Geological drilling process bit pressure multi-step prediction method based on multi-model fusion

Abstract

The invention relates to the field of geological drilling process intelligent monitoring and control, and discloses a geological drilling process bit pressure multi-step prediction method based on multi-model fusion, and the method comprises the steps: collecting the data of a geological drilling process, and carrying out the preprocessing of the data of the geological drilling process; establishing a smooth regression model, a jump classification model and a bit pressure change regression model by using a plurality of models based on the lightweight gradient lifting tree; fusing the plurality of prediction models by adopting a weighted least square method; and on the basis of the fused model, bit pressure prediction values of multiple depth points in the drilling process are generated through multi-step prediction. According to the method, the overall accuracy and robustness of bit pressure prediction are effectively improved, the exploration efficiency can be improved, the operation cost is reduced, and the method has important engineering application value and economic significance.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring and control of geological drilling processes, specifically involving a multi-step prediction method for drilling pressure in geological drilling processes based on multi-model fusion. Background Technology

[0002] Geological drilling is a crucial step in mineral resource exploration. Drill pressure (DP), as a vital indicator of drilling progress and stability, is critically important for improving exploration efficiency, shortening the exploration cycle, and reducing costs. To accurately predict DP changes during drilling, a multi-step prediction method combines drilling depth with relevant parameters to continuously predict DP at multiple depth points. This method can predict DP changes at multiple future depths, starting from the current depth, helping operators understand DP trends in real time and providing guidance for subsequent drilling operations. Through multi-step prediction, operators can not only grasp the current drilling status but also anticipate DP trends, thereby optimizing drilling plans and improving operational efficiency and safety.

[0003] Most current methods for predicting drilling pressure (DPP) rely primarily on time-series data processing techniques to predict DPP at each time step. While these methods can reflect the real-time state of the drilling process to some extent, they often overfit local data and struggle to comprehensively and accurately capture the complex dynamic changes during drilling. Therefore, developing a DPP prediction method that incorporates drilling depth data for geological drilling processes can reduce noise and interference in time-series data, better extract spatial features, and has significant practical implications.

[0004] Currently, single-step prediction methods are the most common approach for drilling pressure forecasting, typically relying on local predictions of the current drilling depth or time point. However, this method is often limited to analyzing the current drilling status and struggles to fully reflect long-term trends and dynamic changes in depth during the drilling process. Therefore, developing multi-step prediction methods can enable multi-step prediction of drilling pressure at multiple depth points, helping operators identify potential risks or changes during future drilling and optimize work plans. Summary of the Invention

[0005] This invention aims to address the technical problem that existing single-step drill pressure prediction methods cannot fully reflect the trend and depth of the drilling process, leading to low drilling efficiency. To solve this problem, this invention provides a multi-step drill pressure prediction method based on multi-model fusion for geological drilling processes.

[0006] This invention provides a multi-step prediction method for drilling pressure in geological drilling processes based on multi-model fusion, specifically including the following steps: S1: Collect data on the geological drilling process and preprocess the geological drilling data; S2: Use multiple lightweight gradient boosting tree models to establish smooth regression model, jump classification model and drill pressure change regression model respectively; S3: Weighted least squares method is used to fuse multiple prediction models; S4: Based on the fusion model, the drilling pressure prediction values ​​at multiple depth points during the drilling process are generated through multi-step prediction.

[0007] A computer device includes at least: one or more processors; and a memory storing one or more computer programs; wherein the processors invoke the computer programs to implement the steps of the multi-step prediction method for drilling pressure in the geological drilling process based on multi-model fusion.

[0008] A computer storage device stores a computer program that is invoked by a processor to implement the steps of the multi-step prediction method for drilling pressure in the geological drilling process based on multi-model fusion.

[0009] The technical solution provided by this invention has the following beneficial effects: 1. Significantly improved prediction accuracy: By integrating three lightweight gradient boosting tree models—smooth regression, jump classification, and drill pressure change regression—the overall accuracy and robustness of drill pressure prediction are effectively improved by comprehensively utilizing the different models' ability to capture different features of the drilling process.

[0010] 2. Comprehensive reflection of drilling dynamics: The jump classification model can identify abnormal fluctuations or sudden changes in drilling pressure, the smooth regression model focuses on trend changes, and the drilling pressure change regression model captures the magnitude of changes. The three work together to make the prediction results more comprehensive and realistic in reflecting the complex dynamic behavior in the drilling process.

