Mine closed goaf pollution release early warning method, system and equipment based on water quality and water level monitoring

Abstract

The application discloses a mine area closed goaf pollution release early warning method based on water quality and water level monitoring, and belongs to the underground water monitoring field. The method comprises the following steps: collecting water level data and representative water samples of water-filled rock layers, performing water chemical and isotope analysis on the water samples, analyzing the direct contribution water volume of each water-filled rock layer, inputting a data set containing dynamic monitoring parameters and future expected parameters into a pre-trained water level prediction model to obtain a predicted water level, calculating the expected exchange flow between the closed goaf and the water-filled rock layer by using the Darcy seepage formula based on the predicted water level and the goaf shape, and calculating the expected pollutant discharge amount in the prediction period. If the expected pollutant discharge amount exceeds a preset safety threshold, a warning for the corresponding closed goaf is generated, so that the early warning of the pollutant discharge of the closed goaf is realized, and timely and scientific decision-making basis is provided for the active prevention and control and precise management and control of the mine area underground water pollution risk.

Description

Methods, systems, and equipment for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring. Technical Field

[0001] This invention relates to the field of groundwater monitoring. More specifically, this invention relates to a method, system, and equipment for early warning of pollution release from closed mining subsidence areas based on water quality and water level monitoring. Background Technology

[0002] After long-term, large-scale mining, a considerable number of closed or abandoned mines have accumulated in many mining areas. The underground chambers and tunnel systems of these abandoned mines are often slowly filled with groundwater after production ceases, forming a special groundwater body with a complex water-rock reaction environment. In this environment, the mine water continuously reacts with residual coal-bearing minerals and metal sulfides, leading to a gradual deterioration of the mine water quality and the accumulation of high concentrations of sulfates, heavy metal ions, and characteristic pollutants, becoming a potential source of pollution.

[0003] The roof strata of abandoned goafs contain well-developed water-conducting fracture zones, establishing a hydraulic connection between them and water-bearing strata. The highly polluted water bodies accumulated in abandoned mines engage in continuous, head-driven hydraulic exchange with surrounding water-bearing strata used for water supply or ecological maintenance. Studies show that dynamic factors such as short-term heavy rainfall in the mining area, periodic changes in drainage volume from surrounding operating mines, or regular extraction from centralized water sources can all cause fluctuations in the water levels of the relevant water-bearing strata. These fluctuations alter the hydraulic gradient between the abandoned goaf and the water-bearing strata, thus affecting the volume and rate of water exchange between the closed goaf and the water-bearing strata.

[0004] The migration and diffusion of pollutants exhibit a significant lag. There is a time window between changes in hydrodynamic conditions and the transport of the pollutant plume within the aquifer and its capture by downstream monitoring points. Current technologies primarily rely on periodic sampling and monitoring of groundwater quality. By the time monitoring data confirms pollution, harmful substances have often already entered the aquifer and diffused, and pollution has already occurred. At this point, emergency response is costly and has limited effectiveness.

[0005] In addition, there may still be operating mines in the mining area, and their drainage activities will continue to disturb the regional flow field. Residents' lives and agricultural production depend on the extraction and utilization of groundwater resources. These activities are superimposed on the natural hydrological processes of abandoned mines, making the hydrodynamic and water quality evolution of the entire mining area's groundwater system extremely complex.

[0006] In summary, there is currently a lack of effective prediction and early warning methods to address the persistent groundwater pollution risk caused by the existence of closed mining areas. There is an urgent need to invent a method that can integrate multi-source monitoring data, simulate pollutant release flux driven by hydrodynamics, and achieve early risk warning in order to fill the technological gap in the current ability to proactively prevent and control groundwater environmental risks in mining areas. Summary of the Invention

[0007] One objective of this invention is to provide a method, system, and equipment for early warning of pollution release from closed mining areas based on water quality and water level monitoring. By integrating multi-source dynamic monitoring data such as water quality and water level, and using a trained neural network model to predict future water head fields, the expected pollutant exchange flux between the closed mining area and the water-bearing rock strata can be quantitatively calculated in advance. This achieves a shift from "post-event discovery" to "pre-event warning," providing timely and scientific decision-making basis for proactive prevention and precise management of groundwater pollution risks in mining areas.

[0008] To achieve these objectives and other advantages according to the present invention, according to one aspect of the present invention, a method for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring is provided, comprising the following steps:

[0009] S1. Locate and monitor all closed goaf areas within the monitoring area that are hydraulically connected to the water-bearing rock strata, collect water level data for each water-bearing rock strata, collect representative water samples within the monitoring area, including: background water samples for each water-bearing rock strata, water samples for each closed goaf area, and mixed water samples for all groundwater discharge bodies, and monitor the flow rate of all groundwater discharge bodies.

[0010] S2. Perform water source analysis on the representative water samples to analyze and obtain the direct water contribution of each water-bearing rock layer to the groundwater discharge bodies in the current monitoring period;

[0011] S3. Construct a first dataset and a second dataset. The first dataset includes the direct water contribution of each water-bearing rock layer, the duration of the period, the monitored flow rate of each groundwater discharge body, atmospheric precipitation, meteorological parameters, and the beginning and end water level data of each water-bearing rock layer during the current monitoring period. The second dataset includes the expected precipitation, expected meteorological parameters, the duration of the prediction period, and the expected flow rate of each groundwater discharge body during the target prediction period.

[0012] S4. Input the first dataset and the second dataset into the water level prediction model, and output the predicted water level data of each water-filled rock layer at the end of the prediction period. The water level prediction model is established and trained based on a neural network.

[0013] S5. For the target closed goaf, based on the predicted water level data of the corresponding water-bearing rock layer, the shape characteristics of the closed goaf, the vertical equivalent seepage path length of the closed goaf, and the vertical equivalent permeability coefficient, the Darcy seepage formula is used to calculate the expected exchange flow rate of the target closed goaf when the associated water-bearing rock layer reaches the predicted water level data.

[0014] S6. Based on the pollutant concentration analysis results and expected exchange flow of water samples from the closed goaf, calculate the expected pollutant discharge of the closed goaf within the prediction period. If the expected pollutant discharge exceeds the preset safety threshold, generate a pollution risk warning for the closed goaf.

[0015] Preferably, water level data for each water-bearing rock layer is collected by deploying water level monitoring wells relative to each water-bearing rock layer; in step S3, the water level data at the beginning and end of the cycle are the initial water level value collected uniformly at the beginning of the current monitoring cycle and the final water level value collected uniformly at the end of the current monitoring cycle by each water level monitoring well corresponding to the water-bearing rock layer; in step S4, the predicted water level data is the water level height of each water level monitoring well at the end of the predicted cycle.

[0016] Preferably, step S5 includes the following steps:

[0017] S51. Based on the predicted water level data of each water level monitoring well, the predicted head height distribution field of each water-bearing rock layer in the prediction period is generated by spatial interpolation.

[0018] S52. Based on the predicted head height distribution field, determine the water inflow section and water outflow section of the water-bearing strata, and calculate the predicted head difference between the two.

[0019] S53. Calculate the expected exchange flow rate between the target closed goaf and the water-bearing rock strata. The calculation formula is as follows:

[0020]

[0021] Where Q is the expected exchange flow rate, in m³ / d; B is the width of the water-bearing stratum affected by the goaf, in m; L is the length of the water-bearing stratum affected by the goaf, in m; ΔH is the predicted hydraulic head difference, in m; h i is the vertical equivalent seepage path length, in meters; K is the vertical equivalent permeability coefficient, in meters per day.

