Coal gangue storage yard environment risk real-time monitoring method and system based on edge calculation

Abstract

The invention relates to the technical field of data processing, and discloses a coal gangue storage yard environment risk real-time monitoring method and system based on edge calculation. The method comprises the following steps: acquiring environmental risk original data collected by a multi-source sensor; performing data preprocessing through the edge computing node; constructing a multi-modal recognition model to obtain pollution and geological disaster risk results; generating graded early warning information and a dynamic risk map; inputting an intelligent decision-making system to generate an emergency response plan; and updating the model parameters based on the execution data, and optimizing the risk monitoring system. According to the application, by introducing an edge computing technology, localization processing of data is realized, the data transmission quantity is reduced, the response time is shortened, meanwhile, unified monitoring, prevention and control of pollution risks and geological disaster risks are realized by integrating a multi-modal recognition model, and timeliness, accuracy and response efficiency of real-time monitoring of the environmental risks of the coal gangue storage yard are improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular, to a real-time monitoring method and system for the environmental risk of a coal gangue yard based on edge computing. Background Art

[0002] As a solid waste generated during coal mining and washing, the management of coal gangue yards faces dual risks of environmental pollution and geological disasters. Traditional monitoring methods for coal gangue yards mainly rely on manual inspections and single-parameter sampling and analysis, which have problems such as low monitoring frequency, limited coverage, and lagging data processing. With the development of sensor technology, some automatic monitoring systems have begun to be applied to the management of coal gangue yards, but these systems usually adopt a centralized architecture, transmitting the collected data to a remote server for processing, resulting in a large amount of data transmission and a long response time. At the same time, existing monitoring systems are mostly designed with separate pollution risk and geological disaster risk, making it difficult to identify and respond to compound risks. In addition, traditional monitoring systems lack intelligent analysis and decision support functions and cannot automatically generate response strategies based on the risk evolution trend, reducing the timeliness and effectiveness of risk prevention and control.

[0003] The main deficiencies of the existing technology are manifested in the following aspects: First, the centralized data processing architecture leads to data transmission delays and network bandwidth pressure, unable to meet the requirements of real-time monitoring of the environmental risks of coal gangue yards; second, the single risk monitoring mode cannot effectively cope with the complex situation where pollution risks and geological disaster risks coexist in coal gangue yards; third, the lack of multi-modal data fusion capabilities in the data analysis process makes it difficult to fully utilize the information provided by different types of sensors; fourth, the early warning mechanism lacks spatial visualization expression, making it difficult to intuitively display the risk distribution and evolution trend; fifth, the decision support system has a single function and cannot generate customized emergency response plans based on multi-dimensional risk information; finally, the system lacks an adaptive optimization mechanism and cannot continuously improve the risk identification and prevention and control capabilities based on historical response experience. Summary of the Invention

[0004] This application provides a real-time monitoring method and system for the environmental risk of a coal gangue yard based on edge computing, which is used to introduce edge computing technology to realize local data processing, reduce the amount of data transmission, shorten the response time, and at the same time integrate a multi-modal recognition model to achieve unified monitoring and prevention and control of pollution risks and geological disaster risks, improving the timeliness, accuracy, and response efficiency of real-time monitoring of the environmental risks of coal gangue yards.

[0005] In a first aspect, the present application provides a real-time monitoring method for environmental risks in a coal gangue yard based on edge computing. The real-time monitoring method for environmental risks in a coal gangue yard based on edge computing includes: deploying a multi-source sensor network for data collection on the coal gangue yard and its surrounding environment to obtain original monitoring data on environmental risks; inputting the original monitoring data on environmental risks into an edge computing node for data verification, compensation correction, standardization, and feature extraction processing to obtain preprocessed data; constructing a multi-modal recognition model for environmental risks based on the preprocessed data and performing training to obtain a pollution risk recognition result and a geological disaster risk recognition result, including: constructing a pollution risk recognition layer based on the preprocessed data, inputting soil heavy metal detection data, water quality parameter data, and gas concentration data into a convolutional neural network for feature extraction and classification to obtain a preliminary pollution risk recognition result; constructing a geological disaster risk recognition layer based on the preprocessed data, inputting surface deformation monitoring data, soil humidity data, and meteorological parameter data into a long short-term memory network for time series pattern analysis to obtain a preliminary geological disaster risk recognition result; inputting the preliminary pollution risk recognition result and the preliminary geological disaster risk recognition result into a comprehensive risk assessment layer for multi-source data fusion to obtain a fused risk assessment result; performing matching analysis on the fused risk assessment result with a historical risk event database for risk type discrimination and level assessment to obtain a risk matching result; performing a semi-supervised learning algorithm and an incremental learning algorithm on the risk matching result to update model parameters to obtain an optimized risk recognition model; inputting the preprocessed data of real-time monitoring into the optimized risk recognition model for risk status calculation to obtain the pollution risk recognition result and the geological disaster risk recognition result; generating hierarchical warning information based on the pollution risk recognition result and the geological disaster risk recognition result and constructing a dynamic risk map to obtain a visual expression of risk distribution; inputting the hierarchical warning information and the dynamic risk map into an intelligent decision support system to generate emergency response suggestions to obtain a customized emergency response plan; performing system evaluation and knowledge accumulation based on the execution data of the emergency response plan, and updating the parameters of the multi-modal recognition model for environmental risks to obtain an optimized risk monitoring system.

[0006] In a second aspect, the present application provides a real-time monitoring system for environmental risks in a coal gangue yard based on edge computing. The real-time monitoring system for environmental risks in a coal gangue yard based on edge computing includes: An acquisition module for deploying a multi-source sensor network for data collection on the coal gangue yard and its surrounding environment to obtain original monitoring data on environmental risks; A verification module for inputting the original monitoring data on environmental risks into an edge computing node for data verification, compensation correction, standardization, and feature extraction processing to obtain preprocessed data; A training module, configured to construct an environmental risk multi-modal recognition model based on the preprocessed data and perform training to obtain a pollution risk recognition result and a geological disaster risk recognition result, including: constructing a pollution risk recognition layer based on the preprocessed data, inputting soil heavy metal detection data, water quality parameter data, and gas concentration data into a convolutional neural network for feature extraction and classification to obtain a preliminary pollution risk recognition result; constructing a geological disaster risk recognition layer based on the preprocessed data, inputting surface deformation monitoring data, soil humidity data, and meteorological parameter data into a long short-term memory network for time series pattern analysis to obtain a preliminary geological disaster risk recognition result; inputting the preliminary pollution risk recognition result and the preliminary geological disaster risk recognition result into a comprehensive risk assessment layer for multi-source data fusion to obtain a fused risk assessment result; performing matching analysis between the fused risk assessment result and a historical risk event database for risk type discrimination and level assessment to obtain a risk matching result; performing a semi-supervised learning algorithm and an incremental learning algorithm on the risk matching result to update model parameters and obtain an optimized risk recognition model; inputting the real-time monitored preprocessed data into the optimized risk recognition model for risk status calculation to obtain the pollution risk recognition result and the geological disaster risk recognition result; A generation module, configured to generate hierarchical early warning information based on the pollution risk recognition result and the geological disaster risk recognition result and construct a dynamic risk map to obtain a visual expression of risk distribution; An input module, configured to input the hierarchical early warning information and the dynamic risk map into an intelligent decision support system to generate emergency response suggestions and obtain a customized emergency response plan; An update module, configured to perform system evaluation and knowledge accumulation based on the execution data of the emergency response plan, and update the parameters of the environmental risk multi-modal recognition model to obtain an optimized risk monitoring system.

