Deep mine ground hydraulic fracturing effect evaluation and monitoring and early warning system

Abstract

This invention discloses a system for evaluating, monitoring, and warning the effectiveness of hydraulic fracturing in deep mines. The system comprises a data acquisition layer, a data processing layer, an intelligent analysis layer, a visualization layer, and an application layer. This invention constructs a standardized engineering database by integrating geological, construction, and microseismic monitoring data; it utilizes fractal geometry theory and dynamic inversion algorithms to build a digital three-dimensional fracture network model; it performs parallel calculations of multi-dimensional quantitative evaluation indicators such as pressure relief effect, fracture system effectiveness, and achievement of engineering objectives; and it combines machine learning models for intelligent diagnosis and graded early warning. This invention solves the problems of existing technologies, such as single evaluation dimensions, reliance on experience, and delayed early warning. It realizes the transformation of fracturing effect evaluation from experience-based judgment to multi-dimensional quantitative intelligent evaluation, and the leap from passive response to proactive early warning, providing precise decision support for safe and efficient mining in deep mines.

Description

Technical Field

[0001] This invention relates to the technical fields of mine safety and disaster prevention, specifically to a deep mine surface hydraulic fracturing effect evaluation and monitoring early warning system that combines the Internet of Things, big data analysis and artificial intelligence. Background Technology

[0002] As mineral resources are mined at greater depths, dynamic disasters such as rock bursts and rockbursts caused by high ground stress are becoming increasingly serious. Surface hydraulic fracturing, as an effective regional stress relief technology, is widely used. However, current assessments of fracturing effectiveness largely rely on post-hoc, isolated single indicators (such as the reduction of microseismic events), lacking a comprehensive evaluation system that integrates multiple parameters, is quantitative, and dynamic. Furthermore, the response of the surrounding rock after fracturing and stress relief is a dynamic process; traditional monitoring methods cannot achieve real-time analysis, early warning, and closed-loop decision support, leading to delayed engineering control and significant safety hazards. Existing technologies suffer from data silos, subjective evaluation, and delayed early warning.

[0003] Therefore, there is an urgent need for an integrated technical solution that can break down data barriers, integrate multi-source information, provide real-time dynamic evaluation, provide intelligent early warning, and support intuitive decision-making, so as to realize the paradigm shift of hydraulic fracturing engineering from "experience-driven" to "data and model-driven", and fundamentally improve the intelligence and scientific level of deep ground pressure disaster prevention and control. Summary of the Invention

[0004] To address, or at least partially address, the technical problems existing in the background art, the present invention provides a system and method for evaluating and monitoring the effects of hydraulic fracturing on the surface of deep mines, which has the capabilities of real-time collaborative monitoring, dynamic evaluation through multi-source data fusion, and intelligent advanced prediction. This facilitates a fundamental shift from experience-based decision-making to data-driven decision-making and helps improve the intelligence and scientific level of deep ground pressure disaster prevention and control.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] In a first aspect, the present invention provides a system for evaluating, monitoring, and early warning of the effect of surface hydraulic fracturing in deep mines. This system includes: a data acquisition layer, a data processing layer, and an intelligent analysis layer, wherein: The data acquisition layer is used to acquire multi-source heterogeneous data, which includes construction parameters, rock mass response parameters, geological design parameters, and mining disturbance parameters. The data processing layer is used to clean and remove anomalies from multi-source heterogeneous data, align it with a unified benchmark in time and space, and store it in the database to provide data support for the intelligent analysis layer. The intelligent analysis layer includes a fracture inversion module, an effect evaluation module, and an early warning analysis module, wherein: the fracture inversion module dynamically inverts the fracture propagation morphology based on processed data; the effect evaluation module calculates a comprehensive evaluation index based on multi-dimensional quantitative evaluation indicators; and the early warning analysis module provides early warning of hydraulic fracturing risks based on a prediction model.

[0007] Optionally, the data acquisition layer acquires multi-source heterogeneous data through an air-to-ground-pore monitoring network, which includes: The parameter direct acquisition equipment directly collects the operating parameters of the fracturing pump injection equipment, including pressure and water injection flow rate, in real time through an industrial IoT gateway; The surface monitoring array, which includes a microseismic monitoring network and surface subsidence monitoring points, is used to capture the global vibration wave field and surface deformation field induced by fracturing and mining. The borehole monitoring chain consists of fiber optic stress sensors and distributed fiber optic acoustic sensing systems arranged in series in monitoring boreholes in the target fracturing layer and key layers above and below, for in-situ, continuous, and distributed measurement of stress and strain fields.

[0008] Optionally, based on the acquired data, microseismic events can be automatically identified using the long-short time window mean ratio method and / or machine learning algorithms.

[0009] Optionally, the data processing layer has a multi-source data access interface, supports the import of real-time data streams and historical data, and transmits the unified data storage structure to the cloud data hub; the data hub has a built-in stream processing engine to perform real-time alignment, interpolation and quality verification of multi-source heterogeneous data, forming standardized time-series data blocks indexed by events.

[0010] Optionally, in the intelligent analysis layer, the crack inversion module uses time-series data as input, and performs crack inversion using a full-time-step fitting inversion algorithm through the PKN crack propagation model, wherein: The PKN crack propagation model is as follows:

[0011] in, For in position and time The width of the crack at that time The height of the crack. For Young's modulus, Poisson's ratio, This refers to the net pressure within the seam. The objective function is defined as:

[0012] in, for Real-world measured data from distributed fiber optic acoustic wave sensing. Based on crack parameters Forward simulation data, For regularization; as new data flows in, the crack parameters are updated in real time using sliding window or recursive Bayesian estimation.