[0022] Preferably, step S2 includes the following steps:

[0023] S21 Perform stable isotope analysis on the representative water samples collected in step S1, and determine and obtain the stable hydrogen and oxygen isotope composition of the background water samples and mixed water samples of each water-bearing rock layer.

[0024] S22. Using the stable hydrogen and oxygen isotope composition of the background water sample from each water-bearing rock layer as the end-member value and the stable hydrogen and oxygen isotope composition of each mixed water sample as the mixing value, input the mixture into the isotope mixing model for calculation to obtain the proportion of water contribution of each water-bearing rock layer to each groundwater discharge body.

[0025] S23. For each water-bearing rock stratum, multiply its contribution ratio to a certain groundwater discharge body by the total monitored flow rate of that discharge body in the current monitoring period to obtain the contribution of the water-bearing rock stratum to the groundwater discharge body. Summarize the total contribution of the water-bearing rock stratum as its direct contribution in the current monitoring period.

[0026] Preferably, the water level prediction model is a BP neural network model optimized by particle swarm optimization algorithm, employing a three-layer feedforward network, including: an input layer, at least one hidden layer, and an output layer; wherein, the input layer has multiple input neuron nodes, the hidden layer has multiple hidden neuron nodes, and each input neuron node is connected to each of the hidden neuron nodes; the hidden neuron nodes are all connected to the output layer; the construction and training process of the water level prediction model is as follows: a first dataset containing multiple historical monitoring periods and a corresponding second dataset are obtained, and all parameters from both are input into the input neuron nodes, using the predicted water level data corresponding to each historical monitoring period as the training target; the initial weights and thresholds of the BP neural network model are globally optimized using particle swarm optimization algorithm, and the input neuron nodes, hidden neuron nodes, and output layer are iteratively trained using error backpropagation algorithm; when the error index of the output result reaches the preset requirement, the training is completed, and the water level prediction model is obtained.

[0027] Preferably, the groundwater discharge bodies include mine drainage water and groundwater source intake within the monitoring area. The expected flow rates of each groundwater discharge body in the second data set include the expected drainage flow rate of each mine and the planned pumping volume of each water source intake point. The formula for calculating the expected drainage flow rate is as follows:

[0028]

[0029]

[0030] Where Q0 is the expected drainage flow rate of the mine, in m³ / d; K0 is the seepage coefficient of the mine, in m / d; S is the drawdown, in m; M is the average thickness of the water-bearing strata corresponding to the mine, in m; R0 is the radius of influence of the mine, in m; and r0 is the equivalent radius of the mine, in m.

[0031] Preferably, the method for calculating the expected pollutant discharge includes the following steps: S61, based on the water level data of the water-bearing rock strata collected in the current monitoring cycle, calculate the instantaneous exchange flow rate Q' of the target closed goaf at the end of the current monitoring cycle, and calculate the expected exchange flow rate Q at the end of the prediction cycle based on the predicted water level data, and take the arithmetic mean of Q' and Q as the expected average exchange flow rate Q. avg ;

[0032] S62. Based on the pollutant concentration analysis results of water samples from the target closed goaf area, determine the average concentration C of the target pollutant in the closed goaf area during the current monitoring period; the formula for calculating the expected pollutant discharge is: , where T is the total duration of the prediction period.

[0033] A pollution release early warning system for closed goaf areas in mining areas is provided for performing the above-mentioned methods, including: a data acquisition module, a data processing module, and an early warning module;

[0034] The data acquisition module collects water level data of each water-bearing rock layer through deployed water level monitoring wells, collects representative water samples within the monitoring area, and monitors the flow rate of all groundwater discharge bodies.

[0035] The data processing module, connected to the data acquisition module, is used to process and predict the acquired data. It includes a water source analysis submodule, which is used to analyze the acquired representative water samples and obtain the direct contribution of each water-bearing rock layer.

[0036] The dataset construction submodule is used to construct the first dataset and the second dataset;

[0037] The water level prediction submodule is used to output the predicted water level data of each water-filled rock layer at the end of the prediction period using the trained water level prediction model, and to calculate the expected exchange flow of the target closed goaf area.

[0038] The pollutant calculation module is connected to the data processing module and obtains the expected exchange flow rate of the target closed goaf and the pollutant concentration of the goaf water sample, and calculates the expected pollutant discharge of the target closed goaf within the prediction period.

[0039] The early warning module is connected to the pollutant calculation module and is used to compare the calculated expected pollutant discharge amount with the preset safety threshold. If the threshold is exceeded, a pollution risk early warning information for the closed goaf area is generated and released.

[0040] An electronic device is provided, including a memory and a processor. The memory stores computer instructions, and the processor executes the computer instructions to perform the above-described method for early warning of pollution release from closed mining areas based on water quality and water level monitoring.

[0041] The present invention has at least the following beneficial effects:

[0042] First, this invention establishes a pollutant release early warning method based on hydrodynamic feedforward, realizing the forward shift of risk prevention. In response to the current lagging monitoring model of "pollute first, monitor later", this invention predicts future water levels and calculates expected exchange flow to assess pollutant discharge, realizing feedforward early warning for pollution sources. It can issue early warnings to water resource management departments before pollutants actually enter the aquifer and spread in closed goaf areas, providing sufficient response time. It is particularly suitable for old mining areas with dense historical mining activities and a large number of closed goaf areas.

[0043] Secondly, by collaboratively analyzing water quality monitoring data and water quantity monitoring data, this invention achieves accurate quantification of pollution in each closed mining area, solving the technical problem of unclear pollution source analysis caused by the fragmentation of water quality and water quantity data in traditional methods. It is applicable to complex mining areas where it is necessary to clearly distinguish the responsibility for mixed pollution from multiple water sources.

[0044] Third, by introducing a neural network model to predict future water levels, a dynamic and reliable prediction of the core driving force of pollution release (hydraulic head difference) is achieved, solving the technical problem of high prediction difficulty and poor timeliness of traditional mechanistic models when dealing with the nonlinear interplay of multiple factors such as meteorology and mining. This method is suitable for early warning scenarios in mining areas with complex hydrodynamic conditions and interference from various dynamic factors.

[0045] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0046] Figure 1 is a schematic diagram of a method flow of a technical solution of the present invention;

[0047] Figure 2 is a schematic diagram of the groundwater migration calculation principle model in one technical solution of the present invention. Detailed Implementation

[0048] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0049] It should be noted that, unless otherwise specified, the experimental methods described in the following embodiments are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified. In the description of this invention, the orientation or positional relationship indicated by the terms is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing this invention and simplifying the description. It does not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0050] As shown in Figure 1, this invention provides a method for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring, comprising the following steps:

[0051] S1. Locate and monitor all closed goaf areas within the monitoring area that are hydraulically connected to the water-bearing rock strata, collect water level data for each water-bearing rock strata, collect representative water samples within the monitoring area, including: background water samples from each water-bearing rock strata, water samples from each closed goaf area, and mixed water samples from all groundwater discharge bodies, and monitor the flow rate of all groundwater discharge bodies.

[0052] Specifically, by systematically collecting and analyzing historical geological exploration reports, mine operation plans, roadway layout maps, and mine closure reports, the spatial distribution of all closed goaf areas (i.e., coal seam areas that have been historically mined out and abandoned) is delineated. Based on historical mining reports of closed goaf areas, water-bearing strata with hydraulic connections are identified. When necessary, geophysical exploration methods (such as high-density electrical resistivity tomography and transient electromagnetic methods) are used to investigate goaf areas where historical mining reports are unclear, in order to determine whether there is a hydraulic connection between them and water-bearing strata. Finally, a list and spatial database of all identified closed goaf areas with hydraulic exchange risks are established.