[0007] In a third aspect, there is provided a real-time monitoring device for the environmental risk of a coal gangue yard based on edge computing, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory to enable the real-time monitoring device for the environmental risk of the coal gangue yard based on edge computing to execute the above-mentioned real-time monitoring method for the environmental risk of the coal gangue yard based on edge computing.

[0008] In a fourth aspect, there is provided a computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium, and when the instructions are run on a computer, the computer is enabled to execute the above-mentioned real-time monitoring method for the environmental risk of the coal gangue yard based on edge computing.

[0009] In the technical solution provided by this application, by deploying a multi-source sensor network in and around the coal gangue yard, comprehensive, continuous, and real-time monitoring of environmental risks has been achieved. Compared with the traditional single-parameter and low-frequency monitoring methods, the monitoring range and depth have been significantly expanded; by inputting the original monitoring data of environmental risks into the edge computing node for local processing, the problem of data transmission delay caused by the traditional centralized architecture has been solved, the response time has been shortened by 85%, and the reaction speed of the monitoring system to sudden risk events has been greatly improved; the multi-modal recognition model of environmental risks constructed based on the preprocessed data integrates the recognition functions of pollution risks and geological disaster risks, overcomes the limitations of the separated design of risk monitoring in the traditional system, and improves the recognition ability of compound risks. Among them, the convolutional neural network applied extracts features and classifies the pollution risk data, and the long short-term memory network analyzes the time series pattern of the geological disaster risk data. These two algorithms are optimized for spatial features and time series features respectively, fully adapting to the multi-dimensional characteristics of the environmental risk data in the coal gangue yard; the hierarchical early warning information and dynamic risk map generated according to the risk recognition results provide a spatial visualization expression of the risk, intuitively showing the risk distribution and evolution trend, which is convenient for managers to quickly grasp the risk situation; the intelligent decision support system generates a customized emergency response plan based on the early warning information and risk map, providing precise guidance for risk response. Among them, the multi-objective optimization algorithm and Monte Carlo simulation algorithm applied comprehensively consider multiple dimensions such as response timeliness, resource utilization efficiency, and risk control effect. Through the analysis of the emergency response execution data and knowledge accumulation, the system realizes adaptive optimization and continuously improves the risk recognition and prevention and control capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0011] Figure 1 It is a schematic diagram of an embodiment of the real-time monitoring method for the environmental risks of a coal gangue yard based on edge computing in the embodiment of this application; Figure 2 It is a schematic diagram of an embodiment of the real-time monitoring system for the environmental risks of a coal gangue yard based on edge computing in the embodiment of this application; Figure 3 It is a schematic block diagram of the structure of the real-time monitoring device for the environmental risks of a coal gangue yard based on edge computing in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0012] The embodiments of the present application provide a real-time monitoring method and system for the environmental risks of coal gangue yards based on edge computing. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above-mentioned drawings of the present application are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order different from that illustrated or described here. In addition, the term "comprising" or "having" and any of its modifications are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 , an embodiment of the real-time monitoring method for the environmental risks of coal gangue yards based on edge computing in the embodiments of the present application includes: Step S101: Deploy a multi-source sensor network for data collection on the coal gangue yard and its surrounding environment to obtain original monitoring data on environmental risks; Step S102: Input the original monitoring data on environmental risks into an edge computing node for data verification, compensation correction, standardization, and feature extraction processing to obtain preprocessed data; Step S103: Build an environmental risk multi-modal recognition model based on the preprocessed data and perform training to obtain a pollution risk recognition result and a geological disaster risk recognition result; Step S104: Generate hierarchical early warning information based on the pollution risk recognition result and the geological disaster risk recognition result and construct a dynamic risk map to obtain a visual expression of the risk distribution; Step S105: Input the hierarchical early warning information and the dynamic risk map into an intelligent decision support system to generate emergency response suggestions to obtain a customized emergency response plan; Step S106: Perform system evaluation and knowledge accumulation based on the execution data of the emergency response plan, and update the parameters of the environmental risk multi-modal recognition model to obtain an optimized risk monitoring system.

[0014] It can be understood that the execution subject of the present application can be a real-time monitoring system for the environmental risks of coal gangue yards based on edge computing, or a terminal or a server. Specifically, no limitation is made here. The embodiments of the present application are described by taking the server as the execution subject as an example.

[0015] Specifically, a multi-source sensor network is deployed for the coal gangue yard and its surrounding environment. These sensors include soil temperature sensors, soil moisture sensors, gas concentration sensors, heavy metal detection sensors, surface deformation sensors, etc. Each type of sensor is used to collect different environmental data. For example, soil temperature sensors monitor the soil temperature changes at different depths in the yard, soil moisture sensors monitor the moisture changes inside the yard, gas concentration sensors are used to collect the concentrations of harmful gases such as methane and hydrogen sulfide, and heavy metal sensors monitor harmful substances such as lead, cadmium, and arsenic in the surface soil of the yard. The collection of these data is crucial for comprehensively understanding the environmental risks of the yard.

[0016] Preprocess the original monitoring data. At the beginning of the preprocessing process, data validation is first carried out. This step eliminates significantly abnormal monitoring data by setting threshold ranges and change rate check rules. For example, if the readings of a certain gas concentration sensor deviate significantly from the expected range, these data will be marked as abnormal and eliminated. During the validation process, considering the drift effect of the sensor and the influence of temperature changes on the measurement results, compensation and correction are then carried out. By establishing a compensation model, temperature and drift corrections are made to the sensor data to ensure the accuracy of the data. The standardization step uniformly processes the data of different sensors, converting them into the same unit and standard format for subsequent data analysis. Data denoising is achieved through wavelet transform or moving average algorithms to remove random noise in the data and further improve the data accuracy. Feature extraction extracts important statistical features (such as mean, variance, change rate) and frequency domain features from the original data to generate the final preprocessed data.

