A high-precision mine water recharge dynamic monitoring system and method

Abstract

The application provides a high-precision mine water recharge dynamic monitoring system and method, which comprises a distributed acoustic data acquisition subsystem, the distributed acoustic data acquisition subsystem is connected with one end of an armored optical cable, the armored optical cable is sent into a recharge drill hole through an armored optical cable laying device, the recharge drill hole is also connected with a mine water recharge subsystem to perform mine water recharge; an acoustic signal processing and feature extraction subsystem comprises a wavelet multi-scale decomposition module, a target frequency band extraction module, a sub-band energy calculation module, a non-target frequency band suppression module and a signal reconstruction module; an intelligent injection channel identification and flow estimation subsystem comprises an injection channel identification module, a flow distribution calculation module and an advantage channel intelligent identification module. The application has full wellbore, high spatiotemporal resolution dynamic monitoring capability, injection profile dynamic analysis and advantage channel intelligent identification capability, high reliability and long service life capability in extreme environment, and recharge process safety and optimization decision support capability.

Description

Technical Field

[0001] This invention belongs to the field of coal mining technology and relates to dynamic monitoring of mine water reinjection, specifically to a high-precision dynamic monitoring system and method for mine water reinjection. Background Technology

[0002] Deep well recharge technology for mine water has been widely used as an effective means of "transfer and storage." However, the effective implementation and operational optimization of this technology urgently require high-precision dynamic monitoring of the recharge process. Currently, mainstream monitoring methods mainly rely on hydrogeological models and ground-based point monitoring methods, which have the following significant limitations:

[0003] First, there is a lack of model dependence and data support. The reinjection process often uses the Darcy unsteady flow equation to describe hydrodynamic behavior and combines it with convection-dispersion equations to simulate solute transport (e.g., Zhao et al., 2021). The accuracy of the model is highly dependent on the accuracy of key parameters such as permeability coefficient, reservoir thickness, and reinjection pressure, requiring calibration using a large amount of field measurement data. Existing monitoring methods cannot provide sufficient, real-time, and high-resolution downhole data, leading to significant uncertainties in model predictions and failing to provide a reliable basis for reinjection optimization.

[0004] Second, ground-based point monitoring has inherent limitations. Firstly, it suffers from low spatial resolution: pressure and flow sensors installed at the wellhead can only acquire single-point data, failing to reflect the water pressure and flow distribution along the depth direction within the reinjection zone, making it difficult to identify dominant seepage channels or blockage risks. Secondly, it exhibits poor environmental adaptability and reliability: deep well reinjection environments often possess high pressure (6–10 MPa) and high salinity characteristics (such as the Cl aquifer of the Liujiagou Formation in the western mining area). - With concentrations up to 40,000 mg / L, traditional electronic sensors are susceptible to chemical corrosion, scaling, and electromagnetic interference, resulting in high failure rates, short lifespans, and difficulty in continuously and stably acquiring effective data. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a high-precision dynamic monitoring system and method for mine water reinjection, thereby solving the technical problem that the accuracy of dynamic monitoring of mine water reinjection in existing technologies needs to be further improved.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A high-precision dynamic monitoring system for mine water reinjection includes a distributed acoustic data acquisition subsystem, which is connected to one end of an armored optical cable. The armored optical cable is sent into the reinjection borehole through an armored optical cable deployment device. The reinjection borehole is also connected to the mine water reinjection subsystem for mine water reinjection.

[0007] The distributed acoustic data acquisition subsystem includes a laser, an optical fiber coupler, an acousto-optic modulator, an amplifier, and a circulator connected in sequence. The circulator is connected to one end of an armored optical cable. The circulator is also connected to a polarization diversity receiver, which is connected to an analog-to-digital converter circuit via a balanced photodetector and another balanced photodetector. The analog-to-digital converter circuit is connected to the input of a field-programmable gate array (FPGA), which is connected to a server. The output of the FPGA is connected to a digital-to-analog converter circuit, which is connected to the input of an RF amplifier. The output of the RF amplifier is connected to the acousto-optic modulator. The optical fiber coupler is also connected to the polarization diversity receiver.

[0008] The present invention also has the following technical features: The server has a built-in acoustic signal processing and feature extraction subsystem as well as an intelligent injection channel identification and flow estimation subsystem.

[0009] The acoustic signal processing and feature extraction subsystem includes a wavelet multi-scale decomposition module, a target frequency band extraction module, a sub-band energy calculation module, a non-target frequency band suppression module, and a signal reconstruction module.

[0010] The intelligent injection channel identification and flow estimation subsystem includes an injection channel identification module, a flow distribution calculation module, and a dominant channel intelligent identification module.

[0011] The advantageous channel intelligent recognition module includes a multi-dimensional feature extraction submodule, a LightGBM intelligent classification submodule, an adaptive optimization submodule, and a result evaluation and output submodule.

[0012] The wavelet multi-scale decomposition module is used to decompose the input signal. Perform discrete wavelet transform to decompose into J Layered zone.

[0013] The target frequency band extraction module is used to extract the frequency band of groundwater flow characteristics.

[0014] The sub-band energy calculation module is used to calculate the total energy of each scale factor.

[0015] The non-target frequency band suppression module is used to set the non-target frequency band coefficients to zero.

[0016] The aforementioned signal reconstruction module is used to reconstruct a signal that retains only the energy characteristics of the target frequency band. .

[0017] The injection channel identification method of the injection channel identification module is as follows: when the conditions are met... , When the number of consecutive depth points is ≥3, it is determined to be a valid injection channel; in: Indicates frequency band energy; Indicates a time window; Indicates the drilling depth; Indicates an empirical threshold; This represents the average energy of the background noise; The reconstructed signal representing the target frequency band energy characteristics.

[0018] The method for calculating the flow distribution of each reinjection channel in the flow distribution calculation module is as follows: ; ; In the formula: Indicates injection of the first The flow rate of each recharge channel; Indicates the sequence number of the reinjection channel; Indicates the total injection volume; Indicates the first Flow weight of each recharge channel; Indicates the first The injection length of each recharge channel; P _ i Indicates the first Average energy of each reinjection channel; Indicates the first The total length of each test segment.

[0019] Indicates the first Total energy of each test segment; This indicates the sequence number of the test segment.