[0053] Regarding the acquisition of water level data for each water-bearing stratum, the first step should be to fully collect and utilize existing hydrological observation wells or water supply wells within the monitoring area whose filter pipe sections accurately correspond to the target water-bearing stratum to obtain historical and real-time water level data. In areas where existing well distribution is sparse or spatially representative, new dedicated water level monitoring wells need to be deployed based on hydrogeological conditions. New well locations should ensure that the filter pipe sections accurately correspond to the target water-bearing stratum to obtain the true hydraulic head value for that stratum. Data acquisition can utilize automated pressure sensors for continuous monitoring or manual measurement at prescribed frequencies. All water level data must be uniformly converted to a reference elevation and recorded and stored in the form of hydraulic head values.

[0054] Representative water samples include background water samples from aquifer strata, collected from water level monitoring wells located in the natural recharge area of ​​the aquifer strata or upstream of it, unaffected by goaf pollution. These samples characterize the original hydrochemical field of the aquifer strata. Water samples from closed goaf areas can be collected through existing boreholes, or, under safe conditions, through dedicated drainage boreholes leading to the goaf's water accumulation area. These samples reflect the water quality after long-term water-rock reaction within the pollution source. Mixed water samples from groundwater discharge bodies must be collected within the monitoring period at all locations within the monitoring area that centrally discharge groundwater to the surface or extract water for production and domestic use. This includes drainage outlets of operating mines and pump outlets of centralized groundwater sources. Mixed water samples represent the combined water quality resulting from various natural or engineering mixing processes. All sampling points must be uniformly coded, and the water sampling process must strictly adhere to the "Technical Specifications for Groundwater Environmental Monitoring," using inert material containers and adding necessary protective agents to ensure the representativeness and authenticity of the water samples.

[0055] In flow monitoring, reliable metering facilities should be installed according to the type of discharge point. For open channel drainage, such as mine drainage ditches, standard thin-walled weirs or Parshall flumes can be installed, along with water level sensors to measure the flow rate. For pipeline drainage, such as pump outlet pipes, electromagnetic or ultrasonic flow meters must be installed. All metering equipment must be calibrated regularly, and flow data should be continuously recorded or accumulated according to monitoring cycles, with units uniformly expressed as cubic meters per day or cubic meters per cycle. This data should be strictly synchronized with water quality and water level monitoring in time, thereby providing an accurate basis for quantifying water flow and subsequent pollution load calculations.

[0056] S2. Perform source water analysis on the representative water samples to determine the direct contribution of each aquifer to the groundwater discharge bodies during the current monitoring period. Specifically, due to differences in lithology, burial conditions, and hydrogeochemical processes experienced by different aquifers, their groundwater typically possesses unique hydrochemical or isotopic "fingerprint" characteristics. The water quality of a groundwater discharge body can be considered as the result of a mixture of its various recharge sources in a certain proportion. By analyzing the fingerprint characteristics of background water samples from each aquifer and the characteristics of mixed water samples from various groundwater discharge bodies, and using a mass balance mixing model for inverse calculation, the proportion of water from each aquifer in the mixed water body can be determined, i.e., the water contribution ratio. Multiplying this ratio by the total flow rate of the discharge body during the monitoring period yields the water contribution of that aquifer to a specific discharge body. Characteristic indicators can be conventional hydrochemical parameters such as major ionic composition and trace components, or they can be more stable hydrogen and oxygen stable isotope compositions with stronger tracer capabilities. Analytical calculations can be completed using a variety of mature hybrid models.

[0057] In one technical solution, step S2 includes the following steps:

[0058] S21 involves performing stable isotope analysis on representative water samples collected in step S1 to determine and obtain the stable hydrogen and oxygen isotope composition of background water samples from each aquifer and mixed water samples. Specifically, the background water samples from each aquifer and mixed water samples from various groundwater discharge bodies in step S1 are sent to a qualified laboratory for analysis. The stable hydrogen and oxygen isotope analysis involves determining the δ¹⁸O⁻ content in the water samples. 2 H and δ 18 The ratio of O. Determination requires high-precision instruments, such as wavelength scanning cavity attenuation spectroscopy, and calibration using international standard reference materials to obtain a set of highly identifiable fingerprint data for each background and mixed water sample.

[0059] S22. Using the stable hydrogen and oxygen isotope compositions of background water samples from each aquifer as endmember values ​​and the stable hydrogen and oxygen isotope compositions of each mixed water sample as mixing values, the results are input into an isotope mixing model for calculation. This yields the contribution ratio of each aquifer to the water volume of each groundwater discharge body. Specifically, the fingerprint data obtained in S21 is processed. For each groundwater discharge body requiring analysis, the δ¹⁸O₁⁻ of its mixed water sample is calculated. 2 H and δ 18 The O value is used as the mixing value. Simultaneously, the isotopic composition of all possible water-bearing strata background water samples that could replenish it is used as potential endmember values. These data are then input into an isotopic mixing model for computation. Taking the IsoSource model, currently a mass balance mixing model, as an example, its basic principle is to iterate through all possible endmember combinations within a set incremental step size (e.g., 1%), selecting those combinations where the difference between the calculated and measured mixing values ​​is less than a preset tolerance, and statistically analyzing the contribution ratio distribution of each endmember in these feasible combinations, finally outputting its average value or feasible range. The reliability of this analysis can be quantitatively assessed by calculating statistical indicators of the model fitting results, such as root mean square error.

[0060] S23. For each aquifer, multiply its contribution ratio to a specific groundwater discharge body by the total monitored flow rate of that discharge body during the current monitoring period to obtain the contribution of that aquifer to the groundwater discharge body. Summarize all the contributions of each aquifer as its direct contribution during the current monitoring period. Specifically, extract the total discharge or extraction volume of each groundwater discharge body within the corresponding monitoring period from the flow monitoring records. Multiply the contribution ratio of each aquifer calculated in S22 by the total flow rate of the corresponding discharge body to obtain the direct contribution of that aquifer to a single discharge point. Finally, sum the contributions of an aquifer to all discharge points; the total sum is the direct contribution of that aquifer through all discharge pathways during the current monitoring period. This data quantitatively characterizes the output flux of each aquifer in the regional water cycle, providing dynamic water input parameters for subsequent water level prediction models.

[0061] Water source apportionment is not necessarily limited to isotopic methods. Under specific conditions, other mature hydrogeochemical methods such as water chemical ion ratio analysis and inert gas tracing can be used for mixed calculations, or multiple methods can be combined to improve the reliability of the apportionment.

[0062] S3. Construct a first dataset and a second dataset. The first dataset includes the direct water contribution from each water-bearing stratum, the duration of the monitoring period, the monitored flow rate of groundwater discharge from each stratum, atmospheric precipitation, meteorological parameters, and the beginning and end water level data of each water-bearing stratum within the current monitoring period. The second dataset includes the expected precipitation, expected meteorological parameters, the duration of the prediction period, and the expected flow rate of groundwater discharge from each stratum within the target prediction period. The beginning and end water level data are the initial water level values ​​collected uniformly at the beginning of the current monitoring period and the final water level values ​​collected uniformly at the end of the current monitoring period from each water level monitoring well corresponding to the water-bearing stratum. The first dataset is used for training the subsequent water level prediction model and learning historical patterns, while the second dataset represents future scenarios and is used to drive the water level prediction model to make predictions. The parameters contained in these two datasets have a direct driving or indicative effect on the water level changes in the water-bearing strata.