[0013] Optionally, in the intelligent analysis layer, the effect evaluation module calculates a comprehensive evaluation index based on multi-dimensional quantitative evaluation indicators using linear weighting, fuzzy comprehensive evaluation, or neural network methods, and classifies the evaluation results into different levels based on the comprehensive evaluation index; wherein, the multi-dimensional quantitative evaluation indicators include pressure relief effect, crack effectiveness, and achievement of engineering objectives.

[0014] Optionally, in the intelligent analysis layer, the early warning analysis module adopts a time series prediction model including ARIMA, LSTM or random forest to predict the trends of pressure, flow and crack propagation path; and based on the prediction results and threshold conditions, triggers a three-level alarm mechanism of normal, attention and early warning, and reduces false alarms and duplicate alarms through association rules and causal analysis.

[0015] Optionally, the system also includes a visualization layer that uses web-based multi-rendering to visualize data and calculation results and supports automatic report generation.

[0016] Optionally, the system also includes an application layer that provides application entry points for real-time monitoring, effect evaluation, early warning management, report generation, and system settings.

[0017] Secondly, the present invention also provides a method for evaluating and monitoring the effect of hydraulic fracturing on the surface of deep mines. This method uses the above-mentioned system to evaluate and monitor the effect of hydraulic fracturing on the surface of deep mines.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects: 1. This invention provides a deep mine surface hydraulic fracturing effect evaluation and monitoring early warning system with real-time collaborative monitoring, multi-source data fusion dynamic evaluation and intelligent advanced prediction capabilities. It solves the problems of single evaluation dimensions and reliance on experience in existing technologies, and realizes the transformation of fracturing effect from experience judgment to multi-dimensional quantitative intelligent analysis, which is conducive to providing accurate decision support for the safe and efficient mining of deep mines.

[0019] 2. By integrating functions such as multi-source data sensing, crack dynamic inversion, multi-dimensional effect collaborative evaluation, multi-parameter fusion intelligent early warning, and visualization decision support, this invention achieves comprehensive monitoring and scientific analysis and evaluation of the entire process of deep ground pressure surface hydraulic fracturing treatment, effectively improving the prevention and control capabilities of mine dynamic disasters.

[0020] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0023] Figure 1 This is a schematic diagram of the overall system architecture provided by the present invention; the diagram illustrates the logical composition of the system platform, including the interaction between the data acquisition layer, data processing layer, intelligent analysis layer, visualization layer, and application layer.

[0024] Figure 2 This is a schematic diagram of the user login interface provided by the present invention; the diagram shows the style of the system platform login interface, through which users can log in to the system platform.

[0025] Figure 3 This is a schematic diagram of the monitoring screen provided by the present invention; the diagram shows a three-dimensional example of the monitoring screen and various functional modules.

[0026] Figure 4 This is a schematic diagram of the layout of the main interface for system monitoring provided by the present invention; the diagram includes a system status bar at the top, a key indicator display area in the middle, a dynamic monitoring chart area, and a quick control bar at the bottom.

[0027] Figure 5 This is a schematic diagram of the data intelligent analysis interface provided by the present invention; the diagram shows the partitioned layout of the analysis interface, including the parameter control panel on the left (data source selection, variable and algorithm configuration) and the multi-tab analysis result display area on the right.

[0028] Figure 6 This is a schematic diagram of the trend analysis interface provided by the present invention; the diagram shows the data actually monitored by the monitoring system and the data predicted by the model, which facilitates comparison and early warning.

[0029] Figure 7 This diagram illustrates the export interface for the data analysis report provided by the present invention; the diagram includes a report generation area, an execution summary area, and a key discovery area.

[0030] Figure 8This is a schematic diagram of the system settings interface provided by the present invention; the diagram includes a menu bar on the left, a save / restore bar in the upper right corner, and a settings area in the middle. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The accompanying drawings are intended to exemplarily illustrate the system architecture, interface layout, data processing flow, and core analysis functions involved in the present invention. Those skilled in the art can understand the essence of the present invention by referring to the accompanying drawings and specific embodiments.

[0032] In the description of this invention, it should be noted that some processes described in this application specification and accompanying drawings include multiple operations that appear in a specific order. However, it should be clearly understood that these operations may be performed in any order or in parallel. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention.

[0033] This invention aims to overcome the systemic shortcomings of existing technologies and provide a technical solution for evaluating and monitoring the effectiveness of surface hydraulic fracturing in deep mines, possessing real-time collaborative monitoring, multi-source data fusion dynamic evaluation, and intelligent advanced prediction capabilities. This achieves a fundamental shift from experience-based decision-making to model- and data-driven decision-making. To achieve the above objectives, this invention utilizes the following technical solution: 1. Real-time collaborative monitoring and data management system: This system constructs an integrated air-ground-aperture real-time collaborative monitoring network with synchronized multi-parameter monitoring, and establishes a unified data fusion center, specifically including: Surface monitoring array: Deploy a high-precision microseismic monitoring network and surface subsidence monitoring points to achieve global capture of the vibration wave field and surface deformation field induced by fracturing and mining; In-hole monitoring chain: In the monitoring boreholes of the target fracturing layer and the key layers above and below, fiber optic grating (FBG) stress sensors and distributed fiber acoustic wave (DAS) systems are arranged in series to realize in-situ, continuous and distributed measurement of stress and strain fields. Direct acquisition of engineering parameters: Through the industrial IoT gateway, the operating parameters of the fracturing pump injection equipment, including pressure and water injection flow rate, are directly collected in real time. Basic Data Integration and Management Center: Multi-source data access interface, supports real-time data stream and historical data import, unified data storage structure and transmission to cloud data hub; The data hub has a built-in stream processing engine to perform real-time alignment, interpolation and quality verification (including data validation, outlier detection and missing value handling) on ​​multi-source asynchronous data, forming standardized time-series data blocks indexed by "events" (such as a single fracturing operation); Supports multi-user and multi-role data access control.