[0020] The multi-dimensional feature extraction submodule is used to automatically extract quantitative feature vectors for model training from the preprocessed acoustic signal and flow data. .

[0021] The LightGBM intelligent classification submodule is a multi-classification model built using the LightGBM framework. The input to the model is a quantitative feature vector. The model outputs a probability prediction of the channel type.

[0022] The adaptive optimization submodule integrates a Bayesian optimizer for automatically searching the adjustable coefficients in the LightGBM intelligent classification submodule. The optimal threshold coefficient .

[0023] The result evaluation and output submodule is used to receive the prediction results from the classification module and output structured data including channel type, prediction confidence and feature importance.

[0024] This invention also protects a high-precision dynamic monitoring method for mine water reinjection, which employs the high-precision dynamic monitoring system for mine water reinjection described above.

[0025] The method includes the following steps: Step 1: Prepare the drill holes and casing for water injection.

[0026] Step 2: System initialization and parameter preset.

[0027] Step 3, Armored Optical Cable Laying and Depth Calibration: Step 301: Secure the armored optical cable at the wellhead and connect it to the distributed acoustic data acquisition subsystem.

[0028] Step 302: Start the distributed acoustic wave data acquisition subsystem and perform a device self-test.

[0029] Step 303: Lowering the armored optical cable.

[0030] Step 304: Dynamic sealing of the armored optical cable.

[0031] Step 305: Connect the mine water reinjection subsystem.

[0032] Step 4: High-precision monitoring of the water injection process: Step 401: Start-up and parameter control of the mine water reinjection subsystem.

[0033] Step 402: High-precision monitoring is performed using a distributed acoustic wave data acquisition subsystem.

[0034] Step 5, Acoustic signal processing and feature extraction: A sound wave signal processing and feature extraction subsystem is used for sound wave signal processing and feature extraction.

[0035] Step Six: Intelligent Injection Channel Identification and Flow Estimation: Step 601, Injection Channel Identification: Injection channel identification is performed using an injection channel identification module.

[0036] Step 602, Flow distribution calculation: The flow distribution calculation module is used to calculate the flow distribution.

[0037] Step 603, Intelligent Identification of Advantageous Channels: The dominant channel is identified using an intelligent identification module.

[0038] Compared with the prior art, the present invention has the following technical effects: (I) This invention possesses full-wellbore, high spatiotemporal resolution dynamic monitoring capabilities: breaking through the spatial limitations of traditional point-based monitoring, it achieves continuous distributed monitoring across the entire depth range along the reinjection wellbore, with a spatial resolution reaching sub-meter levels. This completely overcomes the shortcomings of traditional methods in obtaining detailed hydrodynamic information within the formation, thus improving monitoring accuracy. The system supports real-time data acquisition at high sampling frequencies of up to 10kHz and above, accurately capturing the high-frequency acoustic / vibration signals induced by the flow of injected water in the target formation and their transient changes, truly achieving full real-time, dynamic monitoring of the reinjection process. Based on the characteristics of high-resolution acoustic signals, it can accurately identify effective reinjection formations and highly sensitively sense the hydrodynamic response generated by the flow of injected water in porous / fractured media.

[0039] (II) This invention possesses dynamic analysis capabilities for injection profiles and intelligent identification of dominant channels: By introducing advanced signal processing and energy-length weighted algorithms, it achieves quantitative analysis of the spatial distribution of relative flow rate / velocity of reinjection fluid along the well depth in the target layer, providing direct data support for reinjection efficiency assessment, dominant channel identification, and blockage area location. The accompanying data processing and analysis software integrates signal denoising, feature extraction, energy-length weighted flow rate calculation, and an adaptive threshold identification algorithm based on flow statistics. It successfully plots the relative flow rate / velocity distribution profiles along the depth direction of the open hole section (Ordovician limestone aquifer) at each stage of the water injection test, clearly revealing the non-uniform migration characteristics of the injected water in the reservoir. The proposed adaptive threshold algorithm dynamically sets the judgment threshold based on the mean and standard deviation of the flow rate distribution, achieving intelligent identification of dominant channels, significantly improving the accuracy and engineering applicability of the judgment, and effectively locking in dominant reinjection channels within specific depth ranges.

[0040] (III) This invention possesses high reliability and long lifespan under extreme environments: the system uses passive optical fiber as the core sensing medium, fundamentally overcoming the limitations of traditional electronic sensors under high pressure (6-10 MPa) and high mineralization (such as Cl) conditions. - Despite the corrosion, scaling, and failure issues that easily occur in environments with concentrations as high as 40,000 mg / L and strong electromagnetic interference, this device exhibits excellent environmental adaptability and long-term operational stability, enabling it to continuously and reliably acquire effective monitoring data under harsh conditions of deep well reinjection.

[0041] (IV) This invention provides safety and optimization decision support capabilities for the reinjection process: Through real-time analysis of acoustic signal characteristics, the system can promptly identify abnormal situations such as precursors to blockage, signs of crossflow, and micro-rupture events, enabling early warning and providing a critical time window for proactive intervention and reinjection safety. The high-precision, high spatiotemporal resolution dynamic monitoring data provided (including stratum effectiveness, flow distribution, dominant channel identification, and abnormal signals) provides a direct and reliable basis for optimizing injection parameters, evaluating reinjection effects, and improving reinjection efficiency.

[0042] (V) This invention is particularly applicable to deep well reinjection scenarios of high-salinity mine water in arid mining areas of western my country, providing a highly innovative and reliable core monitoring technology for mine water "transfer and storage," green mining, and groundwater ecological environment protection. The principles and system architecture of this technology can also be extended to other underground fluid transport monitoring fields such as CO2 geological sequestration injection monitoring, geothermal reinjection / mining well monitoring, and aquifer energy storage process monitoring, demonstrating broad engineering application prospects and significant scientific value. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the mine water deep well reinjection monitoring system of the present invention.

[0044] Figure 2 This is a schematic diagram of the distributed acoustic wave data acquisition subsystem.

[0045] Figure 3 This is a schematic diagram of the reinjection dynamic monitoring subsystem.

[0046] Figure 4 This is a panoramic waterfall diagram showing the energy density during the water injection process. The middle part between the two vertical white lines represents the injection stage.

[0047] Figure 5 The water injection channel identification results show that the yellow highlighted part between the two horizontal white lines is the injection channel.

