established in conjunction with relevant alignment algorithms, the formation pore pressure prediction model is updated in real time, and real-time optimization is performed using drilling data.
It improves the accuracy and generalization ability of formation pore pressure prediction, enables the migration and dynamic updating of cross-block models, and enhances the safety and intelligence of drilling operations.
Smart Images
Figure CN122169809A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of oil and gas drilling engineering, and specifically relates to a method, system, medium and equipment for predicting formation pore pressure of a target well based on logging data of an existing well and real-time drilling data of the target well. Background Technology
[0002] Accurate prediction of formation pore pressure is crucial for ensuring drilling safety and efficiency. The core challenges of existing technologies lie in the insufficient generalization ability and lack of dynamic adaptability of the models. Traditional methods, such as the Eaton method, heavily rely on regional experience and accurate positive compaction trend lines, resulting in poor applicability in new exploration areas or complex structural zones. While machine learning-based intelligent methods can handle nonlinear relationships, their performance is highly dependent on large amounts of labeled data, leading to incompatibility in new areas or data-scarce regions. Furthermore, data heterogeneity across different work zones due to variations in logging series and standards makes it difficult to directly transfer and reuse trained models.
[0003] A more prominent problem is that existing predictive models are mostly static "pre-drilling models," which are fixed once established and cannot be dynamically corrected and continuously optimized using real-time logging-while-drilling data during drilling. This results in their inability to respond to real-time changes in the actual formation and limited early warning capabilities. At the same time, the enormous value of real-time logging-while-drilling data, especially the occasional measured pressure points, for correcting and optimizing models has not been effectively utilized.
[0004] In summary, there is currently a lack of methods to predict formation pore pressure in target wells based on logging data from drilled wells and real-time drilling data from the target well. Summary of the Invention
[0005] To address the aforementioned problems, the present invention aims to provide a method, system, medium, and device for predicting formation pore pressure of a target well based on well logging data from drilled wells and real-time drilling data from the target well. This addresses the current difficulty in finding a method for predicting formation pore pressure of a target well based on well logging data from drilled wells and real-time drilling data from the target well.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] In a first aspect, the present invention discloses a method for predicting formation pore pressure in a target well based on logging data from an existing well and real-time drilling data from the target well, comprising: Step A: Obtain the relationship between the logging curve data of the target well and the formation pore pressure label of the target well; Step B: Obtain real-time logging data of the target well during drilling operations, preprocess the target well logging data to remove outliers, and obtain the target well logging dataset after removing outliers. Step C: Substitute the measured dataset of the target well after removing outliers into the initial static prediction model of formation pore pressure. The model obtains the predicted data value of formation pore pressure of the target well based on the relationship between the data of the logging curve of the target well and the formation pore pressure label of the target well.
[0008] Step A includes the following steps: Step A1: Obtain the logging curve data and formation pore pressure labels of the drilled wells. Train the logging curve data and formation pore pressure labels of the drilled wells using the initial static prediction model of formation pore pressure to establish the relationship between the logging curve data and formation pore pressure labels of the drilled wells. The initial static prediction model for formation pore pressure is a formation pore pressure initial static prediction model based on a sequence neural network.
[0009] Step A2: Preprocess the logging data of the drilled wells to remove outliers and obtain the drilled well dataset after outlier removal. ; Step A3: Obtain background data of the target well's target region The target domain background data of the target well is obtained through a relevant alignment algorithm. and the drilled well dataset after removing outliers Establish correlations to obtain the source domain data distribution after correlation alignment transformation; Step A4: Using the source domain data distribution after the relevant alignment transformation as input data, substitute it into the initial static prediction model of formation pore pressure for calculation. The model obtains the relationship between the logging curve data of the target well and the formation pore pressure label of the target well based on the relationship between the logging curve data of the drilled well and the formation pore pressure label of the drilled well. Specifically, the target domain background data of the target well is obtained through a relevant alignment algorithm. and the drilled well dataset after removing outliers Establishing correlations and obtaining the source domain data distribution after correlation alignment transformation includes the following steps: 1) Based on the target well's target domain background data Obtain target domain background data of the target well covariance matrix ; 2) Based on the drilled well dataset after removing outliers Obtain the drilled well dataset after removing outliers covariance matrix ; 3) Based on the relevant alignment algorithm, the target domain background data of the target well is... covariance matrix and the drilled well dataset after outlier removal covariance matrix Establish the correlation to obtain the source domain data distribution after correlation alignment transformation.