[0011] 3. Supports multi-step forward prediction: By adopting a multi-step training and prediction mechanism, it can continuously predict the drilling pressure at multiple depth points in the future, helping operators to understand the trend of drilling pressure changes in advance, providing a reliable basis for adjusting drilling parameters and optimizing work plans, and helping to prevent risks and improve operational safety.

[0012] 4. High adaptability and ease of implementation: Based on a lightweight gradient boosting tree model, it boasts high training efficiency and low resource consumption, making it suitable for real-time or near-real-time drilling monitoring scenarios. The weighted least squares fusion strategy can dynamically adjust weights based on the performance of different models, further enhancing the adaptability and stability of the prediction system.

[0013] 5. Promote intelligent drilling management: This method can provide core prediction functions for automated drilling control and intelligent decision support systems, which helps to promote the development of geological drilling processes towards intelligence and precision, improve exploration efficiency, reduce operating costs, and has important engineering application value and economic significance. Attached Figure Description

[0014] The present invention will be further described below with reference to the accompanying drawings and examples. In the accompanying drawings: Figure 1 This is a schematic diagram of the overall process of a multi-step prediction method for drilling pressure in geological drilling based on multi-model fusion according to the present invention. Figure 2 This is a schematic diagram of the prediction results of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention.

[0016] Example 1 Please refer to Figure 1 This invention provides a multi-step prediction method for drilling pressure in geological drilling processes based on multi-model fusion, the main steps of which are as follows: S1: Collect data on the geological drilling process and preprocess the geological drilling data; It should be noted that step S1 specifically includes: S11. Collect time-series data during the sensor drilling process, including drilling depth, drilling pressure, drilling speed, riser pressure, surface torque, rotational speed, mud pit flow rate, and mud pit density parameters; S12. Perform preliminary cleaning on the collected time-series data to remove sensor malfunctions and abnormal data; S13. Resample the time series data according to the drilling depth, with each 1 meter of drilling depth as a data point, and each data point representing a specific depth and its corresponding time series data.

[0017] Specifically, time-series data is collected during the drilling process, including parameters such as drilling depth, drill pressure, drilling speed, riser pressure, surface torque, rotational speed, mud pit flow rate, and mud pit density. The collected time-series data undergoes preliminary cleaning to remove sensor malfunctions and abnormal data. Then, the time-series data is resampled according to drilling depth, with each meter of drilling depth serving as a data point. Each data point represents a specific depth and its corresponding drill pressure, drilling speed, and other parameters.

[0018] S2: Use multiple lightweight gradient boosting tree models to establish smooth regression model, jump classification model and drill pressure change regression model respectively; It should be noted that step S2 specifically includes the following process: S21. Use a lightweight gradient boosting tree to build a smooth regression model. Input historical geological drilling process data with drilling depth as the main axis, train the model, and the output of the model is the preliminary prediction value of the drilling pressure, which represents the drilling pressure at a specific depth point. S22. Use a lightweight gradient boosting tree to build a jump classification model. The input data is historical geological drilling process data with drilling depth as the main axis. The model is trained and the output of the model is a binary label (e.g., 0 represents normal drilling and 1 represents jump), which is used to determine whether there are abnormal fluctuations or sudden changes in drilling pressure. S23. Construct a regression model for drilling pressure variation using a lightweight gradient boosting tree. The input data is historical geological drilling process data with drilling depth as the main axis. Train the model and the output of the model is the amount of change in drilling pressure, i.e. the magnitude of the change in drilling pressure. It should be noted that in steps S21 to S23, the training of the corresponding model is carried out in a multi-step training with a total of m steps. Each step trains a model separately, and a total of m depth points are predicted.

[0019] That is, for step S21, for multi-step prediction ( m The process involves training a separate smooth regression model at each step. Specifically, the model is trained sequentially to predict future drilling pressure starting from the current depth point, until all predictions are completed. m Drill pressure value at each depth point.

[0020] For step S22, for multi-step prediction ( m The process involves training a separate jump classification model at each step. Specifically, the model is trained sequentially, predicting future drill pressure jumps starting from the current depth point, until all jumps are predicted. m The transition between depth points.