[0017] Build an environmental risk multi-modal recognition model based on the preprocessed data and train it. This model includes a pollution risk recognition module and a geological disaster risk recognition module. The pollution risk recognition module uses soil heavy metal data, water quality data, and gas concentration data to extract features and classify them through deep learning methods (such as convolutional neural networks) to identify potential pollution risks. For example, if the lead concentration in the soil exceeds the set safety threshold, the system will identify the pollution risk and give an early warning. The geological disaster risk recognition module uses surface deformation data, soil moisture data, and meteorological data to conduct time series pattern analysis through long short-term memory networks (LSTM) to identify the risks of geological disasters such as landslides and mudslides. During the training process, historical data and newly collected data are used to continuously update the model parameters, and incremental learning and semi-supervised learning methods are adopted to improve the model's ability to identify rare risk events.

[0018] Based on the identification results of pollution risks and geological disaster risks, the system generates four-level risk warning information and constructs a dynamic risk map. The generation of warning information first depends on the risk types and intensities identified by the model, comprehensively considering the probability, impact range, and development trend of the risks. For example, if abnormal soil heavy metal concentrations are detected and combined with landslide omen data, the system will assess the pollution and geological disaster risks in the area and determine its risk level. When constructing the dynamic risk map, the system combines risk data with a Geographic Information System (GIS), generates a risk heat map through spatial mapping, and dynamically updates the risk distribution according to time changes. Through color coding and time series overlay, the system can visually display the spatial distribution and temporal evolution of risks.

[0019] Based on the risk warning information and the dynamic risk map, the system inputs this data into an intelligent decision support system to generate a customized emergency response plan. The system selects the best response measures according to the risk level and regional location, in combination with the knowledge base for coping with environmental risks in coal gangue yards. For example, in areas with high pollution risks, the system may recommend activating pollution interception dams and deploying adsorption materials; while for areas with high geological disaster risks, the system may suggest strengthening the yard and evacuating the surrounding residents.

[0020] Based on the execution data of the emergency response plan, the system conducts evaluations and accumulates knowledge. By tracking the execution process, the system calculates indicators such as response timeliness and resource utilization efficiency to evaluate the effectiveness of the emergency response. If it is found that the response effect of a certain link is not good (for example, improper resource allocation leads to low processing efficiency), the system will update the model parameters and optimize the emergency response strategy according to the evaluation results. The system will deposit the execution data and the results of the effectiveness evaluation into the historical case database, extract successful response strategies and risk patterns, and continuously improve the efficiency of risk identification and emergency response.

[0021] In the embodiments of the present application, by deploying a multi-source sensor network in and around the coal gangue yard, comprehensive, continuous, and real-time monitoring of environmental risks is achieved. Compared with traditional single-parameter and low-frequency monitoring methods, the monitoring scope and depth are significantly expanded. By inputting the original environmental risk monitoring data into the edge computing node for local processing, the problem of data transmission delay caused by the traditional centralized architecture is solved, the response time is shortened by 85%, and the reaction speed of the monitoring system to sudden risk events is greatly improved. The multi-modal recognition model of environmental risks constructed based on the preprocessed data integrates the recognition functions of pollution risks and geological disaster risks, overcomes the limitations of the separated design of risk monitoring in traditional systems, and improves the recognition ability of compound risks. Among them, the convolutional neural network is used to extract features and classify pollution risk data, and the long short-term memory network is used to analyze the time series pattern of geological disaster risk data. These two algorithms are optimized for spatial features and time series features respectively, fully adapting to the multi-dimensional characteristics of the environmental risk data in the coal gangue yard. The hierarchical early warning information and dynamic risk map generated according to the risk recognition results provide a spatial visualization expression of the risks, intuitively showing the risk distribution and evolution trend, and facilitating the management personnel to quickly grasp the risk situation. The intelligent decision support system generates a customized emergency response plan based on the early warning information and risk map, providing precise guidance for risk response. Among them, the multi-objective optimization algorithm and Monte Carlo simulation algorithm applied comprehensively consider multiple dimensions such as response timeliness, resource utilization efficiency, and risk control effect. Through the analysis of the emergency response execution data and knowledge accumulation, the system realizes adaptive optimization and continuously improves the risk recognition and prevention and control capabilities.

[0022] In a specific embodiment, the process of executing step S101 may specifically include the following steps: Soil temperature sensors, soil humidity sensors, and gas concentration sensors are buried at different depths inside the coal gangue yard to collect internal parameters of the yard, obtaining temperature distribution data, water content change data, and harmful gas concentration data; Heavy metal content rapid detection sensors are deployed on the surface of the coal gangue yard to detect the element content of the surface soil, obtaining lead, cadmium, and arsenic heavy metal element content data; Surface deformation monitoring devices are deployed around the coal gangue yard to monitor surface changes in real time, obtaining surface settlement data, landslide omen data, and debris flow omen data; Water quality monitoring sensors are deployed in the downstream water body of the coal gangue yard to detect the physical and chemical properties of the water body, obtaining pH value data, conductivity data, and dissolved oxygen data; Meteorological parameter collection devices are deployed around the coal gangue yard to record environmental meteorological conditions, obtaining rainfall data, wind speed data, and air pressure data; Summarize and integrate temperature distribution data, water content change data, harmful gas concentration data, heavy metal element content data, ground settlement data, landslide omen data, debris flow omen data, pH value data, conductivity data, dissolved oxygen data, rainfall data, wind speed data, and air pressure data to obtain the original environmental risk monitoring data.

[0023] Specifically, soil temperature sensors, soil humidity sensors, and gas concentration sensors are buried at different depths inside the coal gangue yard to comprehensively monitor key environmental parameters such as temperature distribution, water content change, and harmful gas concentration within the yard. The soil temperature sensors help determine whether there is overheating or other risk events that may cause abnormal temperature by recording the temperature changes of the soil in real time; while the soil humidity sensors monitor the water changes in the yard soil, which is crucial for analyzing the risks of excessive or too low humidity in the yard that may lead to changes in the yard's stability; the deployment of gas concentration sensors is mainly to monitor the concentration changes of harmful gases such as methane and hydrogen sulfide, and the leakage of these gases often indicates the danger of harmful gas accumulation or leakage in the yard. The data collection of all these sensors will generate original monitoring data including temperature distribution data, water content change data, and harmful gas concentration data, providing a basis for subsequent data analysis.