[0034] Automatic identification of microseismic events: The STA / LTA algorithm and machine learning algorithm are used to automatically identify microseismic events, and the double-difference positioning method is used to improve the positioning accuracy. The magnitude, energy, moment tensor, rupture type, etc. of microseismic events are calculated, and the time series, spatial clustering and migration patterns of microseismic events are analyzed.

[0035] 2. Dynamic effect evaluation model for multi-source data fusion: The evaluation system of this invention transcends static, single-index evaluation, establishing a dynamic comprehensive evaluation model driven by multi-source data fusion, the contents of which include: Dynamic and refined inversion of fracture networks: Real-time data stream processing engine, synchronous monitoring of multiple parameters (including water injection pressure, flow rate, fracture length, microseismic events, etc.), supporting millisecond-level data updates; Real-time inversion of fracture propagation morphology based on monitoring data using dynamic inversion algorithms; Establishment of an anomaly detection mechanism, using threshold and pattern recognition algorithms to provide real-time early warning of abnormal operating conditions.

[0036] Multi-dimensional comprehensive evaluation of fracturing effect: Establish a comprehensive evaluation index system, including multiple evaluation indicators such as pressure relief effect, fracture effectiveness, and achievement of engineering objectives; determine the weight model, and determine the weight of each indicator by combining expert experience and data analysis; and visualize the comprehensive evaluation results through radar charts and scoring cards.

[0037] Weight determination method: Initial weights are determined by collecting opinions from domain experts through expert surveys; a judgment matrix is ​​constructed and eigenvectors are calculated using the analytic hierarchy process (AHP) to perform consistency checks; objective weights are calculated based on the dispersion of the data using the entropy weighting method; subjective and objective weights are combined using a combined weighting method to obtain comprehensive weights; finally, the weight coefficients are dynamically adjusted based on actual application effects and historical data.

[0038] Comprehensive evaluation calculation: linear weighting uses fixed weights for simple linear combination; fuzzy comprehensive evaluation establishes a fuzzy relation matrix and uses a weighted average operator for synthesis; neural networks dynamically adjust weights through training to achieve a higher-scoring evaluation model. Evaluation results are graded into five levels: Excellent (≥90), Good (80-89), Average (70-79), Satisfactory (60-69), and Unsatisfactory (<60).

[0039] 3. Intelligent advanced prediction and risk warning mechanism: This invention constructs a machine learning-based early warning analysis module (a multi-parameter fusion intelligent prediction and early warning module), specifically including: Prediction Models: Time series prediction models such as ARIMA, LSTM, and random forest are used to predict the changing trends of key parameters such as pressure and flow rate; crack propagation is predicted based on fluid mechanics principles and monitoring data, including crack propagation path prediction. Risk warning system: Provides early warnings of potential risks based on forecast results and risk assessment models; Multi-level early warning mechanism: It sets three levels: normal, attention, and warning, and the triggering conditions include various types such as thresholds and trends; it adopts association rules and causal analysis to reduce false alarms and duplicate alarms; Alarm push mechanism: Supports multiple methods such as SMS notification, APP push, and email notification, and establishes a complete closed-loop management process for alarm confirmation, processing, feedback and closure.

[0040] 4. Visualization and report generation: This invention employs web-based multi-rendering technology to visualize data and calculation results, and supports automatic report generation, specifically including: Real-time monitoring dashboard: Displays historical trends and real-time changes of key parameters such as pressure and flow rate in the form of an instrument panel; displays equipment status and warning status in the form of red, yellow and green lights; and uses bar charts, line charts, distribution histograms, box plots, heat maps and other methods to display changes or statistical distributions of different parameters. Three-dimensional geological visualization: Establishing three perspectives—3D stereo, planar view, and geological profile—to visualize the on-site geological conditions. It allows for an intuitive view of the distribution of microseismic events and parameters such as pressure at different monitoring points, facilitating timely acquisition of data information. Data filtering and querying: Supports selecting data within a certain range such as time and space; supports filtering with multiple parameters and quick data viewing. Automatic report generation: Automatically extracts the data required for the report from the database; automatically generates corresponding charts based on the data; automatically generates text descriptions based on templates and data; automatically generates conclusions and recommendations based on the analysis results; Report export and sharing: Supports exporting to multiple formats such as PDF, Word, and Excel; supports additional functions such as page setup and adding watermarks; supports sending via email, sharing via link, and sharing via QR code.

[0041] The following is combined with Figures 1 to 8 As shown, the specific implementation methods and working principles of the present invention will be described in detail below: This invention discloses a system for evaluating and monitoring the effectiveness of surface hydraulic fracturing in deep mines, such as... Figure 1 The diagram shown is the core flowchart of the data processing and intelligent analysis of this invention. This diagram concisely illustrates the complete closed loop from data input to application output, clearly mapping to the system architecture of this invention. Wherein: ① Data Input Layer: Acquires multi-source heterogeneous data, including four data sources: “construction parameters”, “rock mass response”, “geological design”, and “mining disturbance”, comprehensively covering all data dimensions required for fracturing effect evaluation.