[0021] In step S23, for the multi-step prediction (40 steps), a separate regression model for drilling pressure change is trained for each step. Specifically, the model is trained sequentially, predicting the future change in drilling pressure starting from the current depth point, until all predictions are completed. m Drill pressure variation at various depth points.

[0022] S3: Weighted least squares method is used to fuse multiple prediction models; Specifically, step S3 is as follows: S31. Perform weighted fusion on the corresponding models to obtain the final predicted value for each depth point. d j ，j =1, 2, 3,..., m The specific formula is as follows:

[0023] in This is the initial drilling pressure prediction value from the smooth regression model. This represents the predicted probability value of drill pressure jump in the jump classification model. This represents the predicted value of the change in drilling pressure from the regression model. w 1, w 2, w 3 represents the weights of the sliding regression model, the jump classification model, and the drilling pressure change regression model, respectively; S32, Weight w 1, w 2, w 3. Based on the least squares method, the prediction error of each model is optimized to determine the value, and the specific calculation is as follows:

[0024] in, q i For the first i The mean squared error of each model on the training set i =1, 2, 3.

[0025] S4: Based on the fusion model, the drilling pressure prediction values ​​at multiple depth points during the drilling process are generated through multi-step prediction.

[0026] It should be noted that step S4 is as follows: Input different drilling data into the fused model in step S3. n Based on historical drilling data, the model provides corresponding depth values ​​at different drilling stages according to the input well's historical data. m Drilling pressure prediction for each step.

[0027] Example 2 This invention discloses a multi-step prediction method for drilling pressure (DPP) during geological drilling based on multi-model fusion. First, relevant data from the geological drilling process is collected and processed. Then, multiple specific prediction models, including a smooth regression model, a jump classification model, and a DPP variation regression model, are used to train the data. Next, the prediction results of the multiple models are fused using weighted least squares to obtain a more accurate DPP prediction value. Finally, based on the fused model, multi-step DPP predictions are provided for different drilling history data. The specific steps are as follows: (1) Collect geological drilling process data to obtain raw sample data Historical data of the geological drilling process is stored in the local database of the industrial control computer in the control room in the form of time-series data. Based on the drilling operation records, historical data of relevant parameters such as drilling depth, drilling pressure, drilling speed, riser pressure, surface torque, rotational speed, mud pit flow rate, and mud pit density are collected to form the raw sample data. A total of 8 detection parameters are included.

[0028] (2) Data preprocessing The collected raw sample data is preprocessed. First, the time-series data is resampled according to drilling depth, with each meter of drilling depth serving as a data point. A database is then constructed based on this format.

[0029] (3) Train multiple models Multiple lightweight gradient boosting tree models were trained for drill pressure prediction. First, a smooth regression model was trained, taking historical drilling data as input, to predict drill pressure at a specific depth. The model was trained for 1200 iterations with a step size of 0.03 and 15 leaf nodes. Second, a jump classification model was trained, taking historical data as input, to predict the probability of jumps. This model was trained for 800 iterations with a step size of 0.04. Finally, a drill pressure change regression model was trained, taking historical data as input, to predict the amount of change in drill pressure. This model was trained for 1200 iterations with a step size of 0.03 and 15 leaf nodes. All models received 160 steps of data as input and predicted drill pressure at 40 steps.

[0030] (4) Multi-model fusion prediction Please refer to Figure 2 , Figure 2 This is a schematic diagram of the prediction results of the present invention; the prediction results obtained from the smooth regression model, the jump classification model, and the drilling pressure change regression model are fused using the weighted least squares method to calculate the weight of each model. w 1. w 2 and w 3. The initial weights are set to [0.34, 0.33, 0.33]. These weights are optimized using the least squares method to ensure a more accurate final fused prediction. Finally, based on the fused weights, the drilling pressure at a future depth of 40 points is predicted to obtain the final drilling pressure prediction result.

[0031] The results of this embodiment show that the present invention can realize multi-step prediction of drilling pressure during geological drilling, which has significant economic and application value.

[0032] Example 3 A computer device includes at least: one or more processors; and a memory storing one or more computer programs; wherein the processors invoke the computer programs to implement the steps of the multi-step prediction method for drilling pressure in the geological drilling process based on multi-model fusion.