[0063] Atmospheric precipitation is the most fundamental recharge source for groundwater systems, and its intensity, duration, and spatial distribution directly determine the magnitude of infiltration recharge flux. The direct water contribution from each aquifer and the monitored flow rates of groundwater discharge bodies accurately quantify the water output from aquifers through natural runoff and human intervention. Meteorological parameters such as temperature, humidity, wind speed, and evaporation, while not directly constituting water input or output, influence the amount of net recharge and are key environmental factors driving seasonal and interannual variations in water levels. The periodic start and end water level data for each aquifer characterize the change in water level over a specific period and are the core target variables that the model needs to learn and predict.

[0064] The expected data in the second dataset can be obtained from conventional and reliable planning and forecasting information sources in this field. For example, expected precipitation and expected meteorological parameter vectors can be obtained by referring to authoritative meteorological forecasts issued by local meteorological departments for the target forecast period. In cases where there are no real-time forecasts, historical data from the same period can be analyzed to make reasonable approximations.

[0065] The expected flow rate of groundwater discharge includes the expected drainage flow rate of each mine and the planned pumping volume of each water intake point. The planned pumping volume is based on the water use plan approved by the water management department or operating unit. These data are all planned indicators in the routine management of mining areas and water resources. For the expected drainage flow rate of operating mines, it can be estimated based on the drainage and drawdown targets determined in the mine production plan, combined with the hydrogeological parameters of the mining area, using a mature steady-flow well flow formula. The specific calculation formula is as follows:

[0066]

[0067]

[0068] Wherein, Q0 is the expected drainage flow rate of the mine, in m³ / d; K0 is the mine seepage coefficient, in m / d; S is the drawdown, in m; M is the average thickness of the corresponding water-bearing strata of the mine, in m; R0 is the radius of influence of the mine, in m; r0 is the equivalent radius of the mine, in m; and lg is the logarithmic function operator. The formula for calculating the radius of influence R0 of the mine adopts a simplified expression of a steady-state well flow model for rapid calculation. This expression is one of the commonly used empirical methods for estimating the radius of influence of unconfined or confined aquifers. In practical applications, users only need to directly substitute S, K0, and r0 with the above units to calculate the radius of influence R0 of the mine. All parameters are derived from the actual geological reports, design documents, and production plans of the mine. When implementing this early warning method, all are known quantities or planned values ​​that can be clearly obtained or calculated, and do not rely on real-time monitoring within the prediction period.

[0069] S4. Input the first dataset and the second dataset into the water level prediction model, and output the predicted water level data of each water-bearing stratum at the end of the prediction period. The water level prediction model is established and trained based on a neural network, and the predicted water level data is the water level height of each monitoring well at the end of the prediction period. Specifically, due to the inherent complexity of the groundwater system in the mining area, the change in water level of the water-bearing stratum is affected by a combination of factors, including meteorological conditions, human activities, and the water balance of the aquifer itself. These factors and the water level are difficult to accurately describe using traditional linear or simple physical models. Feedforward neural networks have strong nonlinear mapping capabilities and adaptive learning capabilities for high-dimensional features. They can automatically mine and learn the implicit, complex patterns and laws between multi-source driving factors and water level changes from historical data through training, without needing to pre-define the physical equation form.

[0070] The prediction period can be flexibly set according to the needs of early warning management, and can be as short as several days or weeks. For example, it can be used to address scenarios that may cause rapid disturbances to the regional hydrodynamic field, such as heavy rainfall, changes in agricultural pumping volume, or the commencement of drainage from newly built mines in the coming weeks. This design allows the model to focus more on learning the response relationship between key driving factors such as meteorology and artificial drainage and water level changes. The first and second datasets input to the model integrate key macroscopic factors affecting water levels. The nonlinear mapping capability of the neural network enables it to learn the dominant changing patterns even with a reasonable degree of prediction error in the input parameters, and output predicted water level data that meets the accuracy requirements of subsequent pollution risk trend assessment.

[0071] In one technical solution, the water level prediction model is a BP neural network model optimized based on the particle swarm optimization algorithm. It employs a three-layer feedforward network, including an input layer, at least one hidden layer, and an output layer. The input layer has multiple input neurons, and the hidden layer has multiple hidden neurons, with each input neuron connected to all the hidden neurons. All hidden neurons are connected to the output layer. The particle swarm optimization algorithm is a swarm intelligence optimization algorithm that simulates the social cooperative foraging behavior of flocks of birds or schools of fish. Its core principle is to guide the search process through the historical optimal experience of individuals and the group, thereby efficiently finding the global optimum in the solution space. In this technical solution, the PSO algorithm is used to optimize the initial connection weights and neuron thresholds of the BP neural network to overcome the shortcomings of traditional BP neural networks, such as slow training speed and convergence to local optima due to random initialization. The water level prediction model can be implemented through programming. MATLAB and its neural network toolbox and optimization toolbox, or the Python language combined with optimization algorithm libraries such as PySwarm and DEAP, as well as deep learning frameworks such as TensorFlow and PyTorch, can all be used to easily realize the coupled construction, training and application of PSO algorithm and BP neural network.

[0072] For example, the construction and training process of a BP neural network model optimized using the particle swarm optimization algorithm is as follows:

[0073] A1. Construct a sample set for model learning and evaluation from continuous historical monitoring records. The sample set consists of a series of consecutive samples arranged chronologically. Each sample is constructed by selecting two adjacent and consecutive historical monitoring periods, denoted as the t-th monitoring period and the (t+1)-th monitoring period, respectively.

[0074] For a given sample, all available information is integrated into an input feature vector X. t X t It consists of a first dataset and a second dataset.

[0075] The parameters of the first dataset include:

[0076] (a) Direct water contribution from each water-bearing rock layer , representing the direct water contribution of the j-th aquifer within the t-th monitoring period. This can be obtained by extracting water sample monitoring reports from the historical groundwater monitoring database and calculating using source apportionment methods.

[0077] (b) Duration of the period T t , represents the time length of the t-th monitoring period, which can be in days and is used to standardize water volume and flow data.

[0078] (c) Monitoring flow rates of groundwater discharge bodies in various regions , representing the mine drainage volume or extraction flow rate monitored for the j-th groundwater discharge body within the t-th monitoring period. This data is sourced from the historical groundwater monitoring database.

[0079] (d) Atmospheric precipitation P t , representing the average daily precipitation or cumulative precipitation during the t-th monitoring period. This data is sourced from a historical meteorological database.

[0080] (e) Meteorological parameter vector, including several key meteorological elements in the t-th monitoring period, such as average temperature, average humidity, and surface evaporation, used to characterize the impact of climate on replenishment and consumption. This data is sourced from a historical meteorological database.

[0081] (f) The periodic start and end water level data of each water-bearing rock layer can be a vector, including the initial water level value at the beginning of the t-th monitoring cycle and the last water level value at the end of the t-th monitoring cycle, defining the change result of the water-bearing rock layer in the t-th monitoring cycle. This data comes from the groundwater historical monitoring database.

[0082] The (t+1)th monitoring period serves as the prediction period when the sample is established. The second dataset parameters describe the known external driving forces in the (t+1)th monitoring period, including:

[0083] (a) Expected precipitation P t+1 , representing the average daily precipitation or cumulative precipitation during the (t+1)th monitoring period, uses measured data from the historical meteorological database during the sample training phase.

