Mine underground water control automatic monitoring device and method

Abstract

The invention provides a mine underground water control automatic monitoring device and method, and belongs to the technical field of mine underground safety monitoring. The device comprises a data acquisition module, a data preprocessing module, an intelligent analysis module, a model fusion and optimization module and a real-time monitoring and control execution module. The data acquisition module is used for acquiring underground multi-source data; the data preprocessing module is used for carrying out abnormal value processing and data standardization operation on underground multi-source data; the intelligent analysis module comprises a water regimen anomaly detection sub-module and a flood risk prediction sub-module which are respectively used for underground water regimen anomaly detection and flood risk prediction; the model fusion and optimization module is used for fusing and optimizing output results of the water regimen anomaly detection sub-module and the flood risk prediction sub-module; and the real-time monitoring and control execution module is used for analyzing real-time monitoring data according to a fusion optimization result and executing water prevention and control measures, so that the problem that monitoring is not timely in the mine underground water prevention and control work is effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of underground mine safety monitoring, and in particular to an automatic monitoring device and method for underground mine water prevention and control. Background Art

[0002] During mining operations, underground water prevention and control remain a critical component in ensuring safe production. Due to the complex and ever-changing geological conditions underground in mines, numerous factors can potentially cause flooding. For example, water from aquifers can flow into the mining area through fissures and faults. Water accumulation in old goafs poses a constant threat to the surrounding operating environment. Mines affected by surface water also face the additional risks posed by external factors such as rainfall. These complex water hazards can easily lead to water inrush accidents, seriously endangering the lives of underground workers and the normal production and operation of the mine.

[0003] Traditional underground mine flood prevention and control efforts rely primarily on regular manual inspections and simple instrument monitoring. Manual inspections require workers to go down the mine at fixed intervals to check water levels and the operation of drainage equipment. This method is not only labor-intensive and inefficient, but also lacks real-time monitoring, making it difficult to detect sudden changes in water conditions. Regarding instrument monitoring, while some mines are equipped with conventional water level and water pressure sensors, these sensors are mostly single-function and have outdated data collection, transmission, and analysis methods. Data is typically collected at a fixed frequency and transmitted via wired channels to a ground control center. Abnormal water conditions are then determined based on simple threshold comparisons, lacking comprehensive analysis of multi-source data and intelligent anomaly prediction capabilities. Once abnormal water conditions occur, the inability to obtain comprehensive and accurate water information and respond quickly makes it difficult to implement effective control measures, which can easily lead to the expansion of flooding and irreversible losses. Summary of the Invention

[0004] In view of the technical problems existing in the background technology, the present application provides an automatic monitoring device and method for underground mine water prevention and control, aiming to solve the problem of untimely monitoring in existing underground mine water prevention and control work.

[0005] In a first aspect, an embodiment of the present application provides an automatic monitoring device for underground water prevention and control in a mine, comprising a data acquisition module, a data preprocessing module, an intelligent analysis module, a model fusion and optimization module, and a real-time monitoring and control execution module; The data acquisition module is used to collect multi-source data downhole; The data preprocessing module is connected to the data acquisition module and is used to perform outlier processing and data standardization operations on the collected downhole multi-source data; The intelligent analysis module is connected to the data preprocessing module and includes a water regime anomaly detection submodule and a flood risk prediction submodule, which are used for underground water regime anomaly detection and flood risk prediction respectively; The model fusion and optimization module is connected to the intelligent analysis module and is used to fuse and optimize the output results of the water anomaly detection submodule and the flood risk prediction submodule; The real-time monitoring and control execution module is connected to the model fusion and optimization module, and is used to perform real-time monitoring and analysis on the real-time monitoring data according to the results of the fusion optimization, and to execute water prevention and control measures according to the results of the real-time monitoring and analysis.

[0006] In the technical solution of the embodiment of the present application, by collecting and preprocessing multi-source data underground, accurate, rich data that can be compared and analyzed at the same scale can be obtained. On this basis, water anomaly detection and flood risk prediction are performed through the water situation anomaly detection submodule and the flood risk prediction submodule, and the results of the two are fused and optimized to form an optimized model that can detect water anomalies and predict flood risks. The model is used for real-time monitoring and analysis of real-time monitoring data, so as to detect anomalies in time so as to take corresponding water prevention and control measures, effectively solving the problem of untimely monitoring in existing underground mine water prevention and control work.

[0007] In some embodiments, the step of handling outliers includes: Perform preliminary screening of data based on the Laida criteria and sensor range; The local outlier factor algorithm is then used to analyze the data obtained after the preliminary screening, and the data points with local outlier factors greater than the preset threshold are identified as outliers and removed from the data sequence.

[0008] In this embodiment, preliminary screening is performed using the Laida criterion and the sensor range, and then the local outlier factor algorithm is used to further confirm outliers. This can accurately eliminate obviously abnormal data points, so that the data used for subsequent analysis and modeling can more accurately reflect the actual water conditions in the well, reduce the interference and misleading of abnormal data on the overall data analysis results, and solve the problem of difficult accurate identification of outliers in existing data processing methods.