[0010] Specifically, the expression for the relevant alignment algorithm is:
[0011] In the formula, This represents the drilled well dataset after outlier removal. The covariance matrix; This represents the background data of the target domain of the target well. The covariance matrix; A represents a linear transformation matrix; AT represents the transpose of the linear transformation matrix A; The covariance matrix represents the features of the source domain data after transformation by the linear transformation matrix A; Denotes the squared Frobenius norm of a matrix; This indicates the search for the optimal linear transformation matrix A.
[0012] The above method also includes the following steps: Step D: Verification of the predicted formation pore pressure data for the target well, including the following steps: Step D1: Obtain the measured value of the formation pore pressure label of the target well; Step D2: Output the predicted value of the formation pore pressure label of the target well through the initial static prediction model of formation pore pressure based on the sequence neural network, and compare the measured value of the formation pore pressure label of the target well with the predicted value of the formation pore pressure label of the target well. Step D3: If the measured value of the formation pore pressure label of the target well is within the error range of the predicted value of the formation pore pressure label of the target well, then the predicted value of the formation pore pressure of the target well is valid; otherwise, repeat step B, re-acquire the measured data set of the target well after removing outliers, and repeat steps C to D3 until the predicted value of the formation pore pressure of the target well is valid.
[0013] Secondly, the present invention also discloses an apparatus for predicting formation pore pressure of a target well based on logging data from a drilled well and real-time drilling data from the target well, comprising: The first processing unit is used to acquire the data of the logging curve of the target well and the relationship between the formation pore pressure label of the target well; The second processing unit is used to obtain the target well's real-time logging data during drilling construction, preprocess the target well's real-time logging data, remove outliers, and obtain the target well's real-time logging dataset after removing outliers. The third processing unit is used to input the measured dataset of the target well after removing outliers into the initial static prediction model of formation pore pressure. The model obtains the predicted data value of formation pore pressure of the target well based on the relationship between the data of the logging curve of the target well and the formation pore pressure label of the target well.
[0014] Thirdly, the present invention also discloses a computer-readable storage medium, characterized in that it stores a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.
[0015] Fourthly, the present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described above.
[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: (I) This invention discloses a method for predicting formation pore pressure of a target well based on logging data of a drilled well and real-time drilling data of the target well. The method includes: obtaining the relationship between the logging curve data of the target well and the formation pore pressure label of the target well, including the following steps: obtaining the real-time drilling logging data of the target well through drilling operations; preprocessing the real-time drilling logging data of the target well to remove outliers and obtaining the target well measured dataset after removing outliers; substituting the target well measured dataset after removing outliers into an initial static prediction model of formation pore pressure, which obtains the predicted formation pore pressure data value of the target well based on the relationship between the logging curve data of the target well and the formation pore pressure label of the target well. This invention predicts formation pore pressure in target wells based on logging data from drilled wells and real-time drilling data from target wells. It not only improves the accuracy of the predicted formation pore pressure data for target wells, but also features high accuracy, strong generalization ability, and dynamic updates. It aims to solve the problem of cross-block model migration and realize real-time optimization of the prediction model using drilling data, ultimately improving the safety and intelligence of drilling operations. Attached Figure Description
[0017] Figure 1 This is a flowchart of a method for predicting formation pore pressure of a target well based on logging data of an existing well and real-time drilling data of the target well, as provided in Embodiment 1 of the present invention. Detailed Implementation
[0018] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.
[0019] To address the current lack of methods for predicting formation pore pressure in a target well based on logging data from drilled wells and real-time drilling data from the target well, this invention discloses a formation pore pressure prediction method. First, it obtains the relationship between the logging curve data of the target well and the formation pore pressure label of the target well. Second, it substitutes the measured dataset of the target well (after removing outliers) into the initial static prediction model of formation pore pressure based on a sequence neural network. This model obtains the predicted formation pore pressure data value of the target well based on the relationship between the logging curve data of the target well and the formation pore pressure label of the target well, thereby improving the accuracy of the predicted formation pore pressure data value of the target well.