[0048] The meanings of the labels in the diagram are as follows: 1-Distributed acoustic data acquisition subsystem, 2-Armored optical cable, 3-Armored optical cable laying device, 4-Recharge borehole, 5-Mine water recharge subsystem, 6-Recharge dynamic monitoring subsystem.

[0049] 101-Laser, 102-Fiber Optic Coupler, 103-Acousto-optic Modulator, 104-Amplifier, 105-Circulator, 106-Polarization Diversity Receiver, 107-Balanced Photodetector, 108-Balanced Photodetector, 109-Analog-to-Digital Conversion Circuit, 110-Field Programmable Gate Array, 111-Server, 112-Digital-to-Analog Conversion Circuit, 113-RF Amplifier.

[0050] 301-Logging winch, 302-Ground fixed pulley, 303-Aerial fixed pulley, 304-Fiber optic cable counterweight, 305-Support. 401 - Deep well reinjection borehole, 402 - Technical casing, 403 - Surface casing, 404 - Orifice connector, 405 - Orifice fiber optic sealing port, 406 - Orifice water injection sealing connector.

[0051] 501-Mine wastewater treatment pond, 502-Recharge pipeline, 503-Industrial water injection pump, 504-Water injection valve.

[0052] 601-Electromagnetic flowmeter, 602-Pressure sensor, 603-Multi-channel USB data acquisition card, 604-Recharge power parameter monitoring master station, 605-Power module.

[0053] The specific content of the present invention will be further explained in detail below with reference to the embodiments. Detailed Implementation

[0054] It should be noted that, unless otherwise specified, all devices, circuits and algorithms in this invention are based on devices, circuits and algorithms known in the prior art.

[0055] Distributed acoustic sensing (DAS) technology, as an emerging monitoring method, uses optical fiber as a continuous sensing medium. By analyzing the phase changes of Rayleigh backscattered light, it can sense acoustic / vibration signals distributed along the optical fiber. This technology has advantages such as being fully distributed, resistant to electromagnetic interference, corrosion-resistant, and intrinsically safe, and has achieved good application results in fields such as oil and gas exploration and seismic monitoring.

[0056] Despite the significant advantages of DAS technology, there are currently no publicly reported applications of its use in dynamic monitoring of deep well reinjection processes in mines. During reinjection, physical mechanisms such as the seepage of injected water in porous / fractured media, its interaction with the formation, and potential microseismic events all excite characteristic acoustic / vibration signals. How to effectively capture and analyze these signals using DAS technology to achieve effective identification and confirmation of reinjection layers, dynamic monitoring of reinjection fluid velocity / flow rate along well depth, and early warning of abnormal events such as blockage, crossflow, and microfractures remains a significant technical challenge.

[0057] Especially in the harsh reinjection environment of high mineralization and high pressure in the western mining areas, developing a high-precision and high-reliability dynamic monitoring system and method based on DAS technology is of great practical significance for ensuring the safety of deep well reinjection operations, evaluating the sealing effect, optimizing reinjection strategies, and improving the utilization efficiency of mine water resources. It is also a key technical problem that urgently needs to be solved.

[0058] The mine water deep well reinjection monitoring system of this invention can also be called a mine water deep well reinjection monitoring system based on distributed acoustic wave sensing. The core innovation of this invention lies in the first application of fully distributed DAS fiber optic sensing technology to the field of mine water reinjection monitoring. By using armored optical cables deployed in the reinjection well as a continuous, full-well-section acoustic detection medium, and combining this with dedicated signal processing and feature interpretation algorithms, this invention can achieve: real-time, continuous capture of acoustic / vibration signals along the well depth direction; accurate identification of effective reinjection sections; dynamic monitoring of the spatial distribution of flow rate / velocity of the reinjection fluid in the target reservoir; and early identification and warning of abnormal events such as blockages and crossflows during the reinjection process.

[0059] The present invention provides a high-resolution, high-reliability and full-coverage real-time dynamic monitoring technology for the mine water "transfer and storage" process, which significantly improves the safety, operational efficiency and long-term reinjection effect evaluation capabilities of reinjection operations.

[0060] The following are specific embodiments of the present invention. It should be noted that the present invention is not limited to the following specific embodiments. All equivalent modifications made based on the technical solutions of this application fall within the protection scope of the present invention.

[0061] Example 1: This embodiment provides a high-precision dynamic monitoring system for mine water reinjection, such as... Figure 1 As shown, it includes a distributed acoustic data acquisition subsystem 1, which is connected to one end of an armored optical cable 2. The armored optical cable 2 is sent into the reinjection borehole 4 through an armored optical cable laying device 3. The reinjection borehole 4 is also connected to a mine water reinjection subsystem 5 for mine water reinjection.

[0062] like Figure 2 As shown, the distributed acoustic wave data acquisition subsystem 1 includes a laser 101, an optical fiber coupler 102, an acousto-optic modulator 103, an amplifier 104, and a circulator 105 connected in sequence. The circulator 105 is connected to one end of the armored optical cable 2. The circulator 105 is also connected to a polarization diversity receiver 106. The polarization diversity receiver 106 is connected to an analog-to-digital converter 109 via balanced photodetectors 107 and 108, respectively. The analog-to-digital converter 109 is connected to the input of a field-programmable gate array 110, which is connected to a server 111. The output of the field-programmable gate array 110 is connected to a digital-to-analog converter 112, which is connected to the input of an RF amplifier 113. The output of the RF amplifier 113 is connected to the acousto-optic modulator 103. The optical fiber coupler 102 is also connected to the polarization diversity receiver 106.

[0063] In this embodiment, the distributed acoustic data acquisition subsystem 1 is located at the ground control center and the entire borehole. Its core functions include generating and emitting high-intensity pulsed laser light into the armored optical cable, receiving and demodulating the Rayleigh backscattered light signal returned from the armored optical cable, and converting the phase change information in the optical signal into a digital electrical signal (DAS data) distributed along the optical fiber that corresponds to the acoustic wave / vibration. Key performance parameters include: spatial sampling interval ≤ 1 meter (preferably sub-meter level, such as 0.2m or 0.5m); sampling frequency ≥ 0.1kHz (preferred according to the length of the monitoring target segment, preferably 4kHz or 10kHz for short target segments and 0.1kHz or 1kHz for long target segments), to meet the requirements for capturing transient hydrodynamic responses.