[0084] (b) Expected meteorological parameter vector, including several key meteorological elements in the (t+1)th monitoring period, such as average temperature, average humidity, and surface evaporation, using measured data from the historical meteorological database during the sample training phase.

[0085] (c) Predicted period duration T t+1 , represents the time length of the (t+1)th monitoring period.

[0086] (d) Expected flow rates of groundwater discharge bodies in various regions , represents the amount of mine drainage or extraction flow rate for the j-th groundwater discharge body in the (t+1)-th monitoring period, which is based on measured data from the historical groundwater monitoring database during the sample training phase.

[0087] The corresponding training target value Y t+1 The vector is formed by the water level height of each monitoring well at the end of the (t+1)th monitoring cycle.

[0088] After randomly shuffling all samples, the training set and validation set are divided according to a preset ratio of 7:3 or 8:2. The maximum and minimum values ​​of each feature are calculated based on the training set, and the normalization formula is applied to process all data.

[0089] A2. Construct a three-layer feedforward neural network consisting of an input layer, a single hidden layer, and an output layer, and initialize the mapping relationship between the population particles and the weights and thresholds of the BP neural network, including hyperparameters such as particle dimension, initial velocity, population size, learning factor, and inertia weight.

[0090] Input layer number of nodes n in The dimension of the input feature vector is equal to the dimension of the input feature vector. This vector consists of all parameters from the first and second datasets, specifically including: the direct water contribution from each aquifer, the duration of the monitoring period, the monitored flow rate of each groundwater discharge body, atmospheric precipitation, meteorological parameters (such as temperature, humidity, and evaporation), the beginning and end water level data of each aquifer, and the expected precipitation, expected meteorological parameters, predicted period duration, and expected flow rate of each groundwater discharge body from the second dataset. The input layer is responsible for receiving the standardized multi-source driving data and passing it forward to the hidden layer.

[0091] Number of hidden layer nodes n hid Based on Kolmogorov's theorem and related empirical formulas, the network can be fine-tuned during actual training to ensure that it has sufficient nonlinear expressive power while avoiding overfitting. The Sigmoid function is chosen as the activation function for the hidden layers, which is suitable for continuous regression problems such as water level prediction.

[0092] Number of output layer nodes n out It equals the total number of water level monitoring wells deployed in each water-bearing rock layer within the monitoring area. Each node corresponds to the predicted water level data of one water level monitoring well at the end of the prediction period. The output layer uses a linear activation function to avoid distorting the prediction range due to the limitations of the activation function.

[0093] All weights and thresholds in the neural network are randomly initialized in a uniform distribution within the interval [−0.5, 0.5]. All weights and thresholds to be optimized in the neural network are sequentially concatenated into a real-valued vector, which forms the position vector of each particle in the particle swarm optimization algorithm. The dimension of this vector is completely determined by the network structure, and its value is the sum of the number of weights from the input layer to the hidden layer, the number of weights from the hidden layer to the output layer, and the number of thresholds from the hidden layer to the output layer. Correspondingly, the particle velocity vector is also initialized in the same dimensional space, typically set to zero or a small random value close to zero, so that it can be controlled and adjusted during iteration.

[0094] Regarding the hyperparameter settings of the particle swarm optimization algorithm, considering both search efficiency and convergence reliability, the population size is set to 30 particles. Both the individual learning factor and the social learning factor are assigned a value of 2.0 to ensure that each particle can learn a balance between its own historical best experience and the globally optimal experience shared by the group. The inertia weight adopts a dynamic decay strategy, with an initial value of 0.9, which linearly decreases to 0.4 as the iteration progresses. This design enables the algorithm to maintain strong global exploration capabilities in the early stages of optimization, avoiding getting trapped in local optima; and enhances local exploitation capabilities in the later stages of optimization, achieving a fine-grained search for the optimal solution region.

[0095] A3. Perform iterative search of the particle swarm optimization algorithm to obtain global initial values ​​for the neural network weights and thresholds. Specifically, each generation of iterative search includes fitness evaluation, particle position and velocity update, and convergence judgment.

[0096] Fitness evaluation decodes the position vector of each particle into corresponding neural network weights and thresholds, and uses the training set data for forward propagation to calculate the predicted water level data for all water level monitoring wells. The reciprocal of the root mean square error between the predicted and actual water levels is used as the fitness value of the particle. The higher the fitness, the better the prediction accuracy of the network under that set of parameters.

[0097] After fitness evaluation, particle position and velocity are updated, and particle states are adjusted accordingly. The velocity-position update formula is:

[0098]

[0099]

[0100] in and Let r1 and r2 represent the velocity and position of particle i in the d-th dimension at the k-th iteration, respectively. r1 and r2 are random numbers in the range [0,1], c1 is the individual learning factor, and c2 is the social learning factor.

[0101] The iteration process continues until the iteration convergence condition is met. The iteration termination condition can be set to meet one of the following convergence conditions: reaching the preset maximum number of iterations or the improvement of the global optimal fitness of the population within 20 consecutive generations is less than a threshold.

[0102] When the iteration terminates, the globally optimal position vector is output. As the optimal initial parameter set for a neural network.

[0103] A4. Using the global optimization initial value as the training starting point for all weights and thresholds of the neural network, the backpropagation algorithm is used for supervised iterative training to finally obtain the water level prediction model.

[0104] Specifically, the normalized training samples are input into the feature vector X. k The data is input into a neural network, where it undergoes a Sigmoid nonlinear transformation sequentially through the input layer and hidden layers. Finally, the predicted water level for each monitoring well is generated at the output layer. Calculate and predict water level Compared with the actual observed water level Y k The error between the input and output layers is calculated, and the error signal is propagated backward from the output layer to the input layer in the reverse order of forward propagation. During this process, the gradient of the error with respect to each weight and threshold is calculated layer by layer using the chain rule.

[0105] Finally, the network parameters are updated based on gradient information. Momentum-driven gradient descent is used to iteratively adjust all weights and thresholds based on the calculated gradients. The learning rate controls the update step size, and the momentum term accelerates convergence and enhances stability.

[0106] The above process is repeated on the training set, and the model performance is monitored on an independent validation set. When the validation error no longer decreases in multiple consecutive iterations, an early stopping mechanism is triggered to terminate training, and the network parameters with the best validation performance at this point are saved as the completed water level prediction model.

[0107] A5. Perform performance verification and evaluation on the trained water level prediction model to confirm that it meets the accuracy and reliability requirements for engineering applications.

[0108] The trained neural network model is ultimately evaluated using an independent test set reserved in A1 that was not used for training and validation. The evaluation employs a set of quantitative metrics, with core metrics including root mean square error (RMSE), Nash efficiency coefficient, and mean absolute error (MAE). Model performance must meet preset engineering early warning standards, such as an RMSE of no more than 0.05 meters for water level prediction and a Nash efficiency coefficient of no less than 0.85. Only models that pass this validation are considered qualified and can be used in subsequent early warning processes.

[0109] S5. For the target closed goaf, based on the predicted water level data of the corresponding water-bearing strata, the shape characteristics of the closed goaf, the vertical equivalent seepage path length of the closed goaf, and the vertical equivalent permeability coefficient, the expected exchange flow rate of the target closed goaf when the associated water-bearing strata reach the predicted water level data is calculated using the Darcy flow formula. Specifically, the closed goaf and the overlying water-bearing strata are connected by a water-conducting fracture zone formed by mining. Driven by the hydraulic head difference formed by natural or artificial disturbance, groundwater in the water-bearing strata will flow into the closed goaf, mix and exchange with the accumulated contaminated water, and then flow out of the closed goaf and into the downstream aquifer. This portion of the contaminated water flowing through the closed goaf is an important influencing factor on the migration and release of pollutants in the water-bearing strata.