[0042] This invention employs an automatic microseismic event identification method to access microseismic data. Its principle and core formula are as follows: The STA / LTA algorithm (long-window mean ratio method) calculates the ratio of the characteristic function's mean within a short-window and a long-window, triggering an event when the ratio exceeds a preset threshold. In practical applications, the original signal is usually preprocessed (e.g., mean removal, trend removal) before calculation. The left-aligned algorithm is a classic STA / LTA algorithm and the default algorithm in many software programs. This algorithm places the current calculation point... Set the left edge of the STA window and the right edge of the LTA window: Short-term average (STA):

[0043] Long-term average (LTA):

[0044] ratio:

[0045] in, Indicates the time index is The sample signal at time 10:00. Indicates at a point in time The signal value; This indicates the number of samples in the short-time window. Indicates the length of the long-term window and the number of samples; This represents the ratio of the means.

[0046] Machine learning recognition algorithms: Machine learning can build high-precision classification models by extracting features from the time domain, frequency domain, and fractal dimension. Commonly used algorithms and their performance include: Support Vector Machines (SVMs) have high accuracy in identifying microseismic events in mines. Their core principle lies in using a classification decision function to determine the category of a microseismic signal.

[0047] In the formula, It is the first in the training set The feature vector of each sample yes The corresponding category label (e.g., +1 indicates a microseismic event, -1 indicates noise). It's a kernel function; its purpose is to calculate new samples. With training samples The inner product in high-dimensional space avoids complex calculations directly in high-dimensional space. and These are the model parameters determined during the training process, which determine the final classification hyperplane. N is the total number of training samples.

[0048] Ultimately, the function The output value determines the type of signal.

[0049] AdaBoost is an ensemble learning algorithm. Its core idea is to combine multiple simple weak learners to construct a powerful strong learner. It exhibits good stability in microseismic identification. The formula is as follows:

[0050] in, It is the first In microseismic identification, a simple decision tree is often used to play the role of a weak learner. It is the first The weights of a weak learner determine its performance (lower error rate). The larger the value; It represents the number of weak learners.

[0051] Double-difference positioning: Double-difference positioning is a relative positioning method designed to eliminate systematic errors caused by inaccurate velocity models. Its core assumption is that when the distance between two events is much smaller than the distance from the event to the station, their ray paths are very similar, and therefore velocity anomalies on the paths can be canceled out.

[0052] Moment tensor inversion: By analyzing microseismic waveform data recorded by sensors and combining it with Green's function, which describes the wave propagation law in rock, the mechanical process of the seismic source can be inferred. Its core is to solve a system of linear equations to obtain a second-order tensor describing the characteristics of the seismic source. By decomposing this tensor into isotropic and double-couple components, it is possible to identify whether the microseismic event is caused by shear slip, rock tension, or compressional fracturing, ultimately providing crucial evidence for judging rock mass stability. However, the accuracy of this algorithm is highly dependent on the accuracy of the formation velocity model and the quality of the data.

[0053] Time series analysis refers to the pattern of microseismic events over time, mainly including indicators such as event rate, B-value, and time interval distribution. The B-value is a key parameter in the Gutenberg-Richter relationship, reflecting the relative proportion of events of varying magnitudes and closely related to the stress state (generally, a low B-value corresponds to high stress, and a high B-value corresponds to low stress or crack development). The commonly used method for estimating the B-value is the maximum likelihood method.

[0054] in, The average magnitude of the event. This is the integrity magnitude threshold. This formula applies to cases where the magnitude is continuously distributed without truncation.

[0055] Spatial clustering analysis aims to group microseismic events according to their spatial location to identify active rupture zones, fracture zones, or potential faults. Commonly used methods include K-means and Gaussian mixture models (GMM), among which: The principle of K-means clustering is to divide the data into K clusters such that the sum of squares within each cluster is minimized. The objective function formula is as follows:

[0056] in, For clusters, It is the cluster center.

[0057] The Gaussian Mixture Model (GMM) assumes that the data is generated by a mixture of multiple Gaussian distributions and estimates the parameters using the Expectation-Maximization (EM) algorithm. The probability that each event belongs to each cluster is given by the posterior probability.

[0058] in, It is observation data The overall probability density function, The mixing coefficient, It is the covariance matrix (the shape can be constrained, such as spherical or congruent).

[0059] ② Data processing layer: This includes three modules: "data cleaning", "spatiotemporal alignment" and "storage". All multi-source heterogeneous data is cleaned, anomalies are removed, and spatiotemporal alignment is performed to unify the benchmark. Then, the data is stored in a hybrid database to provide high-quality data support for upper-level analysis.

[0060] ③ Intelligent Analysis Layer: Includes three core functions: "crack inversion", "effect evaluation" and "early warning analysis". This enables the system to receive processed data, complete dynamic crack modeling, multi-dimensional quantitative evaluation and real-time risk early warning, and generate analysis results.