[0064] In this embodiment, the optical signal transmission path in the distributed acoustic wave data acquisition subsystem 1 is as follows: laser → coupler → acousto-optic modulator → fiber optic amplifier → circulator → armored optical cable; the backscattered signal passes through the circulator → polarization diversity receiver → photodetector, and finally outputs an electrical signal. Specifically, this includes the following two processes.

[0065] First, optical pulse generation: a narrow-linewidth laser generates a highly stable laser source; an acousto-optic modulator modulates the continuous laser into a swept-frequency pulse signal under the precise timing control of the FPGA; an erbium-doped fiber amplifier amplifies the optical pulse power to over 100mW, which is then injected into the armored optical cable via a circulator.

[0066] Second, optical signal demodulation: After the backscattered Rayleigh light is output through the port of the fiber optic circulator, it is connected to the polarization diversity receiver unit to suppress the polarization fading effect; the optical signal output by this unit is fed into the balanced photodetector to complete the high-sensitivity conversion of optical signal to electrical signal and differential output.

[0067] In this embodiment, the analog-to-digital conversion circuit 109 employs an analog-to-digital conversion circuit known in the art. The digital-to-analog conversion circuit 112 employs a digital-to-analog conversion circuit known in the art.

[0068] In this embodiment, the field-programmable gate array 110 can achieve timing control with nanosecond-level precision; the 24-bit ADC completes the analog-to-digital conversion of the electrical signal output by the optical signal demodulation module, and after passing through the Σ-Δ modulator, the output rate can reach 50kHz for time-space-intensity acoustic wave data.

[0069] In this embodiment, the armored optical cable 2 is a special armored single-mode optical cable that is resistant to high pressure, corrosion and has high tensile strength, such as stainless steel tube armored single-mode optical fiber, to ensure long-term stable operation in deep well high pressure and high mineralization fluid, and to effectively transmit hydrodynamic acoustic / vibration signals in the borehole.

[0070] As one specific solution in this embodiment, such as Figure 1As shown, the armored optical cable laying device 3 includes a logging winch 301, a ground fixed pulley 302, an aerial fixed pulley 303, and an optical cable counterweight 304. The aerial fixed pulley 303 is mounted on a support 305 and is located above the reinjection borehole 4. The armored optical cable 2 is wound around the logging winch 301, and the other end of the armored optical cable 2 passes around the ground fixed pulley 302 and the aerial fixed pulley 303 in sequence. The other end of the armored optical cable 2 extends into the reinjection borehole 4, and the other end of the armored optical cable 2 is connected to the optical cable counterweight 304.

[0071] In this embodiment, the logging winch 301 is responsible for monitoring and controlling the lowering depth, speed and tension of the armored optical cable, and has mechanical drive, drilling depth measurement, optical fiber stress measurement, armored optical cable lowering speed measurement and human-machine interaction modules.

[0072] As one specific solution in this embodiment, such as Figure 1 As shown, the reinjection borehole 4 includes a deep well reinjection borehole 401, a technical casing 402 is fitted inside the deep well reinjection borehole 401, a surface casing 403 is fitted outside the technical casing 402, a borehole connector 404 is installed on the top of the technical casing 402, and the borehole connector 404 is provided with a borehole fiber optic sealing port 405 and a borehole water injection sealing connector 406; the other end of the armored optical cable 2 passes through the borehole fiber optic sealing port 405 and enters the deep well reinjection borehole 401.

[0073] In this embodiment, the reinjection borehole 4 provides a stable physical path for the high-pressure reinjection water to flow to the target formation. The borehole connector 404, a key development component of this project, is used to reliably connect the high-pressure water injection pipeline to the casing, ensuring leak-free reinjection. Simultaneously, a dedicated sealing port allows the monitoring armored optical fiber to be introduced into the casing, enabling distributed acoustic sensing throughout the borehole. The borehole fiber optic sealing port 405 is constructed using a graphite packing compression seal. The borehole fiber optic sealing port employs a "segmented graphite packing compression seal" technology to prevent the armored optical cable from being cut or deformed, ensuring the accuracy of the monitoring data.

[0074] As one specific solution in this embodiment, such as Figure 1 As shown, the mine water reinjection subsystem 5 includes a mine wastewater treatment tank 501, which is connected to one end of the reinjection pipeline 502. The other end of the reinjection pipeline 502 is detachably connected to the water injection sealing connector 406 at the orifice of the reinjection borehole 4. An industrial water injection pump 503 and a water injection valve 504 are installed on the reinjection pipeline 502.

[0075] In this embodiment, the mine water reinjection subsystem 5 is located at the surface wellhead and is used to inject mine wastewater into the formation. The mine wastewater source in the mine wastewater treatment tank 501 is mine wastewater that has undergone sedimentation and filtration. The other end of the reinjection pipe 502 is detachably connected to the orifice water injection sealing connector 406 via threads. The water injection valve 504 adjusts its opening degree according to real-time monitored pressure and flow data to optimize the injection profile. Through the efficient cooperation of its components, the mine water reinjection subsystem 5 ensures safe mine water reinjection while protecting formation stability and equipment lifespan.

[0076] As one specific solution in this embodiment, such as Figure 1 As shown, it also includes a reinjection dynamic monitoring subsystem 6, such as Figure 3 As shown, the reinjection power monitoring subsystem 6 includes an electromagnetic flowmeter 601 and a pressure sensor 602. The electromagnetic flowmeter 601 and the pressure sensor 602 are connected to a multi-channel USB data acquisition card 603. The multi-channel USB data acquisition card 603 is connected to the reinjection power parameter monitoring master station 604. The electromagnetic flowmeter 601, the pressure sensor 602, and the multi-channel USB data acquisition card 603 are respectively connected to the power module 605.

[0077] In this embodiment, the multi-channel USB data acquisition card 603, the reinjection dynamic parameter monitoring master station 604, and the power module 605 all adopt multi-channel USB data acquisition cards, reinjection dynamic parameter monitoring master stations, and power modules known in the art. The reinjection dynamic monitoring subsystem 6 is used to measure the total reinjection flow rate and wellhead reinjection pressure parameters.