[0020] Example 1: A method for predicting formation pore pressure in a target well based on logging data from existing wells and real-time drilling data from the target well. Embodiment 1 of this invention provides a method for predicting formation pore pressure in a target well based on logging data from an existing well and real-time drilling data from the target well. (Refer to...) Figure 1 The method includes the following steps: Step A: Obtain the relationship between the logging curve data of the target well and the formation pore pressure label of the target well, including the following steps: Step A1: Obtain the logging curve data and formation pore pressure labels of the drilled wells. Train the logging curve data and formation pore pressure labels of the drilled wells using the initial static prediction model of formation pore pressure to establish the relationship between the logging curve data and formation pore pressure labels of the drilled wells. The specific method for obtaining the logging curve data of drilled wells is to collect historical data of drilled wells from multiple different geological blocks, including various sequences such as drilling depth, natural gamma (GR), sonic transit time (DT), resistivity (RT), density (RHOB), and neutron porosity (CNL). These historical data of drilled wells are used as source domain data.
[0021] The specific method for obtaining formation pore pressure labels for drilled wells is to obtain the true formation pore pressure labels corresponding to each depth point through post-drilling formation testing, well completion testing, or calculation using known empirical formulas. Simultaneously, forward modeling data based on multi-condition geological models are added to the source domain data to expand the training sample space.
[0022] The initial static prediction model for formation pore pressure is based on a sequence neural network. It is worth noting that this sequence neural network-based model is existing technology. The sequence neural network is a Long Short-Term Memory (LSTM) network, which includes an input gate, a forget gate, and an output gate to store and transmit the sequential dependencies of well logging data with depth. The model is trained using optimizers such as Adam, with mean squared error (MSE) as the loss function iteration parameter. The final output is an initial static prediction value for formation pore pressure that has good generalization ability for the target work area.
[0023] Step A2: Preprocess the logging data of the drilled wells to remove outliers and obtain the drilled well dataset after outlier removal. ; The preprocessing includes outlier rejection and data normalization.
[0024] The outliers are rejected by using unknown filtering methods to detect and reject gross errors in the data sequence, but valid outliers are retained.
[0025] The data normalization process involves scaling the data to the [0, 1] interval using a max-min normalization method, where the expression for the max-min normalization method is: (Equation 1) In the formula, This refers to the data before data normalization. and These are the minimum and maximum values of the data range, respectively.
[0026] Step A3: Obtain background data of the target well's target region The target domain background data of the target well is obtained through a relevant alignment algorithm. and the drilled well dataset after removing outliers Establish correlations to obtain the source domain data distribution after correlation alignment transformation; The English name of the correlation alignment algorithm is COR relation Alignment, abbreviated as CORAL, also known as the CORAL algorithm. The CORAL algorithm achieves distribution adaptation by minimizing the difference between the covariance matrices of the source domain and the target domain, while simultaneously unifying the feature space and reducing distribution differences.
[0027] A relevant alignment algorithm is used for domain adaptation to eliminate distribution bias caused by differences in data sources.
[0028] Specifically, the target domain background data of the target well is obtained through a relevant alignment algorithm. and the drilled well dataset after removing outliers Establishing correlations and obtaining the source domain data distribution after correlation alignment transformation includes the following steps: 1) Based on the target well's target domain background data Obtain target domain background data of the target well covariance matrix ; Specifically, based on the logging data of the target well, the background data of the target area is obtained by using the low-frequency background pressure trend data derived from the seismic velocity spectrum of the target area. The data is then processed and normalized to serve as the logging data for the target well; then, based on the target domain background data of the target well... Obtain target domain background data of the target well covariance matrix .
[0029] 2) Based on the drilled well dataset after removing outliers Obtain the drilled well dataset after removing outliers covariance matrix ; 3) Based on the relevant alignment algorithm, the target domain background data of the target well is... covariance matrix and the drilled well dataset after outlier removal covariance matrix Establish the correlation to obtain the source domain data distribution after correlation alignment transformation.
[0030] The Frobenius norm distance between the covariance matrix Cs of the source domain data features and the covariance matrix Ct of the target domain data features of the target well is achieved through a correlation alignment algorithm to achieve distribution alignment.
[0031] Specifically, the expression for the relevant alignment algorithm is: (Equation 2) In the formula, This represents the drilled well dataset after outlier removal. The covariance matrix; This represents the background data of the target domain of the target well. The covariance matrix; A represents a linear transformation matrix; AT represents the transpose of the linear transformation matrix A; The covariance matrix represents the source domain data features after transformation by the linear transformation matrix A; denoted by the squared Frobenius norm of the matrix, it is used to measure the distance between the covariance matrix of the transformed source domain data features and the covariance matrix of the target domain data features; This represents finding the optimal linear transformation matrix A such that the distance between the covariance matrix of the source domain data features after the correlation alignment transformation and the covariance matrix of the target domain data features of the target well is minimized.