[0061] The core idea of ​​dynamic inversion of crack propagation morphology is to use monitoring data (such as distributed fiber optic DAS, microseismic events) as input, and through physical models or data-driven methods, to extrapolate the length, height, width, and propagation direction of cracks in real time. PKN models or KGD models can be used. The key formula for the PKN model is:

[0062] in, w Let H be the crack width, E be the crack height, ν be Young's modulus, ν be Poisson's ratio, and p be the net pressure within the crack. This formula describes the relationship between crack width and pressure.

[0063] The inversion algorithm uses full-time fitting inversion, and the objective function is defined as:

[0064] in, for Real-time measured DAS data For parameter-based Forward modeling data of (crack geometry parameters) This is a regularization term used to overcome the ambiguity of the inversion. As new data flows in, the crack parameters are updated in real time using sliding window or recursive Bayesian estimation (such as Kalman filtering).

[0065] Based on crack inversion results and multi-source monitoring data (microseismic event distribution, stress field changes, displacement monitoring, etc.), a multi-dimensional quantitative evaluation index system can be constructed to automatically calculate the stress relief effect, crack effectiveness, and the degree of achievement of engineering objectives, thereby generating a comprehensive score and grade rating. The quantitative index calculation method is as follows: Establish parameters for evaluating the depressurization effect The percentage reduction in average stress relative to the initial stress within the monitoring area is calculated using the following formula:

[0066] in, To monitor the average stress reduction within the monitoring area, This represents the initial stress.

[0067] Establish parameters for evaluating the effectiveness of cracks The calculation formula is as follows:

[0068] in, To account for the crack volume, For proppant efficiency, This is the actual length. For the design length.

[0069] By taking multiple parameters and weighting them together, a target achievement parameter D is established, and the calculation formula is as follows:

[0070] in, , , The weighting coefficients must satisfy the condition that the sum of all values ​​equals 1, and are determined by the project priority. The actual crack length obtained from the inversion. The desired crack length is to be achieved in the design. This represents the actual crack height. The desired crack height is represented by D. D is a comprehensive score parameter ranging from 0 to 1, with a higher value indicating better target achievement.

[0071] The early warning analysis function integrates real-time monitoring data streams (microseismic event rate, stress change rate, crack propagation speed, etc.), uses the isolated forest algorithm to identify anomalies, and combines a multi-level early warning rule base to achieve dynamic risk early warning.

[0072] Isolation Forest is an efficient unsupervised anomaly detection algorithm suitable for high-dimensional, large-scale datasets. Its core idea is that outliers are "few but distinct," making them easier to isolate in a randomly partitioned feature space (i.e., requiring fewer partitions to reach leaf nodes). This algorithm quantifies the degree of anomaly by constructing multiple isolation trees and calculating the average path length of the samples. The expected value of the average path length of the isolation trees is:

[0073] in For the first A harmonic number, is the number of data subsets in the sample.

[0074] In the algorithm described above, a score closer to 1 indicates a greater anomaly. This score can be used to determine thresholds, categorizing the score as red (greater than 0.8, warning), yellow (greater than 0.6, caution), and green (less than 0.6, normal).

[0075] ④ Visualization layer: Includes "3D model" and "2D chart" functions, representing 3D visualization output and 2D chart output respectively, supporting users to rotate, zoom and query data through "interactive operations".

[0076] ⑤ Application Layer: Includes five application functions: "Real-time Monitoring", "Effect Evaluation", "Early Warning Management", "Report Generation" and "System Settings", providing users with a complete entry point for business functions. Figure 1 The dashed arrow "Configuration" pointing from the application layer to the data processing layer and intelligent analysis layer reflects the system's configurability and flexibility.

[0077] Therefore, Figure 1 The core technical process of this invention is summarized in a highly condensed manner, from data access to intelligent analysis, and then to visualization and application output, forming a complete data-driven closed loop, providing full-process technical support for deep geothermal surface hydraulic fracturing treatment.

[0078] See Figure 2 As shown, this embodiment of the invention provides a unified user login interface. After the user starts the system, the right side of the interface displays username and password input boxes, as well as "Login" and "Register" buttons. The system supports multiple user roles (such as monitors, analysts, and administrators) and associates them with different operation permissions. After entering the correct credentials during login, the system loads the corresponding functional modules and data views according to the role permissions and records login logs to ensure system security. If multiple login attempts fail consecutively, the account will be temporarily locked to prevent brute-force attacks.

[0079] See Figure 3 As shown in the illustration, this embodiment of the invention provides a large monitoring screen interface for intuitively displaying the overall situation of the fracturing area, suitable for centralized monitoring by the dispatch center. The center of the screen displays a three-dimensional view, a planar view, and a cross-sectional view, integrating the following elements: Topographic and stratigraphic models: different colors distinguish lithology; Real-time microseismic event sphere: The size of the sphere represents the magnitude, the color represents the energy level (red > yellow > green), and the position of the sphere corresponds to the coordinates of the event occurrence; Sensor point icons (stress gauge, displacement gauge): Blue squares represent stress sensor monitoring points, and purple squares represent displacement sensor monitoring points. Real-time readings are displayed when the mouse hovers over them.

[0080] Users can rotate, zoom, and pan in a 3D scene using a touchscreen or mouse to view any detail.

[0081] Function cards are distributed around the perimeter of the large screen, including: Data statistics card: Displays the total number of microseismic events, stress measurement points, and displacement measurement points; Microseismic Activity Trend Card: Displays microseismic activity at different times in real time; Real-time stress monitoring card: Displays stress data monitored by stress sensors in real time; Displacement change trend card: Displays the displacement changes monitored by the displacement sensor in real time; Fracturing Project Card: Displays the project name, location, and completion status of the project monitored by the system; Comprehensive evaluation card: Displays the overall performance index (0-100) in the form of a dashboard.