[0009] In some embodiments, the data normalization operation includes: Divide the data into numerical continuous variables, water quality data with a range of values, and categorical variables; For the numerical continuous variables, Z-score standardization method is used for preliminary normalization, and then the preliminary normalized data is mapped to the preset interval through linear transformation; For the water quality data with a range of values, the Min-Max standardization method is used for processing; For the categorical variables, one-hot encoding is first performed and then mapped to a low-dimensional embedding vector space through an embedding layer.

[0010] In this embodiment, by adopting different standardization methods for different types of data, the numerical continuous variables, water quality data with a range of values, and categorical variables in the underground water control monitoring data can be unified on an appropriate scale for comparison and analysis, thereby avoiding the impact of inconsistent data scales on subsequent analysis.

[0011] In some embodiments, the water anomaly detection submodule adopts a hybrid architecture that combines a deep convolutional neural network with a long short-term memory network. By inputting preprocessed data, it is trained using the cross-entropy loss function and the Adam optimizer. During the training process, the hyperparameters are adjusted according to the accuracy, recall rate, and F1 value indicators to obtain a model that can be used for water anomaly detection.

[0012] In this embodiment, by adopting a hybrid architecture combining a deep convolutional neural network with a long short-term memory network, and through reasonable training and optimization, the resulting model can deeply explore the spatiotemporal characteristics of underground water control data, thereby accurately distinguishing different degrees of abnormal water conditions. Compared with traditional detection methods that rely on simple threshold judgments or single model structures, the water anomaly detection submodule provided in this application can effectively solve the problem of insufficient monitoring of the spatiotemporal characteristics of water data, providing a strong guarantee for the timely detection of potential floods.

[0013] In some embodiments, the flood risk prediction submodule screens out key features that are strongly correlated with flood risk through correlation analysis, and then uses principal component analysis to reduce the dimensionality of the key features. Then, a multi-layer perceptron and random forest fusion architecture is used to input the reduced dimensionality key feature data into the multi-layer perceptron and random forest models for training respectively, and then the output results of the two models are fused through the Stacking integration method, and the model is optimized according to the mean square error evaluation index to obtain a model that can be used for flood risk prediction.

[0014] In this embodiment, the flood risk prediction submodule fully utilizes the effective information in the multi-source underground data through feature screening, dimensionality reduction to model construction, fusion and optimization, overcomes the limitations of single model and single feature analysis, and can more accurately predict the level or probability of flood risk, etc., and solves the problem of feature redundancy. Based on this, whether considering complex flood scenarios under the influence of multiple factors such as geological structure, mining activities, or meteorological factors, the flood risk prediction submodule provided by this application can provide relatively reliable prediction results, providing a strong basis for taking corresponding water prevention and control measures in advance.

[0015] In some embodiments, the model fusion and optimization module fuses the output results of the water anomaly detection submodule and the flood risk prediction submodule by using a Bayesian model averaging method, including the following steps: respectively determining the prior probabilities of the model constructed by the water anomaly detection submodule and the model constructed by the flood risk prediction submodule; Calculate the posterior probability weight of each model after fusion based on the prior probability of each model and its performance on the training data; The results of the two models are weighted averaged according to the posterior probability weights.

[0016] In this embodiment, the fusion is performed through the Bayesian model averaging method, which fully considers the prior information of each model and its performance on the training data, reasonably allocates weights and weighted averages the results, avoids the limitations and errors that may exist in a single model, and combines the respective advantages of the water situation anomaly detection submodule and the flood risk prediction submodule, so that the final fusion model can output more accurate and reliable results when judging the water situation and predicting flood risks, and is more in line with the actual complex water situation underground.

[0017] In some embodiments, the real-time monitoring and control execution module continuously receives real-time monitoring data and inputs the real-time monitoring data into the integrated and optimized model after pre-processing. The integrated and optimized model is used to analyze the current water status in real time and predict future flood risks. The results are output and compared with the preset warning threshold. When water abnormalities are detected or the flood risk exceeds the preset warning threshold, the warning mechanism is triggered.

[0018] In this embodiment, by continuously receiving real-time monitoring data and quickly analyzing and processing it, real-time monitoring of the entire water situation in the mine is achieved. Any subtle changes in the water situation and potential flood risks can be captured in a timely manner, and early warnings can be issued in a timely manner, so that mine-related personnel can know the water situation dynamics at the first time, gain valuable time for taking countermeasures, and effectively solve the problem of untimely water situation monitoring in the existing technology.

[0019] In some embodiments, the data acquisition module includes monitoring sensors that collect data according to an adaptive sampling period, and the data collected by the monitoring sensors are transmitted to a data center through a hybrid communication network; the monitoring sensors include water level sensors, water pressure sensors and water quality sensors, and the hybrid communication network includes wired communication and wireless communication.

[0020] In this embodiment, by collecting data according to an adaptive sampling cycle, the data collection density can be rapidly increased at key nodes where water conditions suddenly change, thus resolving the problem of delayed monitoring associated with traditional fixed sampling cycles. Furthermore, by employing a hybrid communication network, the respective advantages of wired and wireless communications can be fully utilized, ensuring the stability of the data transmission process and reducing data loss or transmission interruptions caused by the harsh underground environment. Furthermore, the system can flexibly cover all areas underground where data collection is required, ensuring that all types of sensor data from different locations are fully and accurately transmitted to the data center, providing a sufficient data foundation for subsequent comprehensive analysis.

