CN-121976784-A - Fault brine well repairing method for well and ore salt exploitation

CN121976784ACN 121976784 ACN121976784 ACN 121976784ACN-121976784-A

Abstract

The invention discloses a fault brine well restoration method used in well salt exploitation, which comprises the steps of (1) constructing a directional well at a proper position away from the fault brine well, ensuring the vertical direction as much as possible in the construction process of a straight well section of the directional well, (2) carrying out deflecting at the tail end of the straight well section after reaching a preset depth, keeping the deflecting construction track in smooth transition until the well oblique angle is horizontal, (3) controlling the horizontal displacement and azimuth angle, carrying out horizontal section construction, and keeping the well path smooth until the construction drilling track is accurately close to a target area and communicated with a solution cavity of the fault well, thereby forming a new exploitation channel. According to the invention, the mining channel is reestablished near the fault brine well so as to butt joint the fault brine well, and the mining rate of mine salt is improved.

Inventors

CHEN TAO
WANG LIN
Nie pan
DENG YU

Assignees

江西晶昊盐化有限公司

Dates

Publication Date: 20260505
Application Date: 20260213

Claims (9)

1. A method for repairing a faulty brine well in salt recovery of the well, comprising the steps of: (1) Constructing a directional well at a proper position away from the fault brine well, wherein the directional well is ensured to be vertical as much as possible in the construction process of a straight well section; (2) After reaching a preset depth, deflecting is carried out at the tail end of the straight well section, and the deflecting construction track keeps smooth transition until the well oblique angle is horizontal; (3) And controlling the horizontal displacement and azimuth angle, performing horizontal section construction, and keeping the well path smooth until the construction drilling track accurately approaches the target area and is communicated with the fault well dissolving cavity, so as to form a new exploitation channel.
2.A method of repairing a faulty brine well in salt recovery of a well according to claim 1 wherein the vertical interval is controlled at a well inclination of 3.5 °.
3. The method for repairing a faulty brine well in salt mining of a well according to claim 1, wherein after the beginning of the whipstock, the drilling track is predicted in the process of construction drilling, and the method comprises the steps of: s1, acquiring characteristic parameters in the drilling process, wherein the characteristic parameters are respectively a well inclination angle, an azimuth angle and a dogleg degree as target sequences, the characteristic sequences are comparison sequences, the gray scale association degree of each characteristic parameter and the target sequences is acquired, and the characteristic parameters affecting each target variable are determined; S2, performing stable wavelet transformation on characteristic parameters of each target variable in a specific sliding window to be decomposed into an approximate component and a detail component, extracting the approximate component as a trend component of the target variable; The method comprises the steps of S3, constructing a time convolution network prediction system, wherein the prediction system comprises 8 residual blocks, a stacked convolution block layer and a full connection layer, each residual block consists of a causal expansion convolution layer, weight normalization, a ReLU activation function and a space Dropout, obtaining corresponding characteristic values through causal expansion convolution mapping of input characteristics, aligning output time sequences and input time sequences in the causal expansion convolution layer, performing 0 filling on the left side of a sequence to ensure that future time steps are not introduced, performing weight normalization processing after causal expansion convolution, decomposing weight vectors into direction vectors and amplitude scalar quantities, optimizing respectively, introducing nonlinear factors through the ReLU activation function, reducing network complexity, finally preventing network overfitting through the space Dropout, if input and output dimensions of the residual blocks are consistent, directly adding the input to the output of the residual blocks, if the input and output dimensions are inconsistent, projecting the input to the same dimension as the output through 1X 1 convolution, adding, stacking the plurality of residual blocks, performing step-by-step relation, performing step-by-step prediction on the output of the stacked convolution block layer, and obtaining the final predicted value through the full connection layer.
4. A method for repairing a faulty brine well in salt recovery of a well according to claim 3, wherein the characteristic parameters are expanded into detail coefficients and approximation coefficients of different scales by a series of scale functions and wavelet functions in the stationary wavelet transform decomposition, and the expression of the approximation coefficients of the j+1th layer is: The expression of the detail coefficient of the j+1th layer is: wherein i is the order, j is the layer number, K is the index of the filter coefficient, K is the filter length, n is the index of the discrete time sampling point, and 2 j is the up-sampling multiple of the filter coefficient of the j-th layer; The approximation component of the first layer is selected as the trend component of the target variable.