[0082] In this invention, the comprehensive evaluation calculation employs methods such as linear weighting, fuzzy comprehensive evaluation, and neural networks. The specific algorithms and formulas are as follows: Linear weighting is the simplest and most commonly used comprehensive evaluation method. It assumes that each evaluation indicator is independent and that its contribution to the overall evaluation value is linear. By assigning a weight to each indicator and then summing the indicator values ​​using weighted averages, a comprehensive score is obtained. Given n evaluation indicators, for the j-th evaluation object, the comprehensive score is... for:

[0083] in, For weight values, These are the processed indicator values.

[0084] Fuzzy comprehensive evaluation transforms index values ​​into fuzzy membership degrees by establishing a membership degree function, then obtains a comprehensive membership degree vector through fuzzy relation synthesis operation, and finally obtains the evaluation level based on the principle of maximum membership degree or weighted average.

[0085] Neural networks simulate the structure of neurons in the human brain, using a large amount of sample data to train the network and automatically learn the nonlinear mapping relationship between indicators and comprehensive evaluation. After training, the network can be used to score new samples. The loss function can be represented by the mean squared error (MSE).

[0086] See Figure 4 The image shows the monitoring interface of an embodiment provided by the present invention, which adopts a top-middle-bottom partition layout for easy daily monitoring operations: The top system status bar displays, from left to right, the system name, project mine name, system time, refresh, settings, and exit buttons. Clicking a button performs the corresponding action. Central Key Indicators Display Area: Displays the most important real-time indicators in card format, including: "Pressure Monitoring": Displays the maximum, minimum, and rate of change of the pressure during construction. "Flow Monitoring": Displays the cumulative flow, target flow, and rate of change during the water injection process; "Fractured Monitoring": Displays the fracture propagation rate, fracture width, and rate of change during hydraulic fracturing; "Comprehensive Evaluation": Displays the real-time evaluation index of the current fracturing segment (comprehensive evaluation, fracture propagation, stress release, and permeability) on a 100-point scale. Each card can be clicked to enter the corresponding detailed analysis interface; Dynamic monitoring chart area: Line charts and radar charts are used to display "Real-time Pressure and Flow Curves" and "Multi-dimensional Evaluation Analysis," respectively. Charts support selecting the monitoring time and hovering over data points for detailed viewing. Bottom quick control bar: Provides quick access to frequently used functions, including buttons such as "Start / Pause Data Refresh", "Generate Current Report", "Data Analysis", and "Video Monitoring", allowing users to switch quickly.

[0087] like Figure 5 The image shows the interface for intelligent data analysis, used for in-depth mining of historical or real-time data. It is divided into two parts, left and right. Among them: Left-side parameter control panel: Data source selection area: Select the analysis range from the drop-down menu, such as "Real-time monitoring data", "Historical database", "Start time", and "End time"; Variable configuration area: Supports user-defined independent and dependent variables for parameter analysis, such as "microseismic energy" for the Y-axis and "time" for the X-axis, and different analysis algorithms can also be selected.

[0088] The analysis results are displayed on the right side. Use multiple tabs to display different analysis results: Trend Analysis tab: Displays a list of raw data and calculation results used in the analysis, and supports full-screen display and chart export; The Correlation Matrix tab: When multivariate analysis is selected, it automatically calculates and displays a heatmap of correlation coefficients between variables, with the color intensity representing the strength of the correlation. The Statistical Analysis tab presents analysis results in the form of scatter plots, line charts, bar charts, etc., and the graphs are interactive (select and zoom, label data points, save as image); The Predictive Analytics tab allows users to select different prediction models, set prediction duration and confidence intervals, and generate prediction results, such as... Figure 6 As shown.

[0089] In this invention, the prediction model employs time series prediction models such as ARIMA, LSTM, and random forest. Wherein: The ARIMA model transforms a non-stationary time series into a stationary series through differencing, and then establishes an autoregressive moving average model. Maximum likelihood estimation or least squares estimation is used to estimate the model coefficients. The model is represented as ARIMA(p,d,q), where p is the autoregressive order, d is the differencing order, and q is the moving average order. The model equation is:

[0090] in, The time series is a stationary series after differencing. For constant terms, It is a white noise sequence. These are the autoregressive coefficients (i=1,2,...,p). Let be the moving average coefficients (j=1,2,...,q). Using the fitted model, predict the pressure values ​​k time points ahead and provide the prediction intervals (95% confidence level).

[0091] LSTM models effectively capture long-term dependencies in time series through gating mechanisms, making them suitable for predicting nonlinear and non-stationary sequences. First, the time series is converted to a supervised learning format, and the input features are normalized using Min-Max to accelerate convergence. Input and output layers are designed, with mean squared error (MSE) as the loss function and Adam as the optimizer, with an initial learning rate of 0.001. The dataset is divided into training (70%), validation (15%), and test (15%) sets. Early stopping is used to monitor the validation set loss; training stops when the loss does not decrease for 10 consecutive epochs. After training, the latest time-series data is input, and the model forward propagates to obtain the predicted values.

[0092] Random forests, by ensembled multiple decision trees, perform a non-linear mapping of input features, making them suitable for single-step or multi-step prediction of multivariate inputs. In this implementation, historical time-series data is constructed as a feature vector to predict future values. For an input feature vector X, each tree outputs a predicted value. The final prediction of a random forest is the average of all trees:

[0093] The specific implementation steps are similar to those of the LSTM model. The trained model is input into the feature vector of a new sample, and it can output the average prediction value of all trees.