[0078] As a further embodiment, it also includes a sound wave signal processing and feature extraction subsystem and an intelligent injection channel identification and flow estimation subsystem built into the server 111.

[0079] As a preferred embodiment, the acoustic signal processing and feature extraction subsystem includes a wavelet multi-scale decomposition module, a target frequency band extraction module, a sub-band energy calculation module, a non-target frequency band suppression module, and a signal reconstruction module.

[0080] First, the wavelet multi-scale decomposition module: The wavelet multi-scale decomposition module is used to decompose the input signal. (Time-domain acoustic signal, sampling rate fs = 2000Hz) Perform discrete wavelet transform ( DWT ), decomposed into J The sub-strips are 6 layers in this embodiment.

[0081] The formula for wavelet multi-scale decomposition is: ; In the formula: Indicates the input signal; Indicates the sequence number of the sub-band; Indicates the number of sub-bands; The sequence number representing the detailed complex coefficients; Indicates a time window; These represent the detailed complex coefficients, characterizing high-frequency noise and signal; Represents the approximation coefficient, characterizing the low-frequency background noise at the coarsest scale; The wavelet function is used (the db4 wavelet basis is selected in this embodiment); It is a scaling function.

[0082] In this embodiment, the output structure of wavelet multi-scale decomposition is represented as follows: coeffs=[cA6,cD6,cD5,cD4,cD3,cD2,cD1]; In the formula: cA6 represents the approximation coefficient for scale 6 (0-15.6Hz). cD6~cD1 represent the detail complex coefficients (15.6-1000Hz) at scales 6 to 1.

[0083] Second, the target frequency band extraction module: The target frequency band extraction module is used to extract the frequency bands of groundwater flow characteristics.

[0084] In this embodiment, based on field tests and indoor simulation experiments, the characteristic energy of groundwater flow is concentrated in the range of 20-200Hz. Therefore, scale 3-4 (scale 4: 31.2-62.5Hz, scale 3: 62.5-125Hz) is generally selected as the target frequency band.

[0085] Third, the sub-band energy calculation module: The sub-band energy calculation module is used to calculate the total energy for each scale factor.

[0086] Total energy of scale coefficients ; In the formula: Indicates the first Total energy of the scale factor; Sequence number indicating scale; The sequence number representing the detailed complex coefficients; Indicates the first The first scale Each detailed complex coefficient; For the first The length of the scaling factor.

[0087] Fourth, non-target frequency band suppression module: The non-target frequency band suppression module is used to set the coefficients of non-target frequency bands (low-frequency noise at scale 6-5 and high-frequency noise at scale 2-1) to zero.

[0088] Fifth, signal reconstruction module: The signal reconstruction module is used to reconstruct the signal by retaining only the energy characteristics of the target frequency band. .

[0089] Reconstructed signal of target frequency band energy characteristics The calculation formula is: ; In the formula: Indicates a time window; cD5 and cD4 represent the detail complex coefficients at scales 5 to 6.

[0090] As a preferred embodiment, the intelligent injection channel identification and flow estimation subsystem includes an injection channel identification module, a flow distribution calculation module, and a dominant channel intelligent identification module.

[0091] First, the injection channel identification module: Injection channel identification is based on the correlation band energy analysis of acoustic signals (31.2-125Hz band in this embodiment). The injection channel identification method of the injection channel identification module is as follows: when the following conditions are met... , When the number of consecutive depth points is ≥3 (in this embodiment, the sampling point interval is 1m, and the number of depth points corresponds to a drilling depth of 3m), it is determined to be an effective injection channel; in: Indicates frequency band energy; Indicates a time window; Indicates the drilling depth; This represents the empirical threshold (in this embodiment, =1.5, which can be optimized through on-site calibration according to different geological conditions of the actual project. This represents the average energy of the background noise; The reconstructed signal representing the target frequency band energy characteristics.

[0092] Second, the flow distribution calculation module: The method for calculating the flow distribution of each reinjection channel in the flow distribution calculation module is as follows: ; ; In the formula: Indicates injection of the first The flow rate of each recharge channel; Indicates the sequence number of the reinjection channel; Indicates the total injection volume; Indicates the first Flow weight of each recharge channel; Indicates the first The injection length of each recharge channel; P _ i Indicates the first Average energy of each reinjection channel; Indicates the first The total length of each test segment; Indicates the first Total energy of each test segment; This indicates the sequence number of the test segment.

[0093] In this embodiment, The product takes into account both the size and activity level of the recharge channel.

[0094] Third, the superior channel intelligent recognition module: The dominant channel intelligent identification is the core analysis unit of the monitoring system, which uses machine learning algorithms to achieve accurate identification, classification, and dynamic optimization of the injected channels. The dominant channel intelligent identification module includes a multi-dimensional feature extraction submodule, a LightGBM intelligent classification submodule, an adaptive optimization submodule, and a result evaluation and output submodule.

[0095] Specifically, the multi-dimensional feature extraction submodule is used to automatically extract quantitative feature vectors for model training from the preprocessed acoustic signals and flow data. .

[0096] Quantitative feature vectors ; In the formula: Indicates the first The absolute flow rate of each recharge channel; Indicates the first The relative flow rate percentage of each recharge channel; This represents the standardized score of the traffic. Indicates the first Energy density of each reinjection channel; Indicates the first The length of the injection section of each recharge channel; Indicates the first The burial depth characteristics of each recharge channel; Indicates the first Flow fluctuation characteristics of each recharge channel ; Indicates the first The standard deviation of flow rate in each recharge channel within the time window; Indicates the first The arithmetic mean of the flow rate of each recharge channel within a time window.

[0097] Specifically, the LightGBM intelligent classification submodule is a multi-classification model built using the LightGBM framework, and the input to the model is a quantitative feature vector. The model outputs a probability prediction of the channel type.

[0098] The other end of the LightGBM intelligent classification submodule sequentially passes around the ground-mounted pulley, encompassing three predefined channels: Category 0: Non-dominant channel, i.e. < ; Category 1: Secondary dominant channel, i.e. ≤ ≤ T ; Category 2: Advantageous Channel, i.e. > T ; in: Indicates the judgment threshold. ; This represents the adjustable coefficient.

[0099] In this embodiment, the training objective of the LightGBM intelligent classification submodule is to minimize the weighted multi-class log loss function in order to solve the sample imbalance problem.