5. A fault brine well restoration method for well and mineral salt exploitation according to claim 3 is characterized in that a Bayesian optimization algorithm is adopted to optimize the kernel size and expansion factor of a time convolution network prediction system in the prediction process, firstly, a prediction system is initialized, super-parameter candidate value points are randomly collected to construct an initial proxy model, an existing data training proxy model is used for establishing an acquisition function by a Gaussian regression model, the point of maximizing the acquisition function is selected, the function value of an evaluation point is obtained, the function value is added into a current evaluation point data set, and the optimized super-parameter value is output.
6. The method for repairing a faulty brine well in salt mining of a well according to claim 5, wherein the time convolution network prediction system under the optimized super-parameter value is trained by defining a root mean square error loss function, and performing min-max normalization processing on the historical data for training, wherein the formula is: Wherein the method comprises the steps of In order to normalize the processed data, As the original history data, the data is obtained, Is the maximum value in the original history data, Is the minimum value in the original history data; Dividing normalized data into a training set, a verification set and a test set according to a time window function, wherein the training set and the verification set are used for iterative training, the test set is used for evaluating effects, the training set is used for obtaining a predicted value through a time convolution network prediction system, the predicted value and a true value are compared to calculate a loss function, a gradient of loss relative to network structure model parameters is obtained, model weights are updated according to the gradient, after the training is completed, the verification set is used for evaluating effects, the training is repeated until the loss of the verification set is not reduced any more, a network structure model is saved for testing, and the prediction precision and generalization of the model are obtained by inputting the saved model through the test set.
7. The method for repairing a faulty brine well in salt mining according to claim 1, wherein the number ratio of training set, validation set and test set is training set: validation set: test set = 70:15:15.
8. A prediction system for a drilling track in a construction drilling process is characterized by comprising a data acquisition layer, a data transmission and processing layer, a time convolution network prediction system and an output module, wherein the data acquisition layer is used for acquiring characteristic parameters in the drilling process and sending acquired data to the data transmission and processing layer, the data transmission and processing layer comprises a database for storing historical data, stratum characteristic data and design drilling track data and a data processing module for processing and obtaining input characteristics or training data, the data processing module is used for transmitting the processed data to the time convolution network prediction system for prediction or training, a predicted output value is transmitted to the output module and a predicted result is displayed, the output module at least comprises a predicted result output module for displaying the predicted result and combining the predicted result with the historical track and the design track, and a device state parameter monitoring module for detecting and displaying the running state of a device in real time and guaranteeing normal operation of the construction drilling process.
9. The system for predicting the drilling track in the construction drilling process according to claim 8, wherein the data acquisition layer at least comprises a well inclination angle and azimuth angle detection module, a strong magnetic signal source serving as a magnetic beacon source, a magnetic signal receiving module, a near-bit vibration sensor, a near-bit torque sensor, a near-bit mechanical sensor, a temperature sensor, a magnetometer and a logging-while-drilling acoustic module, wherein the well inclination angle and azimuth angle detection module is used for detecting well inclination angles and azimuth angles of drilling positions in real time, the strong magnetic signal source is arranged in a fault brine well and is used as a magnetic beacon source, the magnetic signal receiving module is arranged on construction drilling equipment and is used for receiving signals of the magnetic beacon source, calculating relative distances and azimuth angles, the near-bit vibration sensor is used for detecting bit vibration conditions and evaluating deviation degrees of test parameters, the near-bit torque sensor is used for testing drilling pressure in real time, the near-bit mechanical sensor is used for measuring guiding forces, the temperature sensor is used for measuring temperature of drilling positions in real time, the magnetometer is used for sensing changes of magnetic field intensity and direction in drilling environments, judging whether received magnetic signals are interfered by ferromagnetic minerals, reliability of monitoring distance and azimuth angles is evaluated.