[0021] In some embodiments, the underground multi-source data includes underground water prevention and control monitoring data, underground geological structure data, mining engineering data and meteorological data.

[0022] In this embodiment, by collecting data from various aspects, the underground water conditions can be reflected more comprehensively and effectively, so that the model trained based on the data can output more accurate and reliable results when judging the water conditions and predicting flood risks.

[0023] In a second aspect, an embodiment of the present application provides a method for automatically monitoring underground water prevention and control in a mine, comprising the following steps: S1. Collect downhole multi-source data through the data acquisition module; S2. Performing outlier processing and data standardization on the collected downhole multi-source data through a data preprocessing module to obtain preprocessed data; S3. Based on the preprocessed data, the water anomaly detection submodule and the flood risk prediction submodule respectively perform water anomaly detection and flood risk prediction; S4. Fusion and optimization of the results of water anomaly detection and flood risk prediction using the model fusion and optimization module; S5. Acquire real-time monitoring data through the real-time monitoring and control execution module, perform real-time monitoring and analysis on the real-time monitoring data according to the results of the fusion optimization, and implement water prevention and control measures according to the results of the real-time monitoring and analysis.

[0024] In the technical solution of the embodiment of the present application, standardized high-quality data can be obtained by comprehensive data collection and preprocessing. Then, through the water situation anomaly detection submodule and the flood risk prediction submodule and the fusion optimization of the two, a model that can detect water situation anomalies and predict flood risks can be formed, which is used for real-time monitoring and analysis of real-time monitoring data, so as to detect anomalies in time so that corresponding water prevention and control measures can be taken.

[0025] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] To more clearly illustrate the technical solution of this application, the following is a brief introduction to the drawings used in this application. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be derived from these drawings without inventive effort.

[0027] Figure 1 This is a module architecture diagram of the automatic monitoring device for underground mine water prevention and control provided in an embodiment of the present application; Figure 2 This is an architectural diagram of a data acquisition module in an automatic monitoring device for underground water prevention and control in a mine provided in an embodiment of the present application; Figure 3 This is an architectural diagram of the intelligent analysis module in the automatic monitoring device for underground mine water prevention and control provided in the embodiments of the present application. DETAILED DESCRIPTION

[0028] The following embodiments of the technical solution of the present application will be described in detail with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application and are therefore only examples and are not intended to limit the scope of protection of the present application.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0030] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0031] In order to solve the technical problem of untimely monitoring in the existing underground water prevention and control work in mines, the present application provides an automatic monitoring device and method for underground water prevention and control in mines. By comprehensively collecting and preprocessing relevant data, standardized high-quality data are obtained, and then based on these data, a model is formed to detect abnormal water conditions and predict flood risks. The model is used to monitor and analyze the real-time monitoring data in real time, so as to detect abnormalities in time and take corresponding water prevention and control measures.

[0032] Please refer to Figure 1-3 In a first aspect, the present application provides an automatic monitoring device for underground water prevention and control in a mine, including a data acquisition module, a data preprocessing module, an intelligent analysis module, a model fusion and optimization module, and a real-time monitoring and control execution module; The data acquisition module is used to collect multi-source data downhole; The data preprocessing module is connected to the data acquisition module and is used to perform outlier processing and data standardization operations on the collected multi-source downhole data; The intelligent analysis module is connected to the data preprocessing module, and includes a water regime anomaly detection submodule and a flood risk prediction submodule, which are used for underground water regime anomaly detection and flood risk prediction respectively; The model fusion and optimization module is connected to the intelligent analysis module to integrate and optimize the output results of the water anomaly detection submodule and the flood risk prediction submodule; The real-time monitoring and control execution module is connected to the model fusion and optimization module to perform real-time monitoring and analysis on the real-time monitoring data according to the results of fusion optimization, and to implement water prevention and control measures according to the results of real-time monitoring and analysis.

[0033] In this application, by collecting and preprocessing multi-source data underground, accurate, rich data that can be compared and analyzed at the same scale can be obtained. On this basis, water anomaly detection and flood risk prediction are performed through the water situation anomaly detection submodule and the flood risk prediction submodule, and the results of the two are fused and optimized to form an optimized model that can detect water anomalies and predict flood risks. It is used for real-time monitoring and analysis of real-time monitoring data, so as to detect anomalies in time and take corresponding water prevention and control measures, effectively solving the problem of untimely monitoring in existing underground mine water prevention and control work.

[0034] Furthermore, in some embodiments, the underground multi-source data includes underground water prevention and control monitoring data, underground geological structure data, mining engineering data and meteorological data. Among them, the underground water prevention and control monitoring data can be collected by setting relevant monitoring sensors underground, and the underground geological structure data and mining engineering data can be obtained from geological exploration reports, mining plans and other channels. For mines that will be affected by surface water, meteorological data can be obtained through meteorological departments, so that the underground water conditions can be reflected more comprehensively and effectively.