[0110] In one technical solution, step S5 includes the following steps:

[0111] S51. Based on the predicted water level data from each water level monitoring well, a predicted head height distribution field for each water-bearing stratum during the prediction period is generated using spatial interpolation. Specifically, the predicted water level corresponding to the discrete water level monitoring wells is transformed into a continuous, spatially distributed head height field. Classical spatial interpolation algorithms such as Kriging interpolation, inverse distance weighted interpolation, or spline function interpolation can be used. These methods can simulate and estimate the head height at any location within the water-bearing stratum at the prediction time based on the known spatial location of the water level monitoring wells and the predicted water level data, combined with geographic statistical laws such as spatial autocorrelation. This ultimately generates a gridded predicted head height distribution field covering the entire monitoring area with a standardized format. This process can be efficiently and systematically completed using the spatial interpolation modules in commercial or open-source GIS software platforms such as ArcGIS, Surfer, and QGIS. Through this step, the predicted water level data output by the water level prediction model is successfully transformed into a predicted head height distribution field with clear spatial distribution characteristics that can be used for hydraulic formula calculations.

[0112] S52. Based on the predicted water head distribution field, determine the water inflow and outflow sections of the water-bearing strata, and calculate the predicted water head difference between them. Specifically, the identification of the water inflow and outflow sections is completed based on a combination of hydrogeological principles and spatial analysis. First, the predicted water head distribution field generated in step S51, the precise boundary map of the target closed goaf, and the distribution map of the water-conducting fracture zone determined according to the geological exploration report or geophysical exploration results are spatially superimposed in the same geographic coordinate system. Based on the migration pattern of groundwater flow from high-water-head areas to low-water-head areas, on the superimposed comprehensive analysis map, the high-water-head area located upstream of the closed goaf spatial range and intersecting with the water-conducting fracture zone is identified as the water inflow section, and the low-water-head area located downstream of the closed goaf spatial range and intersecting with the water-conducting fracture zone is identified as the outflow section.

[0113] Subsequently, from the predicted head height distribution field, the head values ​​of all grid cells or contour nodes in the inflow and outflow sections are extracted respectively, and the difference between the two is calculated as the predicted head difference, which quantitatively reflects the net force driving groundwater to flow into and through the closed goaf at the end of the prediction period.

[0114] S53. Calculate the expected exchange flow rate between the target closed goaf and the water-bearing rock strata. The calculation formula is as follows:

[0115]

[0116] Where Q is the expected exchange flow rate, in m³ / d; B is the width of the water-bearing stratum affected by the goaf, in m; L is the length of the water-bearing stratum affected by the goaf, in m; ΔH is the predicted hydraulic head difference, in m; h i is the vertical equivalent seepage path length, in meters; K is the vertical equivalent permeability coefficient, in meters per day.

[0117] As shown in Figure 2, the path of groundwater flowing through a closed goaf in an aquifer is generalized as follows: starting from the upstream aquifer, it first enters the closed goaf vertically through a water-conducting fracture zone, then flows nearly horizontally for a distance within the goaf's caving zone and tunnel space, and finally re-enters the aquifer vertically through the downstream water-conducting fracture zone. The key resistance controlling the water flow comes from the vertical crossing of the fracture zone. Therefore, the entire process is simplified into an equivalent computational model controlled by vertical seepage.

[0118] The width B and length L of the water-bearing strata affected by the goaf define the extent to which the closed goaf influences the water-bearing strata on a planar surface. These dimensions can be determined from mine maps, geological exploration reports, or geophysical survey results, by measuring the boundary of the target goaf and the hydraulic influence zone it generates within the water-bearing strata. For example, the length L can be determined along a direction parallel to the overall groundwater flow direction, and the width B along a direction perpendicular to this flow direction, based on the water-conducting fracture angle, empirical extension ratio, or historical hydraulic influence analysis. For instance, if a rectangular goaf is 500 meters long and 200 meters wide along the water flow direction, and its hydraulic influence extension is estimated to be approximately 50 meters, then L can be taken as 600 meters (500 meters + 2 × 50 meters) and B as 300 meters (200 meters + 2 × 50 meters).

[0119] Vertical equivalent seepage path length h i The effective vertical development height of the water-conducting fracture zone is equivalent to the vertical distance between the floor of the water-bearing stratum and the roof of the caving zone in the closed goaf below. This parameter needs to be determined by directly obtaining objective geological data from the mining area. In practice, it is necessary to review and analyze the specific geological survey report or hydrogeological exploration report for the target goaf and its surrounding rock, extract key information, and identify the floor depth of the target water-bearing stratum and the roof depth of the caving zone formed by mining on the typical borehole columnar section or comprehensive hydrogeological columnar section attached to the report. The depth difference between the two is h. i ,

[0120] The vertical equivalent permeability coefficient K can be determined by inversion from measured hydrological data during mine production. The calculation formula is as follows:

[0121]

[0122] Where Q1 is the stable water inflow of the closed goaf during production, which is derived from the mine hydrological log; A is the area of ​​the closed goaf, which is derived from the mining design or survey map; and ΔH1 is the original head difference corresponding to the stable water inflow during production.

[0123] S6. Based on the pollutant concentration analysis results and expected exchange flow of water samples from the closed goaf, calculate the expected pollutant discharge from the closed goaf within the prediction period. If the expected pollutant discharge exceeds a preset safety threshold, generate a pollution risk warning for the closed goaf. Specifically, the specific value of the safety threshold needs to be set by the environmental management department or mining enterprise according to the environmental function requirements of the protected water body, based on the promulgated "Groundwater Quality Standard," the control requirements in the project's environmental impact assessment approval documents, or through scientific and compliant methods such as groundwater solute transport model simulation calculations. Specific methods for setting this threshold include the following:

[0124] (a) Based on the pollutant concentration limits specified in standards or specifications such as the Groundwater Quality Standard (GB / T 14848) and the Technical Guidelines for Groundwater Pollution Risk Assessment, and combined with the hydrological parameters of the receiving water body, the pollutant release threshold is converted into an allowable total pollutant release threshold through model calculation.

[0125] (b) Refer to the specific requirements or limits for groundwater pollution prevention and control during the closure period as determined in the Environmental Impact Report and its approval documents for the mining area or similar projects during the development phase.

[0126] (c) Using a groundwater solute transport model, and with the premise of protecting downstream sensitive targets (such as water sources), the maximum allowable pollutant flux at the outlet of the goaf is calculated in reverse, and this is used as a safety threshold.

[0127] Pollution risk early warning information, used as decision support data, is pushed to the mining area environmental management platform and relevant personnel. One or more of the following tiered response actions can be quickly initiated for the target closed goaf:

[0128] Level 1 Response: Conduct a special investigation into the closed mining subsidence area and downstream hydrological pathways identified in the warning, analyze the specific causes leading to increased exchange flow or pollutant concentrations, and preliminarily assess the diffusion trend of the pollution plume. Simultaneously, increase the frequency of water quality monitoring at domestic and agricultural water sources within the affected area to ensure water safety and provide real-time data support for subsequent decision-making.