[0100] Specifically, the adaptive optimization submodule integrates a Bayesian optimizer to automatically search for adjustable coefficients in the LightGBM intelligent classification submodule. The optimal threshold coefficient Its optimization objective is to maximize the macro-average F1 score while controlling the prediction variance. Simultaneously, this module possesses an incremental learning function, which maintains the model's timeliness by periodically updating model parameters using newly acquired water injection data, ensuring the system adapts to dynamic changes in formation conditions. For example, model parameters are updated based on newly acquired water injection data at regular intervals (e.g., monthly) to ensure adaptation to changes in formation conditions.

[0101] In this embodiment, the optimal threshold parameter is automatically learned through Bayesian optimization. :

[0102] In the formula: This represents the macro average F1 score; Indicates the variance of the prediction results; Indicates the adjustable coefficient; This represents the regularization coefficient.

[0103] Specifically, the results evaluation and output submodule receives the prediction results from the classification module and outputs structured data including channel type, prediction confidence, and feature importance. For cases with confidence scores below a set threshold (e.g., 0.7), this module automatically triggers a flag, requesting manual review. Furthermore, the module integrates strategies such as early stopping, cross-validation, and SHAP (SHapley Additive exPlanations) value analysis to ensure the model's generalization ability and interpretability.

[0104] Example 2: This embodiment presents a high-precision dynamic monitoring method for mine water reinjection. To verify the actual monitoring performance of this invention, a field application test was conducted in August 2024 at the OM3 deep reinjection well in a typical mining area of ​​the Ordos Basin. Throughout the test, the high-precision dynamic monitoring system based on distributed acoustic sensing (DAS) proposed in this invention was deployed to collect and analyze in real time the acoustic / vibration signals induced by fluid transport during the water injection process. This system possesses sub-meter spatial resolution and kHz-level temporal sampling capability, enabling it to comprehensively capture the dynamic responses related to the reinjection process within the wellbore.

[0105] This method employs the high-precision mine water reinjection dynamic monitoring system described in Example 1. After preparing the reinjection well and casing, the method includes the following steps: Step 1: Preparation of drill holes and casings for water injection: Drilling and surface casing: The borehole diameter is Φ377mm. After drilling through the Quaternary alluvial layer and entering the intact bedrock for 5m, a Φ325mm×8mm surface casing (inner diameter Φ309mm) is installed.

[0106] Technical casing: Drill to 5m below the top of the target Ordovician limestone (Ordovician limestone) aquifer with a diameter reduced to Φ273mm, and run in a Φ219mm×6mm technical casing (inner diameter Φ207mm) to cement and isolate the upper non-target layer.

[0107] Open-hole section: A Φ153mm drill bit was used to drill through the cement plug at the casing shoe, and a Φ113mm drill bit was used to drill an open-hole section, ultimately reaching approximately 150m into the Ordovician limestone aquifer. This open-hole section is the target recharge layer section for this experiment. Figure 4 As shown.

[0108] Step 2, System Initialization and Parameter Presetting: Start the distributed acoustic wave data acquisition subsystem 1 to perform equipment self-test and calibration.

[0109] Configure core acquisition parameters: spatial sampling interval is 1m, sampling frequency is 2kHz to meet the requirements of transient sound wave capture, and single file storage time is 6s to balance data continuity and storage efficiency.

[0110] Synchronously configure the reinjection dynamic monitoring subsystem 6 (wellhead pressure and flow sensor) and set the acquisition frequency (≥1Hz).

[0111] Step 3, Armored Optical Cable Laying and Depth Calibration: Step 301: Secure the armored optical cable 2 at the wellhead and connect it to the distributed acoustic data acquisition subsystem 1.

[0112] Step 302: Start the distributed acoustic wave data acquisition subsystem 1 and perform a device self-test.

[0113] Step 303: Under the action of the cable counterweight 304 and the control of the logging winch 301, the high-pressure resistant single-mode armored optical cable 2 is slowly lowered to the bottom of the borehole. During the lowering process, the actual depth of the armored optical cable 2 entering the borehole is recorded, and the process is paused every 50m for depth calibration using a cable tension meter, with a calibration error ≤ ±0.1m. Simultaneously, the actual interpreted depth of the armored optical cable 2, as tested by the distributed acoustic data acquisition subsystem 1, is calibrated. Finally, the end of the armored optical cable 2 is located to the target aquifer, which in this embodiment is 719m.

[0114] Step 304, Dynamic sealing of armored optical cable: The armored optical cable 2 passes through the segmented graphite packing seal assembly of the fiber optic sealing port 405, ensuring no bending. Using a torque wrench, tighten the gland in three increments (20 Nm → 35 Nm → 50 Nm) to create a radial pressure gradient seal. Pressurize to 10 MPa and hold for 10 minutes; the strain of the armored optical cable should be ≤0.1%, with no leakage.

[0115] Step 305, Connecting the Mine Water Recharge Subsystem: Tighten the threaded end of the reinjection pipe 501 of the mine water reinjection subsystem 5 to the orifice water injection sealing connector 406 with a torque ≥300Nm; inspect the welded joint to ensure there are no cracks / pores, and perform a leak test.

[0116] Step 4, High-precision monitoring of the water injection process (water injection condition: taking stage 2 as an example): Step 401, Start-up and parameter control of the mine water reinjection subsystem: The mine water reinjection subsystem 5 is started. The industrial water injection pump 503 draws water from the mine wastewater treatment pond 504 and pressurizes it to the target pressure. The water injection valve 504 adjusts the valve opening according to the real-time monitoring data (pressure, flow) of the reinjection power monitoring subsystem 6 to optimize the injection profile.

[0117] In this embodiment, the reinjection parameters are set as follows: Target pressure range: 1.5-2.0 MPa; Duration: 5 hours and 30 minutes (from 18:20:48 to 23:59:48 on August 8, 2024). The flow meter monitored the total water injection volume as 20.52 m³.

[0118] Step 402: High-precision monitoring is performed using a distributed acoustic wave data acquisition subsystem. In this embodiment, the core acquisition parameters of the distributed acoustic wave data acquisition subsystem 1 are configured as follows: the spatial sampling interval is 1m, the sampling frequency is 2kHz to meet the requirements of transient acoustic wave capture, and the storage time of a single file is 6s to balance data continuity and storage efficiency.