[0035] Furthermore, in some embodiments, the data acquisition module includes monitoring sensors that collect data according to an adaptive sampling period, which is used to collect underground water prevention and control monitoring data. Among them, the monitoring sensors include water level sensors, water pressure sensors and water quality sensors. By arranging various types of sensors such as water level, water pressure, water quality, etc. at different key positions underground in the mine, and collecting data according to an adaptive sampling period, the sampling period can be dynamically adjusted according to the real-time water situation changes underground. When the rate of change of water level or water pressure exceeds a preset threshold, the sampling period is shortened. Compared with the traditional fixed sampling period method, the adaptive sampling period adopted in this application can quickly increase the data collection density at key nodes where the water situation suddenly changes, thereby effectively solving the problem of untimely monitoring in traditional underground mine water prevention and control work. Specifically, the sampling period adjustment method of the adaptive sampling period is as follows: under normal working conditions, each monitoring sensor collects data according to a pre-set initial sampling period and calculates the change rate of each data (for example, by calculating the ratio of the difference between the water level data at two adjacent sampling moments to the time interval to obtain the water level change rate; the change rate of other data is the same); when the calculated change rate exceeds the preset threshold, the sampling period adjustment mechanism is triggered to shorten the sampling period to a predetermined ratio, so that data collection is more frequent, so as to capture the rapidly changing details of the water situation.

[0036] Furthermore, in some embodiments, the data acquisition module and the data preprocessing module are connected through a data center, the data acquisition module transmits data to the data center, and the data preprocessing module obtains the data from the data center and preprocesses it. The data collected by the data acquisition module is preferably transmitted to the data center through a hybrid communication network, which includes wired communication methods such as RS485 bus and optical fiber and wireless communication methods such as ZigBee and 5G modules. The wired communication method is suitable for long-distance, large-data volume and main data transmission links with high transmission reliability requirements due to its stability and high bandwidth; the wireless communication method uses its flexibility to conveniently realize short-distance networking communication between sensors and surrounding nodes, and is particularly suitable for local areas in complex underground environments where large-scale wiring is difficult to carry out, thereby building a comprehensive coverage and efficient data transmission channel. This application fully utilizes the respective advantages of wired communication and wireless communication by adopting a hybrid communication network, which not only ensures the stability of the data transmission process and reduces data loss or transmission interruption caused by the harsh underground environment (such as electromagnetic interference, physical damage, etc.), but also can flexibly cover various areas underground where data collection is required, ensuring that various sensor data from different locations can be transmitted to the data center in a complete and accurate manner, solving the problem of limited communication and providing a sufficient data basis for subsequent comprehensive analysis.

[0037] Furthermore, in some embodiments, the step of performing outlier processing by the data preprocessing module includes: Perform preliminary screening of data based on the Laida criterion (3σ principle) and sensor range; The local outlier factor algorithm is then used to analyze the data obtained after the preliminary screening, and the data points with local outlier factors greater than the preset threshold are identified as outliers and removed from the data sequence.

[0038] Among them, the screening standard of the Laida criterion (3σ principle) is that if the difference between a certain data and the mean exceeds 3 times the standard deviation, the data will be judged as an outlier. The specific method is: Suppose a monitoring data sequence is , its mean and standard deviation Calculated using the following formula: = In the above formula, n For data series X The number of data points in , x i For data series X Thei data points, For data series X The mean of For data series X The standard deviation of .

[0039] Based on the Laida criterion (3σ principle), >3 All data points were determined as abnormal values ​​for preliminary screening.

[0040] When performing preliminary screening of data based on the sensor's measurement range, data that exceeds the sensor's normal measurement range is considered an outlier.

[0041] On the basis of the above preliminary screening, the local outlier factor (LOF) algorithm is used for further analysis. The specific method is as follows: for each data point in the data sequence, the local outlier factor is calculated using the local outlier factor algorithm. If the local outlier factor calculated for a data point is greater than a preset threshold (the threshold can be pre-set based on historical data or experience to distinguish the boundary between normal data and abnormal data in terms of outlier degree), the data point is finally confirmed as an outlier and removed from the data sequence to ensure the data quality of subsequent analysis.

[0042] For the small number of missing values ​​generated after removing outliers, you can choose to modify them by using methods such as linear interpolation or radial basis function interpolation based on the spatiotemporal characteristics of the data.

[0043] The embodiment of the present application uses the Laida criterion and the sensor range for preliminary screening, and then uses the local outlier factor algorithm to further confirm the outliers, thereby solving the problem of accurate identification of outliers. It can accurately eliminate obvious abnormal data points caused by factors such as sensor failure and sudden interference, so that the data used for subsequent analysis and modeling can more accurately reflect the actual water conditions in the well, reducing the interference and misleading of abnormal data on the overall data analysis results.

[0044] Furthermore, in some embodiments, the step of performing data normalization by the data preprocessing module includes: Divide the data into numerical continuous variables, water quality data with a range of values, and categorical variables; For numerical continuous variables, the Z-score standardization method was used for preliminary normalization, and then the preliminary normalized data were mapped into the preset interval through linear transformation; For water quality data with a range of values, the Min-Max normalization method is used for processing; For categorical variables, one-hot encoding is first performed and then mapped to a low-dimensional embedding vector space through an embedding layer.

[0045] Specifically, for numerical continuous variables (such as water level, water pressure, etc.) in underground water control monitoring data, the methods for preliminary normalization using the Z-score standardization method include: Suppose the data sequence of a numerical continuous variable is Y , its standardized value Calculate using the following formula: in, For data series Y The mean of For data series Y The standard deviation of For data series Y The i data points, for The calculation method of mean and standard deviation is the same as that of the data series above. X The corresponding calculation method is the same as in , so it will not be repeated here.