[0032] The method of achieving distribution alignment using relevant alignment algorithms avoids complex distribution estimation and subspace projection, and is computationally efficient and stable. This transformation operation converts the source domain data into a shared feature space consistent with the background data distribution of the target work area, achieving unsupervised alignment of cross-block features.
[0033] Step A4: Using the source domain data distribution after the relevant alignment transformation as input data, the model is substituted into the initial static prediction model of formation pore pressure for calculation. The model obtains the relationship between the logging curve data of the target well and the formation pore pressure label of the target well based on the relationship between the logging curve data of the drilled well and the formation pore pressure label of the drilled well.
[0034] Step B: Obtain real-time logging data of the target well during drilling operations, preprocess the target well logging data to remove outliers, and obtain the target well logging dataset after removing outliers.
[0035] During the drilling of the target well, logging data uploaded by the measurement-while-drilling (LWD) system is received in real time. The received data undergoes cleaning and normalization preprocessing sequentially. The preprocessed LWD time-series characteristics are then input into the initial static prediction model of formation pore pressure. The system calculates and continuously outputs the predicted formation pore pressure at the current drill bit depth in real time, enabling real-time early warning of abnormal drilling pressure.
[0036] Step C: Substitute the measured dataset of the target well after removing outliers into the initial static prediction model of formation pore pressure. The model obtains the predicted data value of formation pore pressure of the target well based on the relationship between the data of the logging curve of the target well and the formation pore pressure label of the target well.
[0037] Step D: Verification of the predicted formation pore pressure data for the target well, including the following steps: Step D1: Obtain the measured value of the formation pore pressure label of the target well; Step D2: Output the predicted value of the formation pore pressure label of the target well through the initial static prediction model of formation pore pressure based on the sequence neural network, and compare the measured value of the formation pore pressure label of the target well with the predicted value of the formation pore pressure label of the target well. Step D3: If the measured value of the formation pore pressure label of the target well is within the error range of the predicted value of the formation pore pressure label of the target well, then the predicted value of the formation pore pressure of the target well is valid; otherwise, repeat step B, re-acquire the measured data set of the target well after removing outliers, and repeat steps C to D3 until the predicted value of the formation pore pressure of the target well is valid.
[0038] Specifically, during the drilling of the target well, a dynamic update trigger point is set: when the actual formation pore pressure value at a specific depth is obtained through formation pressure testing while drilling (such as LWD-PWD tool measurement), intermittent core testing, or encountering characteristic strata, the system automatically extracts the actual formation pore pressure label. The logging-while-drilling feature sequence at the corresponding depth is combined with the actual formation pore pressure label to construct a "target domain incremental sample pair" with an actual label.
[0039] The method for predicting formation pore pressure in a target well disclosed in Embodiment 1 of this invention is based on logging data from existing wells and real-time drilling data from the target well. Its core steps include constructing a multi-source database, establishing a pre-trained model through transfer learning, and using drilling data for real-time prediction and online dynamic optimization. This method effectively solves the technical problems of low prediction accuracy, poor model generalization ability, and inability to update the model while drilling in traditional methods in new exploration areas. It can be widely applied to the exploration and development of offshore and onshore oil and gas fields, providing key technical support for drilling safety.
[0040] Example 2: A device for predicting formation pore pressure in a target well based on logging data from drilled wells and real-time drilling data from the target well. Embodiment 2 of the present invention provides an apparatus for predicting formation pore pressure of a target well based on logging data from a drilled well and real-time drilling data from the target well, comprising: The first processing unit is used to acquire the data of the logging curve of the target well and the relationship between the formation pore pressure label of the target well; The second processing unit is used to obtain the target well's real-time logging data during drilling construction, preprocess the target well's real-time logging data, remove outliers, and obtain the target well's real-time logging dataset after removing outliers. The third processing unit is used to input the measured dataset of the target well after removing outliers into the initial static prediction model of formation pore pressure. The model obtains the predicted data value of formation pore pressure of the target well based on the relationship between the data of the logging curve of the target well and the formation pore pressure label of the target well.
[0041] Example 3: A computer-readable storage medium Embodiment 3 of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method of Embodiment 1.