[0119] In this embodiment, the real-time monitoring process of the distributed acoustic wave data acquisition subsystem 1 is as follows: The narrow-linewidth laser emitted by the laser 101 is modulated by the acousto-optic modulator 103 to form a linearly swept frequency optical pulse. The optical power is further enhanced by the amplifier 104 (erbium-doped fiber amplifier) ​​and then injected into the single-mode armored optical cable 2 through the circulator 105. The backscattered Rayleigh signal generated in the backfill borehole 4 interferes with the local narrow-linewidth laser. The polarization fading is eliminated by the polarization diversity receiver 106. The electrical signal is output by the balanced photodetector 107. A field-programmable gate array (FPGA) 110 achieves nanosecond-level precision timing control; a 24-bit ADC performs analog-to-digital conversion, and the output data rate can reach 50kHz after passing through a Σ-Δ modulator. To obtain Rayleigh scattering signals at various locations within the borehole in real time, the FPGA 110 is used to perform frequency mapping on the ADC-converted data stream. First, digital domain beat frequency is performed, followed by rotating vector averaging and phase difference operations to demodulate the intensity and phase information of the acoustic signal at various locations of the armored optical cable 2 in real time.

[0120] Step 5, Acoustic signal processing and feature extraction: A sound wave signal processing and feature extraction subsystem is used for sound wave signal processing and feature extraction.

[0121] Step Six: Intelligent Injection Channel Identification and Flow Estimation: An intelligent injection channel identification and flow estimation subsystem is used for injection channel identification and flow estimation.

[0122] Step 601, Injection Channel Identification: Injection channel identification is performed using an injection channel identification module. In this embodiment, the injection channels identified based on sub-band energy are shown in the attached figure. Figure 5 As shown, these are fiber optic channels 904-924, 936-941, 948-653, 986-989, 995-1001, and 1014-1021, corresponding to drilling depths of 579-599m, 611-616m, 623-628m, 661-664m, 670-676m, and 689-696m, respectively.

[0123] Step 602, Flow distribution calculation: The flow distribution calculation module is used to calculate the flow distribution. In this embodiment, the total injection volume is... It is 20.52 m³. The calculation yields... m 3 , m 3 , 1.49m 3 , m 3 , m 3 , 4.92m 3 .

[0124] Step 603, Intelligent Identification of Advantageous Channels: The dominant channel is identified using an intelligent identification module.

[0125] In this embodiment, the loss function of the LightGBM multi-class classification model adopts the weighted multi-class log loss:

[0126] In the formula: Indicates the first Weighted multi-class log loss for each sample; Indicates the sequence number of the sample; Indicates the number of samples; Indicates category; Represents class weights, used to address the problem of imbalanced samples; Indicates the first Does the true label of each sample belong to a category? c (If the sample) i The true category is c ,but =1, otherwise 0); The model predicts the first i Each sample belongs to category c The probability (i.e. the softmax probability output by the model).

[0127] In this embodiment, taking channel 1 as an example, it is assumed that... =1.0, =1.0, =1.0, Table 1 shows the calculated value. , and .

[0128] Table 1 shows the calculation results.

[0129] In this embodiment, based on the data from this embodiment, the classification results after model training are shown in Table 2.

[0130] Table 2. Intelligent classification results of the dominant channel

[0131] In this embodiment, early stopping is used to prevent overfitting, five-fold cross-validation is used to evaluate the model's generalization ability, feature contribution is analyzed based on SHAP values ​​to achieve model interpretability, and manual review is automatically triggered for samples with low confidence (<0.7).

[0132] Compared with the traditional threshold method, the monitoring method of the present invention improves the accuracy by 12.8%, maintains stable classification performance when the number of channels changes, and provides reliable intelligent decision support for mine water injection monitoring.

Claims

1. A high-precision dynamic monitoring system for mine water reinjection, characterized in that, It includes a distributed acoustic data acquisition subsystem (1), which is connected to one end of an armored optical cable (2). The armored optical cable (2) is sent into the reinjection borehole (4) through an armored optical cable laying device (3). The reinjection borehole (4) is also connected to a mine water reinjection subsystem (5) for mine water reinjection. The distributed acoustic wave data acquisition subsystem (1) includes a laser (101), an optical fiber coupler (102), an acousto-optic modulator (103), an amplifier (104), and a circulator (105) connected in sequence; the circulator (105) is connected to one end of the armored optical cable (2); the circulator (105) is also connected to a polarization diversity receiver (106), which is connected to an analog-to-digital conversion circuit (109) through a first balanced photodetector (107) and a second balanced photodetector (108). The analog-to-digital converter (109) is connected to the input of the field-programmable gate array (110), which is connected to the server (111). The output of the field-programmable gate array (110) is connected to the digital-to-analog converter (112), which is connected to the input of the radio frequency amplifier (113), which is connected to the output of the acousto-optic modulator (103). The fiber optic coupler (102) is also connected to the polarization diversity receiver (106). The server (111) has a built-in acoustic signal processing and feature extraction subsystem and an intelligent injection channel identification and flow estimation subsystem; The acoustic signal processing and feature extraction subsystem includes a wavelet multi-scale decomposition module, a target frequency band extraction module, a sub-band energy calculation module, a non-target frequency band suppression module, and a signal reconstruction module. The intelligent injection channel identification and flow estimation subsystem includes an injection channel identification module, a flow distribution calculation module, and a dominant channel intelligent identification module; The advantageous channel intelligent recognition module includes a multi-dimensional feature extraction submodule, a LightGBM intelligent classification submodule, an adaptive optimization submodule, and a result evaluation and output submodule. The wavelet multi-scale decomposition module is used to decompose the input signal. Perform discrete wavelet transform to decompose into J Sub-zone; The target frequency band extraction module is used to extract the frequency band of groundwater flow characteristics; The sub-band energy calculation module is used to calculate the total energy of each scale factor; The non-target frequency band suppression module is used to set the non-target frequency band coefficients to zero; The aforementioned signal reconstruction module is used to reconstruct a signal that retains only the energy characteristics of the target frequency band. ; The injection channel identification method of the injection channel identification module is as follows: when the conditions are met... , When the number of consecutive depth points is ≥3, it is determined to be a valid injection channel; in: Indicates frequency band energy; Indicates a time window; Indicates the drilling depth; Indicates an empirical threshold; This represents the average energy of the background noise; The reconstructed signal representing the target's frequency band energy characteristics; The method for calculating the flow distribution of each reinjection channel in the flow distribution calculation module is as follows: ； ； In the formula: Indicates injection of the first The flow rate of each recharge channel; Indicates the sequence number of the reinjection channel; Indicates the total injection volume; Indicates the first Flow weight of each recharge channel; Indicates the first The injection length of each recharge channel; P _ i Indicates the first Average energy of each reinjection channel; Indicates the first The total length of each test segment; Indicates the first Total energy of each test segment; Indicates the sequence number of the test segment; The multi-dimensional feature extraction submodule is used to automatically extract quantitative feature vectors for model training from the preprocessed acoustic signal and flow data. ; The LightGBM intelligent classification submodule is a multi-classification model built using the LightGBM framework. The input to the model is a quantitative feature vector. The model outputs a probability prediction of the channel type; The adaptive optimization submodule integrates a Bayesian optimizer for automatically searching the adjustable coefficients in the LightGBM intelligent classification submodule. The optimal threshold coefficient ; The result evaluation and output submodule is used to receive the prediction results from the classification module and output structured data including channel type, prediction confidence and feature importance. The total energy of the scale coefficients mentioned above ; In the formula: Indicates the first Total energy of the scale factor; Sequence number indicating scale; The sequence number representing the detailed complex coefficients; Indicates the first The first scale Each detailed complex coefficient; For the first The length of the scaling factor; The quantitative feature vector ; In the formula: Indicates the first The absolute flow rate of each recharge channel; Indicates the first The relative flow rate percentage of each recharge channel; This represents the standardized score of the traffic. Indicates the first Energy density of each reinjection channel; Indicates the first The length of the injection section of each recharge channel; Indicates the first The burial depth characteristics of each recharge channel; Indicates the first Flow fluctuation characteristics of each recharge channel ; Indicates the first The standard deviation of flow rate in each reinjection channel within the time window. Indicates the first The arithmetic mean of the flow rate of each recharge channel within the time window; The LightGBM intelligent classification submodule predefines three types of channels: Category 0: Non-dominant channel, i.e. < ; Category 1: Secondary dominant channel, i.e. ≤ ≤ T ; Category 2: Advantageous Channel, i.e. > T ; Indicates the judgment threshold. ; This represents the adjustable coefficient.