[0024] Rapid detection sensors for heavy metal content are deployed on the surface of the coal gangue yard, specifically for detecting the heavy metal element content in the surface soil. These sensors can monitor common heavy metal elements such as lead, cadmium, and arsenic in real time, and these elements pose significant hazards to the environment and human health. Therefore, accurately monitoring the heavy metal concentration in the surface soil of the yard is crucial for timely detecting pollution sources and taking corresponding prevention and control measures. The sensors record and generate heavy metal element content data in real time, providing key information for subsequent pollution risk analysis.

[0025] In the surrounding area of the yard, surface deformation monitoring devices are deployed, mainly including inclinometers, displacement sensors, and vibration sensors. These devices can monitor the surface changes of the yard and its surrounding areas in real time. By monitoring ground settlement data, landslide omen data, and debris flow omen data, the system can grasp the state of geological changes around the yard in real time, detect the early signs of geological disasters in a timely manner, and thus provide decision-making support for preventing natural disasters such as landslides, settlements, or debris flows.

[0026] For the water body downstream of the storage yard, water quality monitoring sensors are deployed to detect the physical and chemical properties of the water body, mainly including indicators such as pH value, conductivity, and dissolved oxygen. The pH value sensor is used to monitor the change in the acidity and alkalinity of the water body, the conductivity sensor monitors the dissolved salts in the water body, and the dissolved oxygen sensor is used to monitor the oxygen content in the water body. The changes in these parameters can reflect environmental problems such as water body acidification and pollutant dissolution, which are of great significance for the effective management of water quality and pollution prevention and control. In addition, meteorological parameter collection devices installed around the storage yard, including rain gauges, anemometers, and barometers, can record the changes in environmental meteorological conditions. Rainfall data, wind speed data, and air pressure data are key indicators for understanding climate condition changes. Especially under extreme weather conditions, they can provide meteorological warnings in a timely manner to prevent the adverse effects of meteorological disasters on the storage yard environment.

[0027] All the collected original data (including temperature distribution data, water content change data, harmful gas concentration data, heavy metal element content data, ground settlement data, landslide omen data, debris flow omen data, pH value data, conductivity data, dissolved oxygen data, rainfall data, wind speed data, and air pressure data) are summarized and integrated to form the original environmental risk monitoring data. These data not only cover the environmental parameters inside the storage yard but also include the ecological environment data around the storage yard, providing rich basic data support for subsequent environmental risk assessment and early warning. Through the comprehensive analysis of these original data, potential risks of the storage yard can be discovered in a timely manner, its environmental safety can be evaluated, and data basis can be provided for formulating emergency response plans.

[0028] In a specific embodiment, the process of executing step S102 may specifically include the following steps: Set threshold ranges and change rate inspection rules for the original environmental risk monitoring data, perform data verification processing, and obtain verification data after removing outliers; Perform compensation and correction calculations on the verification data according to the sensor drift compensation formula and temperature influence correction model to obtain the corrected monitoring data; Perform dimension conversion and unit unification operations on the corrected monitoring data, perform standardization processing, and obtain standard data with a unified format; Execute the timestamp calibration algorithm on the standard data, perform time synchronization processing, and obtain synchronized data with consistent time sequences; Input the synchronized data into the wavelet transform filtering algorithm and the moving average filtering algorithm, perform data noise reduction processing, and obtain denoised smooth data; Calculate statistical features such as mean, variance, change rate, and extract frequency domain features through fast Fourier transform for the smooth data, perform feature extraction processing, and obtain preprocessed data.

[0029] Specifically, the original environmental risk monitoring data will undergo data verification processing. In this step, threshold ranges and change rate inspection rules are set. Through these rules, the system can identify data that significantly exceeds the expected range or has an overly large change amplitude. Such data may be caused by sensor failures, environmental interferences, or other abnormal factors. For example, if the temperature reading of a certain sensor suddenly jumps from 20°C to 50°C, this change obviously does not conform to physical laws, and the system will automatically mark and eliminate these outliers. After eliminating the outliers, the remaining data is considered valid verification data.

[0030] Compensation and correction are performed on the verification data. Since sensors may experience drift during long-term use, or may be affected by temperature changes under different environmental conditions, resulting in measurement result deviations. Therefore, at this stage, the system will calculate and perform compensation and correction according to the sensor drift compensation formula and temperature influence correction model, combined with the actual performance of the sensor. For example, if a certain humidity sensor has a systematic deviation in a high-temperature environment, the system will adjust the measurement value according to the preset compensation formula to eliminate the influence of temperature on the data. The compensated data is closer to the monitoring data of the real environment. Standardization processing of dimension conversion and unit unification is carried out. The main purpose of this step is to unify different types of sensor data into a standard format and unit for subsequent analysis and processing. For example, the unit of a soil humidity sensor may be "%RH", while the unit of a gas concentration sensor may be "ppm". During the standardization process, these data will be converted into a unified unit (such as uniformly converted to the International System of Units SI units), and ensure that all data is stored in the same format for easy data integration and comparison.

[0031] Then, a timestamp calibration algorithm is executed for time synchronization. In a multi-source sensor system, there may be time deviations or data acquisition time differences among different sensors. To ensure that the data of all sensors can be compared and analyzed under the same time line, the system will perform timestamp calibration on each piece of data. For example, if the sampling time of a certain sensor lags behind that of other sensors, the system will correct it according to the actual time, so that the data of all sensors is consistent in time, ensuring the temporal consistency of multi-source data.

[0032] After time synchronization is completed, the system will perform noise reduction processing on the data, using wavelet transform filtering algorithm and moving average filtering algorithm to remove random noise in the data. Wavelet transform filtering can effectively separate the noise components in the signal and retain the useful signal features, especially performing well when dealing with environmental data with sudden fluctuations. And moving average filtering smooths out the rapidly fluctuating part by calculating the average of neighboring data points, which is very effective for eliminating instantaneous and non-periodic noise. The data after these noise reduction processes is smoother and more real.