[0042] Example 4: A computer device Embodiment 4 of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method of Embodiment 1. Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for predicting formation pore pressure in a target well based on logging data from an existing well and real-time drilling data from the target well, characterized in that, include: Step A: Obtain the relationship between the logging curve data of the target well and the formation pore pressure label of the target well; Step B: Obtain real-time logging data of the target well during drilling operations, preprocess the target well logging data to remove outliers, and obtain the target well logging dataset after removing outliers. Step C: Substitute the measured dataset of the target well after removing outliers into the initial static prediction model of formation pore pressure. The model obtains the predicted data value of formation pore pressure of the target well based on the relationship between the data of the logging curve of the target well and the formation pore pressure label of the target well.
2. The method according to claim 1, characterized in that, Step A includes the following steps: Step A1: Obtain the logging curve data and formation pore pressure labels of the drilled wells. Train the logging curve data and formation pore pressure labels of the drilled wells using the initial static prediction model of formation pore pressure to establish the relationship between the logging curve data and formation pore pressure labels of the drilled wells. Step A2: Preprocess the logging data of the drilled wells to remove outliers and obtain the drilled well dataset after outlier removal. ; Step A3: Obtain background data of the target well's target region The target domain background data of the target well is obtained through a relevant alignment algorithm. and the drilled well dataset after removing outliers Establish correlations to obtain the source domain data distribution after correlation alignment transformation; Step A4: Using the source domain data distribution after the relevant alignment transformation as input data, the model is substituted into the initial static prediction model of formation pore pressure for calculation. The model obtains the relationship between the logging curve data of the target well and the formation pore pressure label of the target well based on the relationship between the logging curve data of the drilled well and the formation pore pressure label of the drilled well.
3. The method according to claim 2, characterized in that, The target domain background data of the target well is obtained through a relevant alignment algorithm. and the drilled well dataset after removing outliers Establishing correlations and obtaining the source domain data distribution after correlation alignment transformation includes the following steps: 1) Based on the target well's target domain background data Obtain target domain background data of the target well covariance matrix ; 2) Based on the drilled well dataset after removing outliers Obtain the drilled well dataset after removing outliers covariance matrix ; 3) Based on the relevant alignment algorithm, the target domain background data of the target well is... covariance matrix and the drilled well dataset after outlier removal covariance matrix Establish the correlation to obtain the source domain data distribution after correlation alignment transformation.
4. The method according to claim 3, characterized in that, The expression for the relevant alignment algorithm is: In the formula, This represents the drilled well dataset after outlier removal. The covariance matrix; This represents the background data of the target domain of the target well. The covariance matrix; A represents a linear transformation matrix; AT represents the transpose of the linear transformation matrix A; The covariance matrix represents the features of the source domain data after transformation by the linear transformation matrix A; Denotes the squared Frobenius norm of a matrix; This indicates the search for the optimal linear transformation matrix A.
5. The method according to claim 1, characterized in that, Also includes: Step D: Verification of the predicted formation pore pressure data for the target well, including the following steps: Step D1: Obtain the measured value of the formation pore pressure label of the target well; Step D2: Output the predicted value of the formation pore pressure label of the target well through the initial static prediction model of formation pore pressure based on the sequence neural network, and compare the measured value of the formation pore pressure label of the target well with the predicted value of the formation pore pressure label of the target well. Step D3: If the measured value of the formation pore pressure label of the target well is within the error range of the predicted value of the formation pore pressure label of the target well, then the predicted value of the formation pore pressure of the target well is valid; otherwise, repeat step B, re-acquire the measured data set of the target well after removing outliers, and repeat steps C to D3 until the predicted value of the formation pore pressure of the target well is valid.
6. A device for predicting formation pore pressure in a target well based on logging data from an existing well and real-time drilling data from the target well, characterized in that, include The first processing unit is used to acquire the data of the logging curve of the target well and the relationship between the formation pore pressure label of the target well; The second processing unit is used to obtain the target well's real-time logging data during drilling construction, preprocess the target well's real-time logging data, remove outliers, and obtain the target well's real-time logging dataset after removing outliers. The third processing unit is used to input the measured dataset of the target well after removing outliers into the initial static prediction model of formation pore pressure. The model obtains the predicted data value of formation pore pressure of the target well based on the relationship between the data of the logging curve of the target well and the formation pore pressure label of the target well.
7. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method described in any one of claims 1 to 5.
8. A computer device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the method according to any one of claims 1 to 5.