[0046] After Z-score standardization and linear transformation, the distribution of numerical continuous variables becomes more regular, and the mean of the sequence of standardized values ​​becomes 0 and the standard deviation becomes 1. Then, linear transformation is used to map the data to a specific interval, such as the interval [0, 1], to meet the requirements of subsequent analysis or model input, facilitating statistical analysis and model training.

[0047] For water quality data with a specific range of values ​​(such as pH, conductivity, etc.), the steps for processing using the Min-Max normalization method include: Assume that the water quality data sequence is Z , first find the minimum value in the sequence and maximum value Then, the normalized values ​​are calculated according to the following formula: in, For data series Y The i data points, for The normalized value.

[0048] After processing by the above Min-Max standardization method, the standardized data is mapped to interval, so that the data of different water quality indicators can be unified on the same scale for comparison and analysis.

[0049] Similarly, after the categorical variables are mapped through one-hot encoding and embedding layers, the categorical information is retained in a low-dimensional space, while the data dimension is reduced, which can improve computational efficiency.

[0050] Furthermore, in some embodiments, the water anomaly detection submodule within the intelligent analysis module can analyze water anomaly conditions by constructing a model capable of detecting water anomalies. This submodule utilizes a hybrid architecture combining a deep convolutional neural network with a long short-term memory network. By inputting a preprocessed feature dataset, the dataset is divided into a training set, a validation set, and a test set in a certain proportion, annotated with water status labels, and trained using a cross-entropy loss function combined with the Adam optimizer. During the training process, hyperparameters are adjusted based on accuracy, recall, and F1 value metrics to prevent overfitting, resulting in a model capable of detecting water anomalies.

[0051] Specifically, the water anomaly detection submodule utilizes a hybrid architecture combining a deep convolutional neural network (CNN) with a long short-term memory (LSTM) network. This architecture comprises multiple convolutional layers, pooling layers, activation function layers, and LSTM layers. The convolutional layer uses a sliding convolution kernel over the input data (i.e., a preprocessed feature dataset containing various water-related feature information extracted from multiple sources of underground data). This convolution automatically extracts local spatial features of the data, such as correlations between data at different monitoring points and spatial patterns of changes in parameters such as water level and pressure. The pooling layer then downsamples the convolved feature map (e.g., using max pooling or average pooling), further reducing the data dimensionality and extracting key spatial feature information. The LSTM layer is specifically designed to process the long-term dependencies of time series data. Its internal cell state and gating mechanism (input gate, forget gate, output gate, etc.) work together to update the cell state based on the input features of the current moment and the hidden state of the previous moment, and output the hidden state of the current moment. In this way, the dynamic changes of data over time can be learned. Features such as periodic fluctuations in water level and trends in water pressure can be effectively learned.

[0052] This application uses a combination of CNN and LSTM to enable the model to comprehensively analyze the characteristics of underground water data from both spatial and temporal dimensions. Through a hybrid architecture and a reasonable training and optimization process, the model can deeply explore the spatiotemporal characteristics of underground water prevention and control data and accurately distinguish different degrees of abnormal water conditions. Whether it is subtle abnormal changes in water level and water pressure, or complex water fluctuation patterns, they can all be effectively identified. Compared with traditional detection methods that rely on simple threshold judgments or single model structures, this greatly improves the accuracy and reliability of water anomaly detection, solves the problem of insufficient monitoring of the spatiotemporal characteristics of water data, and provides a strong guarantee for the timely detection of potential floods.

[0053] Furthermore, in some embodiments, the flood risk prediction submodule within the intelligent analysis module can predict flood risk by constructing a model that can be used for flood risk prediction. The flood risk prediction submodule uses correlation analysis to screen key features strongly associated with flood risk, then uses principal component analysis to reduce the dimensionality of these key features. Then, using a multi-layer perceptron and random forest fusion architecture, the reduced dimensionality key feature data is fed into the multi-layer perceptron and random forest models for training. The outputs of the two models are then fused using a stacking ensemble method, and the model is optimized using a mean square error evaluation metric to produce a model that can be used for flood risk prediction.

[0054] Specifically, the flood risk prediction submodule uses correlation analysis to select key features that are highly correlated with flood risk. The method includes: for a large number of feature variables extracted from multi-source underground data, the key features that are highly correlated with flood risk are selected by calculating the correlation coefficients between them. For example, for two variables X and Y , the Pearson correlation coefficient method is used to calculate the correlation between two variables rXY , the calculation formula is as follows: rXY= The above formula calculates the correlation between each feature and a flood risk indicator (such as the likelihood of a flood occurring within a certain time period). By setting a correlation threshold and selecting features with an absolute correlation coefficient greater than the threshold as key features, we eliminate features with less significant contributions to flood risk prediction and weaker correlations. This reduces subsequent computational effort and data redundancy, while focusing on factors that significantly impact flood risk.