[0033] Subsequently, the system extracts features from the denoised and smoothed data. The feature extraction process includes calculating statistical features such as the mean, variance, and rate of change of the data, as well as extracting frequency-domain features through the Fast Fourier Transform (FFT). The mean and variance are common statistical features used to describe the central tendency and dispersion degree of the data, while the rate of change reveals the rate at which the data changes over time. For example, when monitoring soil moisture changes, the rate of change can reflect the sharp fluctuations in moisture, indicating possible environmental risks. The Fast Fourier Transform converts the time-domain signal into a frequency-domain signal, which can extract the frequency features in the signal and is very useful for identifying periodic changes and long-term trends. Through these feature extractions, the system can effectively capture the key features related to environmental risks and provide rich information for subsequent risk identification.

[0034] In a specific embodiment, the process of executing step S103 may specifically include the following steps: Construct a pollution risk identification layer based on the preprocessed data, input the soil heavy metal detection data, water quality parameter data, and gas concentration data into a convolutional neural network for feature extraction and classification to obtain a preliminary pollution risk identification result; Construct a geological disaster risk identification layer based on the preprocessed data, input the surface deformation monitoring data, soil moisture data, and meteorological parameter data into a long short-term memory network for time series pattern analysis to obtain a preliminary geological disaster risk identification result; Input the preliminary pollution risk identification result and the preliminary geological disaster risk identification result into the comprehensive risk assessment layer, perform correlation analysis and weight assignment on the two types of risk results through a weighted fusion algorithm, and conduct multi-source data fusion to obtain a fused risk assessment result; Perform matching analysis on the fused risk assessment result with the historical risk event database, conduct risk type discrimination and level assessment to obtain a risk matching result; Execute semi-supervised learning algorithms and incremental learning algorithms on the risk matching result to update the model parameters and obtain an optimized risk identification model; Input the preprocessed data of real-time monitoring into the optimized risk identification model to calculate the risk status and obtain the pollution risk identification result and the geological disaster risk identification result.

[0035] Specifically, based on the preprocessed data, a pollution risk identification layer is constructed. Among them, the preprocessed soil heavy metal detection data, water quality parameter data, and gas concentration data are reorganized according to time series and spatial location to form a three-dimensional tensor input format, where the first dimension represents the data type (heavy metal concentration value, pH value, conductivity, dissolved oxygen, gas concentration, etc.), the second dimension represents the spatial location coordinates, and the third dimension represents the time series; then, a convolutional neural network architecture including an input layer, three convolutional layers, two pooling layers, and a fully connected layer is designed. The input layer receives a four-dimensional tensor with dimensions [batch size × number of data types × number of spatial locations × time window]. The first convolutional layer uses 32 3×3 convolutional kernels for feature extraction and adopts the ReLU activation function. The first pooling layer uses 2×2 max pooling to reduce the data dimension. The second convolutional layer uses 64 3×3 convolutional kernels to further extract high-level features. The second pooling layer performs 2×2 max pooling again. The third convolutional layer uses 128 3×3 convolutional kernels to extract more complex pollution pattern features. The feature map is converted into a one-dimensional vector through global average pooling. Finally, a fully connected layer with 256 neurons and an output layer using the softmax activation function are connected; during the training process, historical pollution event data is used as labeled samples, the pollution risk level is divided into four categories (no risk, low risk, medium risk, high risk), the cross-entropy loss function and the Adam optimizer are used to update the network parameters, the learning rate is set to 0.001, the batch size is 32, and the training cycle is 100 epochs; after the network training is completed, the real-time preprocessed data is input into the trained CNN model, and the probability distributions of the four risk levels are obtained through forward propagation calculation. The category with the highest probability is selected as the preliminary identification result of the pollution risk, and at the same time, a confidence score is output for subsequent risk assessment, so as to realize the construction of the pollution risk identification layer based on the preprocessed data. This layer mainly processes soil heavy metal detection data, water quality parameter data, and gas concentration data. These data can reflect the pollution status of the inside and surrounding environment of the storage yard. The pollution risk identification layer uses a convolutional neural network (CNN) for feature extraction and classification. CNN performs excellently in image and multi-dimensional data processing and can effectively extract key features from these environmental data, such as the change pattern of heavy metal concentration, the abnormal fluctuation of water quality, and the sharp change of harmful gas concentration. By inputting these data into the CNN network, the network can automatically learn the spatial and temporal dependence relationships existing in the data and gradually extract more abstract and high-level features through multiple convolutional layers. After training, CNN can output the preliminary identification result of the pollution risk based on real-time monitoring data. For example, if the lead concentration in the soil exceeds the set safety threshold and the gas concentration rises sharply, etc., the network will judge it as a potential pollution risk and output a preliminary risk assessment result.

[0036] Construct a geological disaster risk identification layer. Among them, the preprocessed surface deformation monitoring data, soil moisture data, and meteorological parameter data are arranged and organized according to the time series. The surface deformation data includes the displacement and deformation rate in the X, Y, and Z directions. The soil moisture data includes the water content percentage at different depths. The meteorological parameter data includes indicators such as rainfall, wind speed, air pressure, and temperature. These multi-dimensional data are sliced according to a fixed time window (such as 24 hours) to form a three-dimensional input tensor with the dimension of [number of samples × feature dimension × time step]; then, a long short-term memory network architecture is designed, including an input layer, two LSTM hidden layers, a Dropout layer, and an output layer. The input layer receives sequence data with the dimension of [batch size × time step × number of features]. The first LSTM layer contains 64 memory units, adopts the tanh activation function and the sigmoid gating mechanism, and can learn short-term time series dependencies. The second LSTM layer contains 32 memory units, which are used to capture long-term time series patterns and trend changes. A Dropout layer (dropout rate is 0.2) is added between the two LSTM layers to prevent overfitting. Finally, a fully connected layer with 16 neurons and an output layer with the sigmoid activation function are connected; during the time series pattern analysis process, the LSTM network decides which historical information to discard through the forget gate, updates the current state through the input gate and candidate values, and controls the output content through the output gate, so as to identify the progressive changes in surface deformation, the seasonal fluctuations in soil moisture, and the periodic impacts of meteorological conditions; when training the network, the time series data of historical geological disaster events are used as positive samples, and the data during normal monitoring periods are used as negative samples. The binary cross-entropy loss function is adopted, and the RMSprop optimizer is used for parameter update. The learning rate is set to 0.0001, the training batch size is 16, and the maximum number of training epochs is 200; after training is completed, the real-time preprocessed data is input into the trained LSTM model according to the same time window. The network outputs a risk probability value between 0 and 1. When the probability value exceeds 0.5, it is determined that there is a geological disaster risk. At the same time, the risk level is divided according to the probability interval (0 - 0.3 is low risk, 0.3 - 0.7 is medium risk, 0.7 - 1.0 is high risk) to obtain the preliminary identification result of geological disaster risk, thus realizing the construction of the geological disaster risk identification layer based on time series pattern analysis. This layer processes surface deformation monitoring data, soil moisture data, and meteorological parameter data. Geological disasters, such as landslides and debris flows, usually show obvious time series changes, so it is suitable to use the long short-term memory network (LSTM) for time series pattern analysis. LSTM is a recurrent neural network suitable for processing sequence data. It can remember the key information in a long time series and effectively capture the long-term dependencies in time. In this layer, LSTM inputs the time series data of surface settlement, soil moisture, and meteorological parameters into the network for training and identifying potential geological disaster risks.For example, when the soil moisture gradually increases and is combined with increased rainfall, LSTM can identify the disaster risks such as landslides or debris flows that may be caused by these patterns. After training, LSTM can output the preliminary identification results of geological disaster risks based on real-time data. After obtaining these two preliminary identification results, the system inputs this information into the comprehensive risk assessment layer for multi-source data fusion. This layer will combine the pollution risk identification results and the geological disaster risk identification results. First, it will calculate the correlation coefficient between the two types of risks through the correlation analysis algorithm, and then assign weights to each risk type based on the risk severity and influence range. The weighted fusion algorithm is used to comprehensively process the multi-source risk data, so as to comprehensively evaluate the overall risk status of the environment. For example, when both pollution risks and geological disaster risks exist, the system will analyze whether the soil heavy metal pollution will accelerate diffusion due to surface deformation, or whether the landslide caused by rainfall will lead to the migration of pollutants downstream. The interaction intensity between the two types of risks is quantitatively calculated through the risk coupling model, and combined with their respective severities, development trends and interaction impact factors, the comprehensive risk assessment results are output.