[0129] Level 2 Response: Optimize the drainage plan and pace of operating mines to reduce disturbance to the regional flow field; maintain or adjust the drainage intensity of mines scheduled to cease operation in order to control the hydraulic gradient in key areas; if necessary, suspend groundwater extraction near the warning area and proactively change the flow field direction to macroscopically weaken the hydrodynamic conditions driving pollutant transport.

[0130] Level 3 Response: For identified pollution release channels or high-risk areas, implement direct engineering intervention. Grout the main water-conducting fissures to reduce the exchange flow at the source; or urgently construct permeable reactive barriers or other emergency treatment facilities along the downstream water flow path of the goaf to intercept and treat released pollutants on-site, controlling the spread of pollution.

[0131] In one technical solution, the method for calculating the expected pollutant emissions includes the following steps:

[0132] S61. Based on the water level data of the water-bearing rock strata collected during the current monitoring period, calculate the instantaneous exchange flow rate Q' of the target closed goaf at the end of the current monitoring period, and calculate the expected exchange flow rate Q at the end of the prediction period based on the predicted water level data. Take the arithmetic mean of Q' and Q as the expected average exchange flow rate Q. avg Specifically, based on the measured water level of the monitoring well at the current moment, the instantaneous head difference of the target closed goaf is calculated, thereby calculating the instantaneous exchange flow rate. The calculation method is consistent with the principle and calculation model of calculating the expected exchange flow rate in step S5. From the perspective of pollutant transport mechanism, for closed goafs where hydraulic connection is dominated by vertical fractures, water exchange can be generalized into a "push flow" mode. The relatively clean groundwater newly entering the goaf will displace and push the original polluted water body downstream. Therefore, the average exchange flow rate Q within the prediction period is used. avg This is used to characterize the average intensity of this displacement effect, thus providing a robust and representative flow estimate for predicting the total amount of polluted water that may be replaced from the goaf throughout the entire prediction period.

[0133] S62. Based on the pollutant concentration analysis results of water samples from the target closed goaf area, determine the average concentration C of the target pollutant in the closed goaf area during the current monitoring period; the formula for calculating the expected pollutant discharge is:

[0134]

[0135] Where T represents the total duration of the prediction period, the average pollutant concentration C is determined through standard laboratory hydrochemical analysis of water samples from the target closed mining area. The analysis process follows national standards such as the "Technical Specifications for Groundwater Environmental Monitoring," selecting appropriate standard analytical methods for concentration determination based on the type of target pollutant predicted for early warning. For example, ion chromatography can be used for sulfate and chloride ions, while inductively coupled plasma mass spectrometry or atomic absorption spectrometry can be used for heavy metal ions such as manganese, iron, and arsenic. For comprehensive indicators such as total dissolved solids and total hardness, corresponding standard physicochemical analysis methods are employed. By statistically analyzing the results of multiple water samples potentially obtained within the same monitoring period, the average concentration C of the target pollutant within that period is calculated.

[0136] A pollution release early warning system for a closed goaf in a mining area, used in the above method, includes: a data acquisition module, a data processing module, and an early warning module;

[0137] The data acquisition module collects water level data from various aquifers through deployed water level monitoring wells, collects representative water samples within the monitoring area, and monitors the flow rate of all groundwater discharge bodies. The data acquisition module can be built using mature products deployed on-site. For example, water level data from various aquifers can be automatically collected using standardized water level monitoring equipment such as submersible pressure sensors or float-type water level gauges; representative water samples can be collected using automatic or manual samplers, and stored and transported according to standardized procedures; flow rate monitoring of groundwater discharge bodies can be achieved using general metering devices such as ultrasonic flow meters, electromagnetic flow meters, or standard measuring weirs and flumes. The raw data collected by the above equipment can be uniformly transmitted to the central data platform via a data acquisition instrument or IoT terminal.

[0138] The data processing module, connected to the data acquisition module, is used to process and predict the acquired data. Internally, it includes a water source analysis submodule, used to analyze representative water samples and obtain the direct water contribution from each aquifer. The data processing module can be implemented based on a computer server and corresponding software system. The water source analysis submodule can integrate or call standardized data output from commercially available laboratory analytical equipment such as ion chromatographs and inductively coupled plasma mass spectrometers, and calculate the water contribution from each aquifer using a built-in hybrid model analytical algorithm (such as IsoSource).

[0139] The dataset construction submodule is used to construct the first dataset and the second dataset. The dataset construction submodule can automatically collect real-time data from the acquisition module, manually input data from the management plan (such as expected pumping volume), and forecast data issued by the meteorological department through the software interface or script, and store them in a formatted manner.

[0140] The water level prediction submodule is used to output the predicted water level data of each water-filled rock layer at the end of the prediction period using a trained water level prediction model, and to calculate the expected exchange flow of the target closed goaf area. The core of the water level prediction submodule is a trained neural network model, which can be packaged into an independent prediction service or computing library. Based on the first and second input datasets, it can quickly output the predicted water level data and the expected exchange flow based on it.

[0141] The pollutant calculation module connects to the data processing module and acquires the expected exchange flow rate and pollutant concentration of the goaf water sample from the target closed goaf. It then calculates the expected pollutant discharge from the target closed goaf within the prediction period. This module can function as a dedicated calculation service or software feature. It receives expected exchange flow rate data from the data processing module and pollutant concentration analysis results from the goaf water sample obtained from the Laboratory Information Management System (LIMS). By executing preset calculation logic, it automatically calculates the expected pollutant discharge from the target closed goaf.

[0142] The early warning module is connected to the pollutant calculation module and is used to compare the calculated expected pollutant discharge amount with a preset safety threshold. If the threshold is exceeded, a pollution risk early warning information is generated and issued for the closed goaf area. The early warning module can be built as a software platform integrating a rule engine and information publishing functions. It automatically compares the received expected pollutant discharge amount with the safety thresholds for different pollutants that are pre-entered into the database. Once the limit is exceeded, the system generates and issues structured pollution risk early warning information to designated managers or departments according to a preset communication protocol through platform interface pop-ups, SMS, email, or dedicated application push notifications. The information usually includes key information such as the goaf area number, pollutants exceeding the standard, predicted discharge amount, and degree of exceeding the threshold.

[0143] An electronic device is characterized by comprising a memory and a processor, wherein the memory stores computer instructions, and the processor executes the computer instructions to perform the aforementioned method for early warning of pollution release in closed mining areas based on water quality and water level monitoring. The electronic device can be any terminal device including mobile phones, laptops, desktop computers, tablets, PDAs (Personal Digital Assistants), POS (Point of Sales), in-vehicle computers, etc.

[0144] Computer instructions are stored in a readable storage medium, such as a computer floppy disk, USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., including several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0145] It should be noted that although the steps are described in a specific order above, this does not mean that they must be performed in that order. In fact, some of these steps can be executed concurrently, or even in a different order, as long as the required functionality is achieved. The number of devices and processing scale described herein are for simplification of the invention; applications, modifications, and variations of this invention will be readily apparent to those skilled in the art.