[0055] After filtering out the key features, since these features are often still numerous and may have high dimensions, principal component analysis (PCA) is then used to reduce the dimensionality of the high-dimensional features. The specific steps include: First calculate the covariance matrix of the data matrix composed of these key features: C= in, Indicates the The feature vector of the samples, is the mean of the eigenvector. Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalue and the corresponding eigenvector , sort the eigenvectors from large to small according to their eigenvalues. Then select the first By projecting the original eigenvectors into the selected principal component space, the high-dimensional features are transformed into low-dimensional space, which further reduces the dimension of the data while retaining most of the key information and simplifies the data complexity of subsequent model processing.

[0056] This application leverages the effective information from multi-source underground data through feature screening and dimensionality reduction, model construction, fusion, and optimization. This overcomes the limitations of single models and single feature analysis, addresses feature redundancy, and can accurately predict the level and probability of flood risk. Whether considering complex flood scenarios influenced by multiple factors, including geological structure, mining activities, and meteorological factors, this application can provide relatively reliable predictions, providing a strong basis for taking appropriate flood prevention and control measures in advance.

[0057] Furthermore, in some embodiments, the model fusion and optimization module fuses the output results of the water anomaly detection submodule and the flood risk prediction submodule by using a Bayesian model averaging method, including the following steps: Determine the prior probabilities of the models constructed by the water anomaly detection submodule and the flood risk prediction submodule respectively; Calculate the posterior probability weight of each model after fusion based on the prior probability of each model and its performance on the training data; The results of the two models are weighted averaged according to the posterior probability weights.

[0058] Among them, the prior probability is the initial probability value assigned to the reliability or applicability of each model without considering the current specific data based on domain knowledge and past experience. It reflects our subjective judgment on the relative advantages and disadvantages of each model before obtaining new data.

[0059] More preferably, the model fusion and optimization module also has a dynamic weight adjustment mechanism and an incremental learning mechanism, which can dynamically adjust the weights according to the real-time monitoring data characteristics and the performance of each model under different working conditions, and use new data to update the optimization model, monitor the model performance in real time, and trigger retraining or optimization processes when the performance degrades. Specifically, the real-time monitoring data characteristics cover the various characteristics presented by multi-source data such as water level, water pressure, water quality, etc. collected in real time underground, such as the rate of water level change, the amplitude of water pressure fluctuations, and the trend of water quality changes. Different working conditions include different stages of underground mining activities (such as development, mining, etc.), different seasons (considering the difference between rainy season and dry season when affected by surface water, etc.), and whether there are sudden geological changes.

[0060] This application uses the Bayesian model averaging method for fusion, fully considering the prior information of each model and its performance on the training data, reasonably allocating weights and weighted averaging the results, avoiding the limitations and errors that may exist in a single model, and combining the respective advantages of the water anomaly detection submodule and the flood risk prediction submodule, so that the final fusion model can output more accurate and reliable results when judging the water situation and predicting flood risks, and is more in line with the actual complex water conditions in the well.

[0061] Furthermore, in some embodiments, the real-time monitoring and control execution module continuously receives real-time monitoring data, and inputs the real-time monitoring data into the integrated and optimized model after pre-processing, analyzes the current water status in real time and predicts future flood risks through the integrated and optimized model, outputs the results and compares them with the preset warning threshold, and triggers the warning mechanism when water abnormalities are detected or the flood risk exceeds the preset warning threshold.

[0062] Specifically, the real-time monitoring and control execution module continuously receives real-time monitoring data from sensors deployed at various locations underground (including water level sensors, water pressure sensors, water quality sensors, and so on) via a communication network constructed underground (e.g., a hybrid communication method including a wired RS485 bus, optical fiber, and wireless ZigBee and 5G modules). This raw data often has varying quality, such as outliers, inconsistent data formats, and inconsistent dimensions. Therefore, upon receiving the data, it is processed according to a pre-defined data preprocessing process. The data preprocessing method is consistent with that used in the data preprocessing module and will not be further elaborated here.

[0063] Afterward, the preprocessed real-time monitoring data is rapidly fed into a fused and optimized intelligent analysis model. This intelligent analysis model is derived from the previously described model fusion and optimization module, which integrates and optimizes the outputs of the water anomaly detection submodule and the flood risk prediction submodule. It is capable of comprehensively analyzing complex underground water conditions, accurately identifying water anomalies, and predicting flood risks. Upon receiving real-time data, the model uses its internal algorithm architecture and learned patterns to analyze the current underground water status in real time. For example, it determines whether the water level is within the normal range, whether the water pressure is stable, and whether there are any abnormal changes in water quality. It also uses historical data and current data trends to predict the risk of flooding in the future. If a water anomaly (such as a parameter like water level, pressure, or quality exceeding its normal fluctuation range) is detected, or if the flood risk exceeds a preset safety level, an early warning mechanism is immediately triggered.

[0064] Among them, the triggering of the early warning mechanism can be achieved in various ways. For example, the sound and light alarm in the underground operation area will send out sound and light alarm signals to remind the underground on-site workers to pay attention to changes in water conditions in time and take corresponding risk avoidance measures; at the same time, the early warning information will be sent to the ground control center and the mobile devices of relevant managers (such as mobile phones, tablets, etc.) through the mine's internal communication system (such as wired broadcasting, internal network, etc.), ensuring that relevant personnel can obtain information about abnormal water conditions or increased flood risks in the first time so that they can respond quickly.