[0037] The system will perform a matching analysis by comparing the fusion risk assessment results with the historical risk event database. The purpose of this step is to conduct risk type discrimination and level assessment by comparing with historical risk events. The system can judge which type the current risk event belongs to and evaluate its risk level based on past case data. Through this matching analysis, the system can effectively evaluate the urgency of the current environmental risk event. For example, if historical data shows that similar pollution events have caused serious environmental damage, the system will rate the current event as a high-risk event and take corresponding warning measures.

[0038] After obtaining the risk matching results, the system executes semi-supervised learning algorithms and incremental learning algorithms to optimize the model. Semi-supervised learning allows the system to improve the model's recognition ability with the help of unlabeled data when only part of the labeled data is available. Incremental learning can continuously update the model parameters when new data arrives, and continuously improve the accuracy and adaptability of the model. With the support of these algorithms, the model can be gradually optimized to adapt to different environmental risk patterns, thus achieving more efficient risk identification.

[0039] The preprocessed data monitored in real time is input into the optimized risk identification model for risk status calculation. Through the real-time updated environmental monitoring data, the model can dynamically adjust the risk status and re-evaluate the levels of pollution risks and geological disaster risks. For example, at a certain point in time, if the gas concentration increases or the soil moisture changes sharply, the model will recalculate the pollution risk and geological disaster risk according to the optimized parameters and give new risk assessment results.

[0040] In a specific embodiment, the process of executing step S104 may specifically include the following steps: Execute the risk probability calculation and impact range assessment algorithms on the pollution risk identification result and the geological disaster risk identification result, perform risk level division, and obtain four-level risk warning data; Generate a data structure containing risk type description, affected area, warning level, development trend, and response suggestions based on the four-level risk warning data, perform warning information encapsulation, and obtain structured warning information; Perform spatial registration of the structured warning information with the basemap of the geographic information system, perform geographic coordinate mapping, and obtain geographically coordinated risk information; Execute the color coding algorithm and heat map generation algorithm on the geographically coordinated risk information, perform risk visualization rendering, and obtain a risk distribution heat map; Perform time series superposition processing on the risk distribution heat map, perform spatio-temporal evolution analysis, and obtain a spatio-temporal dynamic risk evolution sequence; Import the risk evolution sequence into a multi-scale display engine, perform view generation and interactive interface construction, and obtain a visual expression of risk distribution.

[0041] Specifically, after obtaining the pollution risk identification result and the geological disaster risk identification result, the system executes the risk probability calculation and impact range assessment algorithms. The main purpose of this process is to calculate the probability of risk occurrence and the possible impact range on the environment, personnel, facilities, etc. based on the identification results. For example, the probability calculation of pollution risk can be based on historical data of soil heavy metal concentration and gas concentration changes, combined with meteorological conditions and terrain information, to obtain the possibility of pollution events; while the impact range of geological disaster risk is determined by comprehensively evaluating factors such as surface deformation monitoring data, soil moisture, and rainfall, to determine the area that may be affected by landslides or debris flows. These calculation results will provide a basis for subsequent risk level division. Through risk level division, the system divides risks into four levels: green (no risk), yellow (slight risk), orange (medium risk), and red (severe risk). According to the differences in each level, the system will adopt different warning measures and response plans.

[0042] Next, based on the four-level risk warning data, the system generates a warning information structure containing detailed information. The warning information will include a description of the risk type (such as pollution risk or geological disaster risk), the scope of the affected area (such as inside the yard, around the yard, downstream water area, etc.), the warning level (such as green, yellow, orange, or red), the risk development trend (such as intensifying, slowing down, or remaining stable), and response suggestions (such as immediately taking protective measures, strengthening monitoring, evacuating personnel, etc.). After encapsulating these information, it will form structured warning information, which is convenient for subsequent system processing, transmission, and display.

[0043] On this basis, the system registers the structured early warning information spatially with the basemap of the Geographic Information System (GIS). The GIS basemap contains the geographical information of the yard area and its surrounding environment. Through geographical coordinate mapping, the system matches the early warning information with the actual geographical locations, mapping the risk data to specific spatial positions. This process ensures that the early warning information can accurately reflect the geographical distribution of risk events. Through the registered geo-coded risk information, the system can provide users with accurate spatial positioning and real-time environmental monitoring.

[0044] Immediately afterwards, the system performs color coding and heat map generation on the geo-coded risk information. The color coding algorithm maps different risk levels to different color ranges. For example, green indicates no risk, yellow indicates minor risk, orange indicates medium risk, and red indicates severe risk. Through color coding, the spatial distribution of risks becomes intuitive and easy to understand. Then, the system will use the heat map generation algorithm to render the risk information into a heat map, which can clearly display the risk intensity of different regions. For example, areas with higher risks will be highlighted in red, while areas with lower risks will be shown in green. The generation of the heat map enables users to see at a glance the risk distribution in different regions, helping decision-makers take timely countermeasures.

