A method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging

Abstract

This invention relates to the field of hydrogeological exploration technology and discloses a method for accurately predicting aquifer inflow by integrating conventional and electromagnetic flow logging. The method includes the following steps: Step 1: Well logging data acquisition. Electromagnetic flow logging and conventional logging are performed simultaneously in the target exploration borehole to obtain static and dynamic flow response amplitude curves, apparent resistivity curves, natural gamma curves, and sonic transit time data. This method for accurately predicting aquifer inflow by integrating conventional and electromagnetic flow logging data constructs a key parameter system covering four dimensions: "water storage space - recharge power - water storage capacity - permeability resistance." It also utilizes the powerful nonlinear fitting capability of a generalized regression neural network (GRNN) to establish a prediction model, effectively overcoming the shortcomings of traditional theoretical formulas in adapting to complex geological conditions.

Description

Technical Field

[0001] This invention relates to the field of hydrogeological exploration technology, specifically to a method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging. Background Technology

[0002] In the field of mineral resource extraction, aquifer inrush is a major cause of water hazards, and accurate prediction of inrush volume is crucial for ensuring safe production. Traditional methods mainly rely on pumping tests combined with theoretical formulas such as the Dubuis formula for calculation. However, these methods have significant drawbacks such as long cycle time, high cost, and large engineering workload. Furthermore, the theoretical formulas have poor adaptability under complex and heterogeneous geological conditions, making it difficult to meet the prediction accuracy requirements of modern efficient exploration and disaster prevention.

[0003] Well logging technology is an efficient means of obtaining subsurface geological information. Conventional well logging (such as resistivity and natural gamma logging) can be used to identify aquifers, but it is difficult to directly and accurately obtain key quantitative parameters such as permeability. Electromagnetic flow logging can directly measure the velocity and flow rate of fluid in the borehole, reflecting the water conductivity of the aquifer, but it cannot independently complete the comprehensive identification of formation lithology. Currently, existing technologies have not yet formed a mature method to effectively integrate the data from both methods and utilize intelligent algorithms to explore the complex nonlinear relationship between parameters and water inflow to achieve efficient and accurate prediction, resulting in the underutilization of the comprehensive utilization value of well logging data. Summary of the Invention

[0004] To address the problems mentioned in the background section, this invention provides the following technical solution: a method for accurately predicting aquifer inflow by integrating conventional and electromagnetic flow logging, comprising the following steps:

[0005] Step 1: Well logging data acquisition. Simultaneously perform electromagnetic flow logging and conventional logging within the target exploration borehole to acquire static and dynamic flow response amplitude curves, apparent resistivity curves, natural gamma curves, and sonic transit time data. Step 2: Aquifer identification and parameter extraction. Based on the well logging data acquired in Step 1, identify the aquifer location and calculate the aquifer thickness M, aquifer floor position H2, and aquifer porosity φ. s and aquifer clay content Four key hydrogeological parameters; Step 3: Construct a water inflow prediction model based on a generalized regression neural network (GRNN), using the aquifer thickness M, bottom plate location H2, and porosity φ. s mud content As input features, the measured water inflow is used as the output label to train and validate the GRNN model; Step 4: Aquifer water inflow prediction. The four key parameters obtained from Step 1 and Step 2 of the borehole to be predicted are input into the validated GRNN model, and the predicted aquifer water inflow value of the borehole is output.

[0006] Preferably, the selection criteria for the four key hydrogeological parameters in step 2 are as follows: aquifer thickness M represents the water storage space, aquifer floor position H2 represents the recharge force, and aquifer porosity φ... s Characterizing water storage capacity, aquifer clay content Characterizes water permeability resistance.

[0007] Preferably, the construction of the GRNN prediction model in step 3 specifically includes: 3.1 Data preprocessing: collecting the four key parameters and corresponding measured water inflow from multiple boreholes in the target area, forming a dataset and performing min-max standardization, and randomly dividing it into training set and test set;

[0008] 3.2 Model Training: The four parameters of the training set are used as input features, and the measured inflow rate is used as the output label. These are input into the GRNN model for training. The smoothing factor of the model is optimized within a preset range using a grid search method. 3.3 Model Validation: The input features of the test set are substituted into the trained model, and the average relative error between the predicted inflow rate and the measured value is calculated. If the average relative error does not exceed 10%, the model is considered to have passed validation.