[0094] Cluster Analysis tab: Automatically generates clustering results and the number of samples in different clusters in the form of a scatter plot, based on the selected clustering model and the number of clusters; Analysis Reports tab: such as Figure 7 As shown, you can click the button according to the text prompts to generate a PDF report, share the report, and export the corresponding chart attachments.

[0095] The system settings interface provided in this invention can be found in the embodiments. Figure 8 It adopts a layout of left-side menu bar - right-side settings area - upper right corner save and restore bar, used to configure system running parameters.

[0096] Left-side menu bar: Lists configurable modules grouped by function, including: "System Settings": Configure system language, display settings, and data storage range, etc. "Monitoring parameters": Configure the normal range of fracturing operation parameters, sensor sampling frequency, data storage period, etc. "Evaluation Model": Configure the indicator weights of the multi-dimensional evaluation model (can be manually adjusted or restored to the default entropy weight method combination weights with one click), and can choose whether to enable automatic model updates; "Early Warning Rules": Supports setting thresholds (such as upper limit of microseismic energy, upper limit of stress change rate) and corresponding early warning levels for each monitoring parameter; users can choose the early warning notification method (email, SMS, telephone). "Data Management": Select data source type and access API interface, import file data (Excel, CSV, text file), export data range, clean up abnormal duplicate data, statistical data, etc.

[0097] The central settings area dynamically switches between displayed settings items based on the selected menu item on the left. For example, when "Evaluate Model" is selected, the central area displays the weight sliders for the seven dimensions, as well as a "Restore Defaults" button.

[0098] The save / restore section in the upper right corner provides "Save All Settings," "Restore Defaults," and "Help" buttons. Clicking "Save All Settings" writes the current configuration to the database; clicking "Restore Defaults" restores factory settings; clicking "Help" displays common operations and chart operation methods.

[0099] The following example of a typical hydraulic fracturing operation illustrates the specific workflow of the intelligent analysis engine of this invention.

[0100] Step 1: Data Acquisition and Preprocessing Prior to fracturing operations, the system had already entered the geological model and fracturing design parameters for the target area. After operations commenced, the data acquisition layer collected real-time data on pumping pressure, flow rate, and other operational data. Simultaneously, the microseismic monitoring system continuously recorded induced microseismic events, and stress gauges and displacement gauges collected stress and displacement changes at a frequency of 1 Hz. All data was initially cleaned by edge computing nodes before being uploaded to the data center.

[0101] Step 2: Crack dynamic inversion: The intelligent analysis engine's fracture inversion module receives microseismic event data in real time and uses the Geiger localization algorithm and moment tensor inversion to update the 3D model of the fracture network every 5 minutes. For example, 30 minutes after fracturing in a certain section, the inversion shows that the fractures mainly extend in a northeast direction, with a length of about 120 meters and a height of about 25 meters, forming a complex fracture network. The system pushes the inversion results to the 3D visualization layer for real-time display.

[0102] Step 3: Multi-dimensional effect evaluation: At the end of the fracturing operation, the comprehensive evaluation module automatically triggers a phase evaluation. For example, based on all the collected data, scores are calculated for four parameters: fracture propagation range, stress release effect, permeability improvement, and microseismic activity (fracture 82, stress 95, permeability 88, microseismic activity 100). The crack propagation range score is obtained by multiplying the ratio of the actual crack length to the design length by 100. For example, when the design half length is 120 m and the actual inverted half length is 98.4 m, the ratio is 0.82 and the score is 82. The stress release effect score is directly calculated as the percentage of the average stress reduction in the monitoring area to the initial stress. For example, if the initial stress is 30 MPa and the average stress drops to 1.5 MPa after fracturing, the stress reduction of 28.5 MPa accounts for 95%, and the score is 95. The permeability improvement score is calculated by the achievement rate of the difference between the pre- and post-pressure permeability relative to the design target. For example, if the pre-pressure permeability is 0.01 mD, the post-pressure permeability is 0.88 mD, and the design target is 1.0 mD, the achievement rate is approximately 0.88, and the score is 88. The microseismic activity score is based on the ratio of the actual total microseismic energy (or number of events) to the design target. A perfect score of 100 is awarded if the actual value meets or exceeds the target. For example, if the design target energy is 10... 5 J, Actual monitored energy: 1.2 × 10⁻⁶ 5 J, with an achievement rate ≥1, scores 100.

[0103] By combining weights (crack 0.3, stress 0.3, permeability 0.2, microseismic activity 0.2) using the entropy weight method, the comprehensive effect index is calculated to be 0.3. 82 + 0.3 95 + 0.2 88+0.2 100 = 90.5, the rating is "Excellent". The system automatically generates an evaluation report and displays the index change curve on the main monitoring interface.