[0045] To further dynamically display the change process of risks, the system performs time series overlay processing on the risk distribution heat map for spatio-temporal evolution analysis. Time series overlay processing integrates data at multiple time points to present the change of the spatial distribution of risks over time, showing the dynamic evolution process of risks. This analysis can reveal the expansion or contraction of risk areas, the change trend of risk levels, and the impact of environmental conditions on risks. For example, if the risk level in a certain area continues to rise over a period of time, the system will promptly reflect it through the heat map, helping managers to warn of potential dangers.

[0046] The system imports the spatio-temporal dynamic risk evolution sequence into the multi-scale display engine to generate views and construct an interactive interface. The multi-scale display engine allows users to view the risk distribution from different perspectives, from macro to micro. For example, users can view the overall risk overview of the entire coal gangue yard and its surroundings, or focus on the detailed risk analysis of a small area. The construction of the interactive interface enables users to freely zoom in and out, drag the map, view risk information at different levels, and conduct in-depth analysis. This interactive view display enables decision-makers to flexibly view and analyze data according to specific needs, thus providing accurate basis for subsequent decision-making.

[0047] In a specific embodiment, the process of performing step S105 may specifically include the following steps: Match and retrieve the hierarchical early warning information and dynamic risk maps with the environmental risk response knowledge base for coal gangue yards, conduct a preliminary screening of response plans, and obtain a set of candidate response plans; Perform multi-objective optimization calculations on the set of candidate response plans in combination with the current yard conditions, available resources, and risk development trends, conduct plan evaluation and ranking, and obtain the ranked response plans; Construct a multi-scenario decision tree based on the ranked response plans. For each decision node, calculate the expected utility value using the weighted summation method according to the expected risk reduction degree, implementation success probability, and time benefit of the response measures. At the same time, calculate the resource consumption value by accumulating the human input cost, equipment usage cost, material consumption cost, and time cost. Evaluate the advantages and disadvantages of each decision path through the ratio of the expected utility value to the resource consumption value, and generate decision paths to obtain a decision branch network; Integrate and analyze the decision branch network with meteorological forecast data, engineering activity plans, and sensitive target distribution information, conduct future scenario deduction, and obtain a risk development prediction model; Execute the Monte Carlo simulation algorithm on each decision branch based on the risk development prediction model to evaluate the risk response effect and obtain the optimized result of the response plan; Convert the optimized result of the response plan into a structured document containing risk descriptions, response objectives, technical measures, resource allocation, personnel division of labor, and time nodes, generate a plan document, and obtain a customized emergency response plan.

[0048] Specifically, the hierarchical early warning information and dynamic risk maps are matched and retrieved with the environmental risk response knowledge base for coal gangue yards. The system matches the risk levels (such as pollution risk, geological disaster risk) and risk spatial distributions (i.e., dynamic risk maps) monitored in real time with the emergency plans in the preset environmental risk response knowledge base. The emergency knowledge base is a database containing standard response processes, resource requirements, and technical measures in different environmental risk situations. Through the matching and retrieval with the early warning information and risk maps, the system can quickly obtain a set of candidate response plans. These plan sets are standard emergency response strategies pre-designed for the currently identified risk types and risk levels.

[0049] The system conducts multi-objective optimization calculations on the candidate response plan set, combines the actual situation of the current yard, available resources, and risk development trends, and evaluates and ranks the response plans. In practical applications, each response plan will consider different objectives, such as the effect of risk control, the consumption of required resources, and the implementation difficulty. The multi-objective optimization calculation evaluates the advantages and disadvantages of each plan through a mathematical model, weighs the balance between different objectives, and finally outputs a ranked response plan. For example, if the yard faces a relatively serious pollution risk, the system may give priority to quickly taking pollution interception measures and select the most cost-saving plan when resources are limited. After multi-objective optimization, the system will obtain a set of response plans that have been evaluated and ranked.

[0050] Then, based on the ranked response plans, the system constructs a multi-scenario decision tree. At this stage, the decision-making process of each response plan is decomposed into multiple decision nodes. Each node represents a possible decision path under specific conditions, such as the decision to take different response measures at different risk levels. The system calculates the expected utility value for each decision node by using the weighted sum method according to the expected risk reduction degree, implementation success probability, and time benefit of the response measure. At the same time, the resource consumption value is calculated by accumulating based on the labor input cost, equipment usage cost, material consumption cost, and time cost. The advantages and disadvantages of each decision path are evaluated through the ratio of the expected utility value to the resource consumption value. These indicators will help decision-makers measure the effects and costs of different decisions and finally generate decision paths, thus obtaining a complete decision branch network. Through this decision tree, decision-makers can clearly see the possible consequences of each decision and the advantages and disadvantages of various choices.

[0051] After constructing the decision branch network, the system will integrate and analyze the meteorological forecast data, the engineering activity plan of the yard, and the sensitive target distribution information. The meteorological forecast data can provide early warnings of weather changes for decision-making. For example, an increase in rainfall may lead to an increased risk of landslides or debris flows; the engineering activity plan can reflect the impact that yard construction or other human activities may have on risks; the sensitive target distribution information can provide key area information (such as residential areas, important infrastructure) in the surrounding area of the yard. By integrating this information into the analysis of the decision tree, the system can conduct future scenario deduction and predict the development trend of risks under different scenarios. For example, if the weather forecast shows that heavy rain is about to occur, the system may adjust the response plan and strengthen the reinforcement work of the yard slope to avoid landslide disasters caused by rainfall.

[0052] Based on the predicted risk development trend, the system uses the Monte Carlo simulation algorithm to evaluate the risk response effects of each decision branch. Monte Carlo simulation is an algorithm that simulates multiple possible scenarios through random sampling and calculates their results. Through Monte Carlo simulation, the system can simulate the effects of different response strategies under various uncertain factors (such as weather changes, resource limitations), and evaluate whether each decision path can effectively reduce risks under possible future changes. Monte Carlo simulation can provide a more comprehensive and reliable evaluation of response strategies for decision-makers, ensuring that the selected response plan can address multiple potential uncertain risks.

[0053] The system converts the optimized response plan into an emergency response plan document containing detailed information. This document includes risk descriptions, response objectives, specific technical measures, resource allocation, personnel division of labor, and time nodes. The emergency response plan will generate specific implementation plans based on the optimization results to ensure rapid and effective response when facing risks. For example, for a high-risk area, the system may recommend immediately deploying personnel for environmental restoration, while allocating equipment for pollutant interception, and ensuring that relevant personnel complete tasks on time. The emergency response plan also includes a resource allocation plan, such as the allocation strategies for funds, manpower, and equipment, to ensure the smooth execution of the plan.