[0009] Preferably, the aquifer thickness M in step 2 is calculated using the formula M=H2-H1 based on the aquifer top depth H1 and bottom depth H2 determined by electromagnetic flow logging.

[0010] Preferably, the porosity φ of the aquifer in step 2 s Based on the sonic transit time data obtained from conventional well logging, using the formula φ s The calculation yields the result, where Δt is the measured acoustic time difference, and Δt_m For the acoustic transit time of the rock skeleton, t_f is the acoustic transit time of the pore fluid.

[0011] Preferably, the aquifer clay content in step 2 is... Based on the natural gamma ray GR data obtained from conventional well logging, using the formula... The calculation yields GR, where GR is the measured value. and These are the minimum and maximum values ​​of the natural gamma for that layer, respectively.

[0012] Preferably, the qualitative judgment of the aquifer in step 2 is based on the following: analyzing the static flow response amplitude curve obtained from electromagnetic flow logging. If the curve shows an increasing trend from bottom to top, it is determined to be a water-bearing layer; if it shows a decreasing trend, it is determined to be a water-absorbing layer; if there is no obvious change, it is determined to be an impermeable layer. The inflection point of the amplitude change of the curve is the aquifer interface.

[0013] Preferably, the electromagnetic flow logging in step 1 includes: obtaining a static flow response amplitude curve by measuring under natural static conditions of the borehole, and obtaining a dynamic flow response amplitude curve by measuring under three pumping tests with different drawdowns.

[0014] Preferably, step 2 also includes a step of calculating the borehole water inflow based on the electromagnetic flow logging data through the "amplitude-velocity-flow rate" conversion system, specifically: calculating the water flow velocity v = A·K based on the flow response amplitude A and the instrument calibration coefficient K, and then calculating the water inflow Q = v·S based on the borehole cross-sectional area S.

[0015] Preferably, the method is applicable to the prediction of aquifer water inflow and the prevention of water hazards during the mining of coal and metal mineral resources.

[0016] Compared with existing technologies, this invention provides a method for accurate prediction of aquifer inflow that integrates conventional and electromagnetic flow logging, and has the following beneficial effects:

[0017] 1. This method for accurately predicting aquifer inflow by integrating conventional and electromagnetic flow logging data constructs a key parameter system covering four dimensions: "reservoir space, recharge power, storage capacity, and permeability resistance." It utilizes the powerful nonlinear fitting capability of a generalized regression neural network (GRNN) to establish a prediction model, effectively overcoming the shortcomings of traditional theoretical formulas in adapting to complex geological conditions. Examples show that the average relative error of the prediction can be reduced to approximately 4.76%, improving accuracy by more than 75% compared to the traditional pumping test combined with formula method (which typically has an error of 15%-25%).

[0018] 2. This method for accurately predicting aquifer inflow, which integrates conventional and electromagnetic flow logging, eliminates the need for large-scale, long-term field pumping tests, relying primarily on well logging operations to acquire data. The entire process from data acquisition to prediction for a single well can be shortened to approximately 2 days, with a cost of about 20,000 yuan. Compared to traditional stratified pumping tests (approximately 9 days and 72,000 yuan), this method reduces time by 77.8% and costs by 72.2%, demonstrating significant economic benefits.

[0019] 3. This method for accurately predicting aquifer inflow by integrating conventional and electromagnetic flow logging differs from traditional pumping tests, which require large-scale extraction of groundwater and disturb the original flow field and stress state of the aquifer. This invention uses only logging methods, avoiding damage to the original hydrogeological conditions of the aquifer caused by pumping operations, making the evaluation process more environmentally friendly and reliable.