Description

Fault brine well repairing method for well and ore salt exploitation Technical Field The invention relates to the technical field of salt production processes, in particular to a fault brine well repairing method used in well and ore salt exploitation. Background In the well and ore salt, partial stone salt layer is produced in a layered or quasi-layered manner in the salt-containing section, basically maintains the original normal layer sequence relation with the top and bottom plates, and has normal layer position and production. The stone salt layer has gentle appearance, multiple layers of stone salt layers, dan Yanceng and clastic rock layers alternately appear to form prosodic inter-layers with equal thickness or different thickness, and the prosodic layers repeatedly appear in the salt-containing section to form a multi-layer structure of the stone salt deposit. In the existing well and salt exploitation, directional butt-joint drilling technology is often used for mineral exploitation, and the method greatly increases the volume of a solution cavity and improves the brine yield and the raw brine quality. Because of the drilling communication, the time for building the groove is shortened, the recovery rate of the development reserve is greatly improved, and the quality and the quantity of halogen water are greatly improved. But for the blockage of brine channels between convection wells caused by the collapse of clastic rock formations or other reasons, the problem of blockage cannot be solved in a better way. Disclosure of Invention To this end, the invention provides a method for repairing a faulty brine well in the exploitation of salt from the well, comprising the steps of: (1) Constructing a directional well at a proper position away from the fault brine well, wherein the directional well is ensured to be vertical as much as possible in the construction process of a straight well section; (2) After reaching a preset depth, deflecting is carried out at the tail end of the straight well section, and the deflecting construction track keeps smooth transition until the well oblique angle is horizontal; (3) And controlling the horizontal displacement and azimuth angle, performing horizontal section construction, and keeping the well path smooth until the construction drilling track accurately approaches the target area and is communicated with the fault well dissolving cavity, so as to form a new exploitation channel. Further, the vertical well section is controlled to have a well inclination less than or equal to 3.5 degrees. Further, after the deflecting is started, the drilling track is predicted in the construction drilling process, and the method comprises the following steps: s1, acquiring characteristic parameters in the drilling process, wherein the characteristic parameters are respectively a well inclination angle, an azimuth angle and a dogleg degree as target sequences, the characteristic sequences are comparison sequences, the gray scale association degree of each characteristic parameter and the target sequences is acquired, and the characteristic parameters affecting each target variable are determined; S2, performing stable wavelet transformation on characteristic parameters of each target variable in a specific sliding window to be decomposed into an approximate component and a detail component, extracting the approximate component as a trend component of the target variable; The method comprises the steps of S3, constructing a time convolution network prediction system, wherein the prediction system comprises 8 residual blocks, a stacked convolution block layer and a full connection layer, each residual block consists of a causal expansion convolution layer, weight normalization, a ReLU activation function and a space Dropout, obtaining corresponding characteristic values through causal expansion convolution mapping of input characteristics, aligning output time sequences and input time sequences in the causal expansion convolution layer, performing 0 filling on the left side of a sequence to ensure that future time steps are not introduced, performing weight normalization processing after causal expansion convolution, decomposing weight vectors into direction vectors and amplitude scalar quantities, optimizing respectively, introducing nonlinear factors through the ReLU activation function, reducing network complexity, finally preventing network overfitting through the space Dropout, if input and output dimensions of the residual blocks are consistent, directly adding the input to the output of the residual blocks, if the input and output dimensions are inconsistent, projecting the input to the same dimension as the output through 1X 1 convolution, adding, stacking the plurality of residual blocks, performing step-by-step relation, performing step-by-step prediction on the output of the stacked convolution block layer, and obtaining the final predicted value through the full connec