[0104] Step 4: Dynamic Early Warning: Two days after the fracturing operation concluded, the mining face approached the fracturing area. Real-time monitoring data showed a sharp increase in the frequency of microseismic events from an average of 5 times / hour to 30 times / hour, with the stress curve exhibiting an accelerated decline (exceeding the threshold of 0.2 MPa / h), and the displacement gauge detecting minor anomalies. The early warning module first triggered a primary warning (rule-triggered, yellow warning). Subsequently, the LSTM prediction model, based on the current trend, predicted that the energy release within the next hour might exceed 1000 J (red warning threshold). Considering multiple parameter characteristics, the system determined that there was a risk of rock mass instability, automatically escalating to an orange warning, and pushing the alert to the on-duty engineer via interface pop-ups, SMS, etc., indicating that it might be due to "mining-induced stress redistribution leading to microseismic activation." Upon receiving the warning, the engineer immediately notified the site to suspend operations, implement reinforced support measures, and record the response feedback in the system. 24 hours later, microseismic activity returned to normal, the warning was lifted, and a complete response loop was formed.

[0105] In summary, this invention integrates multi-source data sensing, crack dynamic inversion, multi-dimensional effect collaborative evaluation, multi-parameter fusion intelligent early warning, and visualization decision support to achieve comprehensive monitoring and scientific evaluation of the entire process of deep ground pressure surface hydraulic fracturing treatment, effectively improving the prevention and control capabilities of mine dynamic disasters.

[0106] Those skilled in the art will understand that embodiments of the present invention can be provided as systems, methods, devices, or computer program products, etc. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0107] It should be noted that the word "comprising" does not exclude the presence of components or steps not listed in the claims. The words "a" or "an" preceding a component do not exclude the presence of a plurality of such components. This invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer.

[0108] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0109] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A system for evaluating, monitoring, and early warning of the effectiveness of surface hydraulic fracturing in deep mines, characterized in that, The system comprises: a data acquisition layer, a data processing layer, and an intelligent analysis layer, wherein: The data acquisition layer is used to acquire multi-source heterogeneous data, which includes construction parameters, rock mass response parameters, geological design parameters, and mining disturbance parameters. The data processing layer is used to clean and remove anomalies from multi-source heterogeneous data, align it with a unified benchmark in time and space, and store it in the database to provide data support for the intelligent analysis layer. The intelligent analysis layer includes a fracture inversion module, an effect evaluation module, and an early warning analysis module, wherein: the fracture inversion module dynamically inverts the fracture propagation morphology based on processed data; the effect evaluation module calculates a comprehensive evaluation index based on multi-dimensional quantitative evaluation indicators; and the early warning analysis module provides early warning of hydraulic fracturing risks based on a prediction model.

2. The system according to claim 1, characterized in that, The data acquisition layer acquires multi-source heterogeneous data through an air-to-ground-pore monitoring network, which includes: The parameter direct acquisition equipment directly collects the operating parameters of the fracturing pump injection equipment, including pressure and water injection flow rate, in real time through an industrial IoT gateway; The surface monitoring array, which includes a microseismic monitoring network and surface subsidence monitoring points, is used to capture the global vibration wave field and surface deformation field induced by fracturing and mining. The borehole monitoring chain consists of fiber optic stress sensors and distributed fiber optic acoustic sensing systems arranged in series in monitoring boreholes in the target fracturing layer and key layers above and below, for in-situ, continuous, and distributed measurement of stress and strain fields.

3. The system according to claim 1, characterized in that, Based on the acquired data, microseismic events are automatically identified using the long-short time window mean ratio method and / or machine learning algorithms.

4. The system according to claim 1, characterized in that, The data processing layer has a multi-source data access interface, supports the import of real-time data streams and historical data, and transmits data to the cloud data hub in a unified data storage structure. The data hub has a built-in stream processing engine that performs real-time alignment, interpolation and quality verification on multi-source heterogeneous data to form standardized time-series data blocks indexed by events.

5. The system according to claim 4, characterized in that, In the intelligent analysis layer, the crack inversion module uses time-series data as input, employs the PKN crack propagation model, and uses a full-time-step fitting inversion algorithm to perform crack inversion, wherein: The PKN crack propagation model is as follows: in, For in position and time The width of the crack at that time The height of the crack. For Young's modulus, Poisson's ratio, This refers to the net pressure within the seam. The objective function is defined as: in, for Real-world measured data from distributed fiber optic acoustic wave sensing. Based on crack parameters Forward simulation data, For regularization; as new data flows in, the crack parameters are updated in real time using sliding window or recursive Bayesian estimation.

6. The system according to claim 1, characterized in that, In the intelligent analysis layer, the effect evaluation module calculates a comprehensive evaluation index based on multi-dimensional quantitative evaluation indicators using linear weighting, fuzzy comprehensive evaluation, or neural network methods, and classifies the evaluation results into different levels based on the comprehensive evaluation index; wherein, the multi-dimensional quantitative evaluation indicators include pressure relief effect, crack effectiveness, and achievement of engineering objectives.

7. The system according to claim 1, characterized in that, In the intelligent analysis layer, the early warning analysis module uses time series prediction models, including ARIMA, LSTM, or random forest, to predict the trends of pressure, flow, and crack propagation paths; and based on the prediction results and threshold conditions, it triggers a multi-level alarm mechanism to reduce false alarms and duplicate alarms through association rules and causal analysis.

8. The system according to claim 1, characterized in that, The system also includes a visualization layer that uses web-based multi-rendering to visualize data and calculation results, and supports automatic report generation.

9. The system according to claim 1, characterized in that, The system also includes an application layer, which provides application entry points for real-time monitoring, effect evaluation, early warning management, report generation, and system settings.

10. A method for evaluating, monitoring, and early warning of the effect of surface hydraulic fracturing in deep mines, characterized in that, This method uses the system described in any one of claims 1–9 to evaluate and monitor the effectiveness of hydraulic fracturing on the surface of deep mines.

Images