[0020] 4. This method for accurately predicting aquifer inflow, integrating conventional and electromagnetic flow logging, combines efficient geophysical logging technology with advanced intelligent algorithms. The process is clear and highly operable. It is applicable not only to loosely porous aquifers but also demonstrates excellent predictive capabilities for complex geological conditions such as fractured aquifers. It provides an efficient, accurate, and universal technical means for hydrogeological exploration, water hazard assessment, and prevention and control strategies in coal mines, metal mines, and other mineral resource extraction areas, possessing broad engineering application and promotion value. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of an aquifer parameter calculation method that integrates flow logging information according to the present invention;

[0022] Figure 2 This is a schematic diagram illustrating the qualitative judgment of aquifers based on amplitude curves, velocity curves, and flow rate curves according to the present invention.

[0023] Figure 3 This is a diagram showing the borehole prediction effect and error in Example S-5 of the present invention;

[0024] Figure 4 This is a diagram showing the aquifer identification and parameter data of the present invention;

[0025] Figure 5 This is a diagram showing the basic information data of the aquifer in this invention;

[0026] Figure 6 This is a diagram of the core parameter data extracted in this invention;

[0027] Figure 7 This is a comparative data chart showing the water inflow prediction results of different methods in this invention;

[0028] Figure 8 This is a data chart showing the comparison results of different indicators in this invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] Please see Figure 1-8 The present invention provides a technical solution:

[0031] Example

[0032] Implementation target: Taking the exploration borehole ZK-1 in a coal mine in Shaanxi as an example, this borehole passes through the Quaternary loose aquifer.

[0033] I. Well Logging Data Acquisition

[0034] Equipment preparation:

[0035] Electromagnetic flow logging tool: PSCL-1 type, flow measurement range 0-10 m³ / h.

[0036] Conventional logging system: KH-2 integrated logging instrument, which can acquire curves of apparent resistivity, natural gamma, sonic transit time, compensated density, etc.

[0037] Auxiliary equipment: logging cable, depth encoder, data acquisition system, submersible pump (for pumping tests).

[0038] Data collection steps:

[0039] Conventional logging: The KH-2 logging tool is lowered to the bottom of the borehole and pulled up at a constant speed with a sampling interval of 0.05 m to obtain the apparent resistivity curve, natural gamma curve and sonic transit time curve of the entire borehole section.

[0040] Electromagnetic flow logging (static): Under the natural state of the borehole without pumping operations, the PSCL-1 electromagnetic flow logging instrument is lowered to the bottom of the borehole. Measurements are taken at depths where aquifers may exist (estimated based on geological data), both during the lifting and lowering phases, to obtain the flow response amplitude curve under static conditions.

[0041] Electromagnetic flow logging (dynamic): A submersible pump is installed inside the borehole and lowered below the aquifer. Three steady-flow pumping tests are conducted sequentially at different drawdown depths (e.g., S1=3m, S2=6m, S3=9m). After each drawdown stabilizes, the lifting and lowering measurement process is repeated to obtain the dynamic flow response amplitude curves corresponding to the three drawdown depths.

[0042] Depth calibration: Ensure accurate matching of depth coordinates between conventional logging and electromagnetic flow logging, with errors controlled within ±0.1m.

[0043] II. Aquifer Identification and Parameter Extraction

[0044] Aquifer identification:

[0045] Analyze the static flow response amplitude curve. For example... Figure 2 As shown, at a depth of 32.5 m, the amplitude curve shows a clear inflection point from small to large, and the amplitude shows an increasing trend with increasing depth, indicating that this is the top interface of the aquifer (H1=32.5 m). At a depth of 40.5 m, the amplitude curve shows an inflection point from large to small, indicating that this is the bottom interface of the aquifer (H2=40.5 m).

[0046] Parameter calculation:

[0047] Aquifer thickness M: Based on the identified top and bottom plate depths, the formula is: M = H2 - H1 = 40.5 m - 32.5 m = 8.0 m.

[0048] Location H2 at the bottom of the aquifer: directly taken as 40.5 m.

[0049] Aquifer porosity φ s Within the 32.5 m to 40.5 m aquifer section, the average acoustic transit time curve Δt = 285 μs / m was read. The sandstone framework transit time Δt_ma = 180 μs / m and the formation water transit time Δt_f = 620 μs / m are known. Substitute these values ​​into the formula to calculate:

[0050] φ s = (285 - 180) / (620 - 180) ≈ 0.238 (i.e. 23.8%).