[0146] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring, characterized in that, Includes the following steps: S1. Locate all closed goaf areas within the monitoring area that are hydraulically connected to the water-bearing rock strata, collect water level data for each water-bearing rock strata, and collect representative water samples within the monitoring area. These representative water samples include: background water samples from each water-bearing rock strata, water samples from each closed goaf area, and mixed water samples from all groundwater discharge bodies. Monitor the flow rate of all groundwater discharge bodies. S2. Perform source analysis on the representative water samples, analyze and obtain the direct contribution of each water-bearing rock strata to the groundwater discharge bodies in the current monitoring period. S3. Construct a first dataset and a second dataset. The first dataset includes the direct contribution of each water-bearing rock strata, the duration of the period, the monitored flow rate of each groundwater discharge body, atmospheric precipitation, meteorological parameters, and the beginning and end water level data of each water-bearing rock strata in the current monitoring period. The second dataset includes the expected precipitation and expected meteorological parameters in the target prediction period. S4. Input the first dataset and the second dataset into the water level prediction model, and output the predicted water level data of each water-bearing rock layer at the end of the prediction period. The water level prediction model is established and trained based on a neural network. S5. For the target closed goaf, based on the predicted water level data of the corresponding water-bearing rock layer, the shape characteristics of the closed goaf, the vertical equivalent seepage path length of the closed goaf, and the vertical equivalent permeability coefficient, calculate the expected exchange flow of the target closed goaf when the associated water-bearing rock layer reaches the predicted water level data using the Darcy seepage formula. S6. Based on the pollutant concentration analysis results of the water sample in the closed goaf and the expected exchange flow, calculate the expected pollutant discharge of the closed goaf within the prediction period. If the expected pollutant discharge exceeds the preset safety threshold, generate a pollution risk warning information for the closed goaf.

2. The method for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring as described in claim 1, characterized in that, Water level data for each water-bearing rock layer is collected by setting up water level monitoring wells relative to each water-bearing rock layer; in step S3, the water level data at the beginning and end of the cycle are the initial water level value collected uniformly at the beginning of the current monitoring cycle and the last water level value collected uniformly at the end of the current monitoring cycle by each water level monitoring well corresponding to the water-bearing rock layer; in step S4, the predicted water level data is the water level height of each water level monitoring well at the end of the predicted cycle.

3. The method for early warning of pollution release from closed mining areas based on water quality and water level monitoring as described in claim 2, characterized in that, Step S5 includes the following steps: S51, based on the predicted water level data of each water level monitoring well, generate the predicted head height distribution field of each water-bearing rock layer during the prediction period using spatial interpolation; S52, based on the predicted head height distribution field, determine the water inflow section and water outflow section of the water-bearing rock layer, and calculate the predicted head difference between them; S53, calculate the expected exchange flow rate between the target closed goaf and the water-bearing rock layer, using the following formula: Where Q is the expected exchange flow rate, in m³ / d; B is the width of the water-bearing stratum affected by the goaf, in m; L is the length of the water-bearing stratum affected by the goaf, in m; ΔH is the predicted hydraulic head difference, in m; h i is the vertical equivalent seepage path length, in meters; K is the vertical equivalent permeability coefficient, in meters per day.

4. The method for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring as described in claim 1, characterized in that, Step S2 includes the following steps: S21, performing stable isotope analysis on representative water samples collected in step S1, determining and obtaining the stable hydrogen and oxygen isotope compositions of background water samples from each aquifer and each mixed water sample; S22, using the stable hydrogen and oxygen isotope compositions of background water samples from each aquifer as endmember values ​​and the stable hydrogen and oxygen isotope compositions of each mixed water sample as mixed values, inputting them into the isotope mixing model for calculation, to obtain the water volume contribution ratio of each aquifer to each groundwater discharge body; S23, for each aquifer, multiplying its water volume contribution ratio to a certain groundwater discharge body by the total monitored flow rate of that discharge body in the current monitoring period, to obtain the water volume contribution of that aquifer to that groundwater discharge body, summing up all the water volume contributions of the aquifer as its direct water volume contribution in the current monitoring period.

5. The method for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring as described in claim 1, characterized in that, The water level prediction model is a BP neural network model optimized by particle swarm optimization algorithm, employing a three-layer feedforward network, including: an input layer, at least one hidden layer, and an output layer; wherein, the input layer has multiple input neuron nodes, the hidden layer has multiple hidden neuron nodes, and each input neuron node is connected to each of the hidden neuron nodes; the hidden neuron nodes are all connected to the output layer; the construction and training process of the water level prediction model is as follows: a first dataset containing multiple historical monitoring periods and a corresponding second dataset are obtained, and all parameters from both are input into the input neuron nodes, using the predicted water level data corresponding to each historical monitoring period as the training target; the initial weights and thresholds of the BP neural network model are globally optimized using particle swarm optimization algorithm, and the input neuron nodes, hidden neuron nodes, and output layer are iteratively trained using an error backpropagation algorithm; training is completed when the error index of the output result reaches the preset requirements, thus obtaining the water level prediction model.

6. The method for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring as described in claim 1, characterized in that, The groundwater discharge bodies include mine drainage water and groundwater source intake within the monitoring area. The expected flow rates of each groundwater discharge body in the second dataset include the expected drainage flow rate of each mine and the planned pumping volume of each water source intake point. The formula for calculating the expected drainage flow rate is as follows: Where Q0 is the expected drainage flow rate of the mine, in m³ / d; K0 is the seepage coefficient of the mine, in m / d; S is the drawdown, in m; M is the average thickness of the water-bearing strata corresponding to the mine, in m; R0 is the radius of influence of the mine, in m; and r0 is the equivalent radius of the mine, in m.

7. The method for early warning of pollution release from closed mining goaf areas based on water quality and water level monitoring as described in claim 1, characterized in that, The method for calculating the expected pollutant discharge includes the following steps: S61, based on the water level data of the water-bearing rock strata collected in the current monitoring cycle, calculate the instantaneous exchange flow rate Q' of the target closed goaf at the end of the current monitoring cycle, and calculate the expected exchange flow rate Q at the end of the prediction cycle based on the predicted water level data, and take the arithmetic mean of Q' and Q as the expected average exchange flow rate Q. avg S62. Based on the pollutant concentration analysis results of water samples from the target closed goaf, determine the average concentration C of the target pollutant in the closed goaf during the current monitoring period; the formula for calculating the expected pollutant discharge is: Where T is the total duration of the prediction period.

8. A pollution release early warning system for closed mining goaf areas, used to execute the pollution release early warning method for closed mining goaf areas based on water quality and water level monitoring as described in any one of claims 1 to 7, characterized in that, include: Data acquisition module, data processing module, and early warning module; The data acquisition module collects water level data of each water-bearing rock layer through deployed water level monitoring wells, collects representative water samples within the monitoring area, and monitors the flow rate of all groundwater discharge bodies. The data processing module, connected to the data acquisition module, is used to process and predict the acquired data. It includes a water source analysis submodule, which is used to analyze the acquired representative water samples and obtain the direct contribution of each water-bearing rock layer. The dataset construction submodule is used to construct the first dataset and the second dataset; the water level prediction submodule is used to use the trained water level prediction model to output the predicted water level data of each water-bearing rock layer at the end of the prediction period, and to calculate the expected exchange flow of the target closed goaf; the pollutant calculation module is connected to the data processing module to obtain the expected exchange flow of the target closed goaf, the pollutant concentration of the goaf water sample, and to calculate the expected pollutant discharge of the target closed goaf within the prediction period; the early warning module is connected to the pollutant calculation module to compare the calculated expected pollutant discharge with a preset safety threshold, and if it exceeds the threshold, to generate and issue a pollution risk early warning information for the closed goaf.

9. An electronic device, characterized in that, The method includes a memory and a processor, wherein the memory stores computer instructions, and the processor executes the computer instructions to perform any one of the methods for early warning of pollution release from closed mining areas based on water quality and water level monitoring, as described in any one of claims 1 to 7.

Images