2. The high-precision mine water reinjection dynamic monitoring system as described in claim 1, characterized in that, The armored optical cable laying device (3) includes a logging winch (301), a ground fixed pulley (302), an aerial fixed pulley (303), and an optical cable counterweight (304). The aerial fixed pulley (303) is mounted on a support (305) and is located above the reinjection borehole (4). The armored optical cable (2) is wound around the logging winch (301). The other end of the armored optical cable (2) passes around the ground fixed pulley (302) and the aerial fixed pulley (303) in sequence. The other end of the armored optical cable (2) extends into the reinjection borehole (4). The other end of the armored optical cable (2) is connected to the optical cable counterweight (304).

3. The high-precision mine water reinjection dynamic monitoring system as described in claim 1, characterized in that, The reinjection borehole (4) includes a deep well reinjection borehole bare hole (401), a technical casing (402) is fitted inside the deep well reinjection borehole bare hole (401), a surface casing (403) is fitted outside the technical casing (402), a borehole connector (404) is installed on the top of the technical casing (402), and a borehole connector (404) is provided with a borehole fiber optic sealing port (405) and a borehole water injection sealing connector (406); the other end of the armored optical cable (2) passes through the borehole fiber optic sealing port (405) and enters the deep well reinjection borehole bare hole (401).

4. The high-precision mine water reinjection dynamic monitoring system as described in claim 1, characterized in that, The mine water reinjection subsystem (5) includes a mine wastewater treatment tank (501), one end of which is connected to a reinjection pipeline (502), and the other end of which is detachably connected to the orifice water injection sealing connector (406) of the reinjection borehole (4); an industrial water injection pump (503) and a water injection valve (504) are provided on the reinjection pipeline (502).

5. The high-precision mine water reinjection dynamic monitoring system as described in claim 1, characterized in that, It also includes a reinjection power monitoring subsystem (6), which includes an electromagnetic flowmeter (601) and a pressure sensor (602) installed on the reinjection pipeline (502) of the mine water reinjection subsystem (5). The electromagnetic flowmeter (601) and the pressure sensor (602) are connected to a multi-channel USB data acquisition card (603), which is connected to the reinjection power parameter monitoring master station (604). The electromagnetic flowmeter (601), the pressure sensor (602), and the multi-channel USB data acquisition card (603) are respectively connected to the power module (605).

6. A high-precision dynamic monitoring method for mine water reinjection, characterized in that, The high-precision dynamic monitoring method for mine water reinjection adopts the high-precision dynamic monitoring system for mine water reinjection as described in any one of claims 1 to 5; The high-precision dynamic monitoring method for mine water reinjection includes the following steps: Step 1: Prepare the drill hole and casing for water injection; Step 2: System initialization and parameter preset; Step 3, Armored Optical Cable Laying and Depth Calibration: Step 301: Fix the armored optical cable (2) at the wellhead and connect it to the distributed acoustic data acquisition subsystem (1); Step 302: Start the distributed acoustic wave data acquisition subsystem (1) and perform device self-test; Step 303, lower the armored optical cable (2); Step 304, Armored optical cable (2) dynamic sealing; Step 305, connect the mine water reinjection subsystem (5); Step 4: High-precision monitoring of the water injection process: Step 401, Start-up and parameter control of the mine water reinjection subsystem (5); Step 402: High-precision monitoring is performed using a distributed acoustic wave data acquisition subsystem (1); Step 5, Acoustic signal processing and feature extraction: A sound wave signal processing and feature extraction subsystem is used for sound wave signal processing and feature extraction. Step Six: Intelligent Injection Channel Identification and Flow Estimation: Step 601, Injection Channel Identification: Injection channel identification is performed using an injection channel identification module; Step 602, Flow distribution calculation: The flow distribution calculation module is used to calculate the flow distribution. Step 603, Intelligent Identification of Advantageous Channels: The dominant channel is identified using an intelligent identification module.

Images