[0033] A computer storage device stores a computer program that is invoked by a processor to implement the steps of the multi-step prediction method for drilling pressure in the geological drilling process based on multi-model fusion.

[0034] The preferred embodiments of the present invention disclosed above are only for the purpose of illustrating the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation described herein. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can better understand and utilize the present invention.

Claims

1. A multi-step prediction method for drilling pressure in geological drilling processes based on multi-model fusion, characterized in that, Includes the following steps: S1: Collect data on the geological drilling process and preprocess the geological drilling data; S2: Use multiple lightweight gradient boosting tree models to establish smooth regression model, jump classification model and drill pressure change regression model respectively; S3: Weighted least squares method is used to fuse multiple prediction models; S4: Based on the fusion model, the drilling pressure prediction values ​​at multiple depth points during the drilling process are generated through multi-step prediction.

2. The multi-step prediction method for drilling pressure in geological drilling process based on multi-model fusion according to claim 1, characterized in that, Step S1 specifically includes: S11. Collect time-series data during the sensor drilling process, including drilling depth, drilling pressure, drilling speed, riser pressure, surface torque, rotational speed, mud pit flow rate, and mud pit density parameters. S12. Perform preliminary cleaning on the collected time-series data to remove sensor malfunctions and abnormal data; S13. Resample the time series data according to the drilling depth, with each 1 meter of drilling depth as a data point, and each data point representing a specific depth and its corresponding time series data.

3. The multi-step prediction method for drilling pressure in geological drilling process based on multi-model fusion according to claim 1, characterized in that, Step S2 specifically includes the following processes: S21. Use a lightweight gradient boosting tree to build a smooth regression model. Input historical geological drilling process data with drilling depth as the main axis, train the model, and the output of the model is the preliminary prediction value of the drilling pressure, which represents the drilling pressure at a specific depth point. S22. Use a lightweight gradient boosting tree to build a jump classification model. The input data is historical geological drilling process data with drilling depth as the main axis. Train the model and the output of the model is a binary label to determine whether there are abnormal fluctuations or sudden changes in drilling pressure. S23. Construct a regression model for drilling pressure variation using a lightweight gradient boosting tree. The input data is historical geological drilling process data with drilling depth as the main axis. Train the model and the output of the model is the amount of change in drilling pressure, i.e. the magnitude of the change in drilling pressure.

4. The multi-step prediction method for drilling pressure in geological drilling process based on multi-model fusion according to claim 3, characterized in that, In steps S21 to S23, the training of the corresponding model is carried out by multi-step training, which includes a total of m steps. Each step trains a model separately, and a total of m depth points are predicted.

5. The multi-step prediction method for drilling pressure in geological drilling process based on multi-model fusion according to claim 4, characterized in that, Step S3 is as follows: S31. Perform weighted fusion on the corresponding models to obtain the final predicted value for each depth point. d j ，j =1, 2, 3, ..., m The specific formula is as follows: in This is the initial drilling pressure prediction value from the smooth regression model. This represents the predicted probability value of drill pressure jump in the jump classification model. This represents the predicted value of the change in drilling pressure from the regression model. w 1, w 2, w 3 represents the weights of the sliding regression model, the jump classification model, and the drilling pressure change regression model, respectively; S32, Weight w 1, w 2, w 3. Based on the least squares method, the prediction error of each model is optimized to determine the value, and the specific calculation is as follows: in, q i For the first i The mean squared error of each model on the training set i =1, 2, 3.

6. The multi-step prediction method for drilling pressure in geological drilling process based on multi-model fusion as described in claim 5, characterized in that, Step S4 is as follows: Input different drilling data into the fused model in step S3. n Based on historical drilling data, the model provides corresponding depth values ​​at different drilling stages according to the input well's historical data. m Drilling pressure prediction for each step.

7. A computer device, characterized in that, It includes at least: one or more processors; a memory storing one or more computer programs; wherein the processor calls the computer programs to implement: the steps of the multi-step prediction method for drilling pressure in geological drilling process based on multi-model fusion as described in any one of claims 1-6.

8. A computer storage device, characterized in that, A computer program is stored, which is invoked by a processor to implement the steps of the multi-step prediction method for drilling pressure in the geological drilling process based on multi-model fusion as described in any one of claims 1-6.

Images