[0051] aquifer clay content Read the natural gamma curve value GR=75API for the same layer and query the pure sandstone layer in this borehole. = 45 API, pure mudstone section =150API. Substitute into the formula to calculate:

[0052] = (75 - 45) / (150 - 45) ≈ 0.286 (i.e. 28.6%).

[0053] Inflow verification calculation:

[0054] The dynamic flow response amplitude at the maximum drawdown (S3=9m) is taken as A=2.1 mV, and the instrument calibration coefficient is K=0.15 m / (s·mV). The flow velocity is calculated as: v = A·K = 2.1 * 0.15 = 0.315 m / s.

[0055] Given a borehole diameter of 150 mm and a cross-sectional area S = π*(0.15 / 2)² ≈ 0.0177 m².

[0056] Calculate the inflow rate: Q = v·S = 0.315 * 0.0177 ≈ 0.00557 m³ / s ≈ 20.05 m³ / h. This value can be used as a label for model training or as a reference for result verification.

[0057] III. Construction and Training of GRNN Prediction Model

[0058] Dataset preparation:

[0059] Data were collected from 10 boreholes in the study area that had completed pumping tests, including four key parameters for each borehole (H2, M, φ). s , ) and the measured inflow Q (such as Figure 4 Example). In this embodiment, the data from the ZK-1 well is used as the sample to be predicted and is not used in model training.

[0060] The four input parameters of the nine training holes were respectively... The data is standardized and mapped to the [0,1] interval. For example, for aquifer thickness M, its maximum value is... =12.5m, minimum value =1.5m, then the standardized value of a certain hole M=8.0m is (8.0-1.5) / (12.5-1.5) ≈ 0.591.

[0061] Model training:

[0062] The standardized 9 groups (H2, M, φ) s , The data is used as input features, and the corresponding 9 Q-reals are used as output labels to construct a training set.

[0063] A GRNN model was constructed using the `newgrnn` function in MATLAB, with the number of neurons in the pattern layer equal to the number of training samples (9). The optimization range for the smoothing factor (spread) was set to [0.1, 1.0], with a step size of 0.1.

[0064] Leave-one-out cross-validation was used to iterate through all smoothing factors. The average relative error of the model's predictions on the training set was calculated under different smoothing factors. The results showed that the average relative error of the training set was the smallest (5.2%) when the smoothing factor was 0.3. Therefore, the optimal smoothing factor was determined to be 0.3, and the model training was completed.

[0065] Model validation:

[0066] To initially verify the model's performance, two wells were randomly selected from the nine training wells as a temporary test set, and the remaining seven were used for training. After training the model using the optimal smoothing factor (0.3), predictions were made on the two test wells.

[0067] The relative errors between the predicted inflow Qpredicted and the measured value Qactual were calculated to be 4.8% and 6.1%, respectively, with an average of 5.45%, which is less than the threshold of 10%. This preliminarily indicates that the model performance is qualified and can be used for new well prediction.

[0068] IV. Aquifer Inflow Prediction (corresponding to claim 1)

[0069] The four parameters extracted from borehole ZK-1 to be predicted (H2=40.5m, M=8.0m, φ) s =23.8%, The value (28.6%) was standardized. The maximum and minimum values ​​used were consistent with those used in the training set.

[0070] The standardized vectors are then input into the trained GRNN prediction model.

[0071] The model output is the standardized predicted inflow rate. After inverse standardization, the final predicted inflow rate Q_predict = 15.8 m³ / h is obtained.

[0072] Comparative verification: The inflow rate Qtransmitted = 17.8 m³ / h calculated by traditional stratified pumping test combined with the Dubuis formula, and the actual observed stable inflow rate Qactual = 15.0 m³ / h.

[0073] The relative error predicted by the method of this invention is approximately 5.3% = |15.8 - 15.0| / 15.0 * 100%.

[0074] The relative error of the traditional method prediction is approximately 18.7% (|17.8 - 15.0| / 15.0 * 100%).