[0065] By continuously receiving real-time monitoring data and rapidly analyzing and processing it, the system enables real-time monitoring of the entire underground water situation, promptly capturing any subtle changes in the water situation and potential flood risks. Whether it's a sudden surge in water levels or abnormal fluctuations in water pressure within a short period of time, or an increase in flood risk based on long-term data trend analysis, it can accurately detect and issue timely warnings, allowing mine personnel to be informed of water status dynamics immediately, buying valuable time to implement countermeasures and solving the problem of untimely water monitoring.

[0066] More specifically, detailed decision-making rules can be developed based on the mine's pre-established flood emergency plan and relevant safety regulations, combined with water situation analysis and forecasting results. Different flood risk levels correspond to different response measures. Specific decision recommendations are generated based on real-time water situation matching decision rules. These recommendations are promptly delivered to relevant personnel through the mine's internal information system and mobile applications, while also controlling and executing relevant equipment. Based on this, the real-time monitoring and control execution module can operate based on the mine's carefully prepared flood emergency plan and related strict safety regulations. The flood emergency plan details the various measures the mine should take in different flooding scenarios. It is based on the mine's past hydrogeological surveys, past flood cases, and estimates of various possible underground water conditions. It clearly defines key aspects such as the response process and division of responsibilities for different flood severity levels. Relevant safety regulations, from a more macro perspective, define the basic requirements and operational guidelines for flood prevention and control to ensure underground safety. The real-time monitoring and control execution module matches the real-time acquired and analyzed water conditions with established decision-making rules to generate specific decision recommendations. For example, if the current water analysis shows a slight rise in water levels but remains within normal fluctuations, and the flood risk is predicted to be low, the generated decision recommendation would include maintaining the operation of existing drainage equipment, increasing the frequency of inspections at specific monitoring points, and notifying relevant monitoring personnel to focus on key areas. By formulating decision rules based on professional emergency plans and safety regulations, combined with accurate water analysis and forecast results, the generated decision recommendations are scientifically sound and responsive to actual water conditions. This avoids the previous practice of relying on experience or blindly responding to floods, and addresses the lack of scientific basis for decision-making. Different flood risk levels correspond to different response measures, enabling precise responses to flood situations. This allows mines to take the most appropriate and effective measures to control flood risks and ensure the safety of underground operations in the face of various complex water changes.

[0067] In a second aspect, an embodiment of the present application provides a method for automatically monitoring underground water prevention and control in a mine, comprising the following steps: S1. Collect downhole multi-source data through the data acquisition module; S2. Performing outlier processing and data standardization on the collected downhole multi-source data through a data preprocessing module to obtain preprocessed data; S3. Based on the preprocessed data, the water anomaly detection submodule and the flood risk prediction submodule respectively perform water anomaly detection and flood risk prediction; S4. Fusion and optimization of the results of water anomaly detection and flood risk prediction using the model fusion and optimization module; S5. Acquire real-time monitoring data through the real-time monitoring and control execution module, perform real-time monitoring and analysis on the real-time monitoring data according to the results of the fusion optimization, and implement water prevention and control measures according to the results of the real-time monitoring and analysis.

[0068] Specifically, in the data collection step, multiple sensors and other data sources are integrated to collect multi-source data from underground wells, and adaptive sampling and hybrid communication networks are used to ensure the comprehensive and timely transmission of data. Next, the data preprocessing step performs outlier processing and standardization on the collected data to improve data quality. In the intelligent analysis step, the water anomaly detection submodule and the flood risk prediction submodule respectively use specific architectures and algorithms to perform water anomaly detection and flood risk prediction on the preprocessed data. Subsequently, the model fusion and optimization step uses Bayesian model averaging combined with a dynamic weight adjustment mechanism to fuse and optimize the results of the two submodules, enhancing the overall prediction accuracy and adaptability. Finally, the real-time monitoring and control execution step receives real-time data, conducts monitoring and analysis based on the fusion and optimization results, triggers warnings in the event of anomalies or exceeding risk levels, and pushes recommendations according to decision rules to control relevant equipment to implement water prevention and control measures.

[0069] In summary, the present application provides an automatic monitoring device and method for underground water prevention and control in mines, which belongs to the field of underground mine safety monitoring technology. The device includes a data acquisition module, a data preprocessing module, an intelligent analysis module, a model fusion and optimization module, and a real-time monitoring and control execution module; the data acquisition module is used to collect multi-source data underground; the data preprocessing module is used to perform outlier processing and data standardization operations on the multi-source data underground; the intelligent analysis module includes a water situation anomaly detection submodule and a flood risk prediction submodule, which are used for underground water situation anomaly detection and flood risk prediction respectively; the model fusion and optimization module is used to fuse and optimize the output results of the water situation anomaly detection submodule and the flood risk prediction submodule; the real-time monitoring and control execution module is used to analyze the real-time monitoring data according to the results of the fusion optimization and implement water prevention and control measures, thereby effectively solving the problem of untimely monitoring in underground mine water prevention and control work.

[0070] It should be noted that the present application is not limited to the above-mentioned embodiments. The above-mentioned embodiments are merely examples, and any embodiments having substantially the same structure and effect as the technical concept within the scope of the present application are all included in the technical scope of the present application. In addition, without departing from the scope of the present application, any other embodiments that can be conceived by those skilled in the art and that combine some of the constituent elements in the embodiments are also included in the scope of the present application.