[0075] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for accurate prediction of aquifer inflow rate by integrating conventional and electromagnetic flow logging, characterized in that, Includes the following steps: Step 1: Well logging data acquisition. Electromagnetic flow logging and conventional logging are performed simultaneously in the target exploration borehole to obtain static and dynamic flow response amplitude curves, apparent resistivity curves, natural gamma curves, and sonic transit time data. Step 2: Aquifer Identification and Parameter Extraction. Based on the logging data obtained in Step 1, the location of the aquifer is identified and the aquifer thickness M and the location of the aquifer floor are calculated. H 2 Aquifer porosity φ s and aquifer clay content Four key hydrogeological parameters; Step 3: Construct a water inflow prediction model based on a generalized regression neural network (GRNN), using the aquifer thickness M and the bottom plate location as parameters. H 2 Porosity φ s mud content Using measured water inflow as the input feature and actual water flow as the output label, the GRNN model is trained and validated. Step 4: Aquifer inflow prediction. Input the four key parameters obtained from Step 1 and Step 2 of the borehole to be predicted into the validated GRNN model, and output the predicted aquifer inflow value of the borehole.

2. The method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging as described in claim 1, characterized in that, The selection criteria for the four key hydrogeological parameters mentioned in step 2 are as follows: aquifer thickness M represents the water storage space, aquifer floor location H2 represents the recharge force, and aquifer porosity... φ s Characterizing water storage capacity, aquifer clay content Characterizes water permeability resistance.

3. The method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging as described in claim 1, characterized in that, Step 3, constructing the GRNN prediction model, specifically includes: 3.1 Data preprocessing: Collect the four key parameters and corresponding measured water inflow from multiple boreholes in the target area, form a dataset, perform min-max standardization, and randomly divide it into training and test sets; 3.2 Model Training: The four parameters of the training set are used as input features, and the measured inflow rate is used as the output label. These are then input into the GRNN model for training. The smoothing factor of the model is optimized within a preset range using a grid search method. 3.3 Model Validation: Substitute the input features of the test set into the trained model and calculate the average relative error between the predicted inflow and the measured value. If the average relative error does not exceed 10%, the model validation is successful.

4. The method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging as described in claim 1, characterized in that, The aquifer thickness M mentioned in step 2 is the aquifer top depth determined by electromagnetic flow logging. H 1 and base plate depth H 2 Through formula Calculated.

5. The method according to claim 1, characterized in that, The porosity of the aquifer mentioned in step 2 φ s Based on the sonic transit time data obtained from conventional well logging, using the formula... Where Δt is the measured acoustic time difference, Sound wave time difference.

6. The method according to claim 1, characterized in that, The aquifer clay content mentioned in step 2 Based on the natural gamma ray GR data obtained from conventional well logging, using the formula... The result is given, where GR is the measured value. These are the minimum and maximum values ​​of the natural gamma for that layer, respectively.

7. The method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging as described in claim 1, characterized in that, The qualitative judgment of the aquifer in step 2 is based on the following: analyzing the static flow response amplitude curve obtained from electromagnetic flow logging. If the curve shows an increasing trend from bottom to top, it is determined to be a water-bearing layer; if it shows a decreasing trend, it is determined to be a water-absorbing layer; if there is no obvious change, it is determined to be an impermeable layer. The abrupt change in amplitude of the curve is the aquifer interface.

8. The method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging as described in claim 1, characterized in that, The electromagnetic flow logging described in step 1 includes: obtaining the static flow response amplitude curve by measuring under natural static conditions of the borehole, and obtaining the dynamic flow response amplitude curve by measuring under three pumping tests with different drawdowns.

9. The method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging as described in claim 1, characterized in that, Step 2 also includes the step of calculating the borehole water inflow based on the electromagnetic flow logging data through the "amplitude-velocity-flow rate" conversion system. Specifically, the water flow velocity v = A·K is calculated based on the flow response amplitude A and the instrument calibration coefficient K, and the water inflow Q = v·S is calculated based on the borehole cross-sectional area S.

10. The method for accurate prediction of aquifer inflow by integrating conventional and electromagnetic flow logging as described in claim 1, characterized in that, The method is applicable to the prediction of aquifer water inflow and the prevention of water hazards during the mining of coal and metal mineral resources.

Images