Claims

1. An automatic monitoring device for underground water prevention in mines, characterized in that: It includes data acquisition module, data preprocessing module, intelligent analysis module, model fusion and optimization module, and real-time monitoring and control execution module; The data acquisition module is used to collect multi-source data downhole; The data preprocessing module is connected to the data acquisition module and is used to perform outlier processing and data standardization operations on the collected downhole multi-source data; The intelligent analysis module is connected to the data preprocessing module and includes a water regime anomaly detection submodule and a flood risk prediction submodule, which are used for underground water regime anomaly detection and flood risk prediction respectively; The model fusion and optimization module is connected to the intelligent analysis module and is used to fuse and optimize the output results of the water anomaly detection submodule and the flood risk prediction submodule; The real-time monitoring and control execution module is connected to the model fusion and optimization module, and is used to perform real-time monitoring and analysis on the real-time monitoring data according to the results of the fusion optimization, and to execute water prevention and control measures according to the results of the real-time monitoring and analysis.

2. The automatic monitoring device for underground mine water prevention and control according to claim 1 is characterized in that: The steps of outlier processing include: Perform preliminary screening of data based on the Laida criteria and sensor range; The local outlier factor algorithm is then used to analyze the data obtained after the preliminary screening, and the data points with local outlier factors greater than the preset threshold are identified as outliers and removed from the data sequence.

3. The automatic monitoring device for underground mine water prevention and control according to claim 1 is characterized in that: The data standardization operation includes: Divide the data into numerical continuous variables, water quality data with a range of values, and categorical variables; For the numerical continuous variables, Z-score standardization method is used for preliminary normalization, and then the preliminary normalized data is mapped to the preset interval through linear transformation; For the water quality data with a range of values, the Min-Max standardization method is used for processing; For the categorical variables, one-hot encoding is first performed and then mapped to a low-dimensional embedding vector space through an embedding layer.

4. The automatic monitoring device for underground mine water prevention and control according to claim 1 is characterized in that: The water anomaly detection submodule adopts a hybrid architecture that combines a deep convolutional neural network with a long short-term memory network. It inputs preprocessed data and uses the cross-entropy loss function and Adam optimizer for training. During the training process, hyperparameters are adjusted according to the accuracy, recall rate, and F1 value indicators to obtain a model that can be used for water anomaly detection.

5. The automatic monitoring device for underground mine water prevention and control according to claim 1 is characterized in that: The flood risk prediction submodule screens out key features that are highly correlated with flood risk through correlation analysis, and then uses principal component analysis to reduce the dimensionality of the key features. Then, a multi-layer perceptron and random forest fusion architecture is adopted to input the reduced dimensionality key feature data into the multi-layer perceptron and random forest models for training respectively. The output results of the two models are then fused through the Stacking integration method, and the model is optimized according to the mean square error evaluation index to obtain a model that can be used for flood risk prediction.

6. The automatic monitoring device for underground mine water prevention and control according to claim 1 is characterized in that: The model fusion and optimization module fuses the output results of the water anomaly detection submodule and the flood risk prediction submodule by using the Bayesian model averaging method, including the following steps: respectively determining the prior probabilities of the model constructed by the water anomaly detection submodule and the model constructed by the flood risk prediction submodule; Calculate the posterior probability weight of each model after fusion based on the prior probability of each model and its performance on the training data; The results of the two models are weighted averaged according to the posterior probability weights.

7. The automatic monitoring device for underground mine water prevention and control according to claim 1 is characterized in that: The real-time monitoring and control execution module continuously receives real-time monitoring data and inputs the real-time monitoring data into the integrated and optimized model after pre-processing. The integrated and optimized model is used to analyze the current water situation in real time and predict future flood risks. The results are output and compared with the preset warning threshold. The warning mechanism is triggered when water abnormalities are detected or the flood risk exceeds the preset warning threshold.

8. The automatic monitoring device for underground mine water prevention and control according to claim 1 is characterized in that: The data acquisition module includes monitoring sensors that collect data according to an adaptive sampling period, and the data collected by the monitoring sensors are transmitted to the data center through a hybrid communication network; the monitoring sensors include water level sensors, water pressure sensors and water quality sensors, and the hybrid communication network includes wired communication and wireless communication.

9. The automatic monitoring device for underground mine water prevention and control according to claim 1, characterized in that: The underground multi-source data includes underground water prevention and control monitoring data, underground geological structure data, mining engineering data and meteorological data.

10. A method for automatic monitoring of underground water prevention in mines, characterized in that: The steps include: S1. Collect downhole multi-source data through the data acquisition module; S2. Performing outlier processing and data standardization on the collected downhole multi-source data through a data preprocessing module to obtain preprocessed data; S3. Based on the preprocessed data, the water anomaly detection submodule and the flood risk prediction submodule respectively perform water anomaly detection and flood risk prediction; S4. Fusion and optimization of the results of water anomaly detection and flood risk prediction using the model fusion and optimization module; S5. Acquire real-time monitoring data through the real-time monitoring and control execution module, perform real-time monitoring and analysis on the real-time monitoring data according to the results of the fusion optimization, and implement water prevention and control measures according to the results of the real-time monitoring and analysis.

Images