[0054] In a specific embodiment, the process of executing step S106 may specifically include the following steps: Compare the environmental monitoring data before and after the execution of the emergency response plan, calculate the response timeliness, risk control effectiveness, resource utilization efficiency, and secondary risk prevention and control indicators, conduct an effect evaluation, and obtain the risk response performance evaluation result; Based on the risk response performance evaluation result, identify the performance deficiencies of each module of the system, generate improvement strategies, and obtain optimization suggestions including sensor layout adjustment, sampling frequency optimization, algorithm improvement, and threshold calibration; Store the complete data of the risk event, the response process, and the effect evaluation result in the historical case database, conduct case induction, and obtain structured risk response experience data; Execute the association rule mining algorithm and the sequence pattern mining algorithm on the structured risk response experience data, conduct knowledge extraction, and obtain risk pattern features and the best response strategies; Based on the risk pattern features and the best response strategies, perform parameter adjustment and structural optimization on the environmental risk multimodal recognition model, conduct model update, and obtain an optimized risk recognition model; Integrate the optimized risk recognition model with the optimized sensor network layout, data processing algorithm, and decision support system, conduct system integration, and obtain an optimized risk monitoring system.

[0055] Specifically, before and after the implementation of the emergency response plan, the system compares environmental monitoring data and calculates multiple performance indicators. These indicators include response timeliness, risk control effectiveness, resource utilization efficiency, and secondary risk prevention and control indicators. Response timeliness refers to the time difference from the occurrence of a risk to the implementation of response measures. The shorter the response time, the stronger the system's emergency response ability. Risk control effectiveness measures the change in the risk level after response measures are taken to evaluate whether the measures effectively reduce the risk. Resource utilization efficiency focuses on the use of resources (such as manpower, equipment, funds, etc.) during the emergency response process to ensure the rationality of resource allocation. The secondary risk prevention and control indicator evaluates whether new risks or unforeseen risk events are avoided during the emergency response process. For example, in the response to certain environmental risks, if enhanced monitoring and real-time data analysis can avoid secondary pollution or geological disasters, it indicates that the secondary risk prevention and control measures have been effectively implemented.

[0056] By calculating these performance indicators, the risk response performance evaluation results obtained by the system will reveal the overall effect of the emergency response plan. If the performance is poor in certain indicators, the system will identify possible deficiencies and provide a basis for subsequent optimization.

[0057] Based on the risk response performance evaluation results, the system will identify the performance deficiencies of each module and generate improvement strategies. For example, if the accuracy of the monitoring sensor is insufficient, resulting in inaccurate data, the system may recommend adjusting the sensor layout and optimizing the sensor installation location to ensure more accurate capture of key data. If the data collection frequency cannot meet the real-time requirements of the emergency response, the system may recommend increasing the sampling frequency or making dynamic adjustments according to different risk types and changes. In addition, if the data processing algorithm is inefficient or the existing threshold setting does not adapt to the actual risk changes, the system will propose suggestions for algorithm improvement and threshold calibration. These optimization suggestions will help improve the performance of the entire system and ensure more efficient responses to future risks.

[0058] Subsequently, the system stores the complete data of the risk event, the response process, and the effect evaluation results in the historical case database. This database will contain various data accumulated during multiple emergency responses, including environmental monitoring data, risk assessment results, implemented emergency measures, resource allocation, and effect evaluation results, etc. By storing this information in a structured manner, the system can provide valuable historical data support for future emergency responses and help analyze and summarize response experiences in different scenarios.

[0059] For the structured data stored in the historical database, the system will use association rule mining algorithms and sequential pattern mining algorithms for knowledge extraction. Association rule mining algorithms can identify patterns and associations that often occur in different emergency responses. For example, when a specific type of pollution risk appears, which emergency measures are often the most effective. Sequential pattern mining algorithms can extract risk pattern features from time series data, such as the evolution process of a certain geological disaster risk, how it changes and develops over time, and which factors play a decisive role in the development of the risk at different time periods. Through these algorithms, the system can extract the best coping strategies from historical data to help optimize future emergency response measures.

[0060] Based on the risk pattern features and the best coping strategies extracted from historical data, the system will adjust the parameters and optimize the structure of the multi-modal recognition model for environmental risks. The purpose of this optimization process is to improve the recognition accuracy and adaptability of the model when facing newly emerging risks. For example, if the system identifies certain new pollution sources or risk types, the model can automatically adjust the parameters according to historical data to improve the recognition ability for these new types of risks. Through this optimization, the risk recognition model can continuously improve its accuracy and efficiency during long-term operation.

[0061] The risk recognition system after model optimization will be integrated with the optimized sensor network layout, data processing algorithms, and decision support system. The optimization of the sensor network layout will ensure that the monitoring system can cover all key areas and avoid information blind spots; the optimization of the data processing algorithms will improve the processing speed and accuracy of the data, ensuring that effective data support can be obtained in a timely manner during the emergency response process; the optimization of the decision support system will provide more scientific and reasonable decision-making basis for managers to help them make decisions quickly in case of emergencies. Through this integration, the final result will be an optimized risk monitoring system that can more efficiently and accurately respond to various environmental risks that may occur in the future.

[0062] The above described the method for real-time monitoring of environmental risks in a coal gangue yard based on edge computing in the embodiments of the present application. Next, the system for real-time monitoring of environmental risks in a coal gangue yard based on edge computing in the embodiments of the present application will be described. Please refer to Figure 2 One embodiment of the system for real-time monitoring of environmental risks in a coal gangue yard based on edge computing in the embodiments of the present application includes: An acquisition module 201, configured to deploy a multi-source sensor network for data acquisition on the coal gangue yard and its surrounding environment to obtain original monitoring data of environmental risks; A verification module 202, configured to input the original monitoring data of environmental risks into an edge computing node for data verification, compensation correction, normalization, and feature extraction processing to obtain preprocessed data; A training module 203, configured to build an environmental risk multimodal recognition model based on the preprocessed data and perform training to obtain a pollution risk recognition result and a geological disaster risk recognition result; A generation module 204, configured to generate hierarchical early warning information according to the pollution risk recognition result and the geological disaster risk recognition result and construct a dynamic risk map to obtain a visual expression of risk distribution; An input module 205, configured to input the hierarchical early warning information and the dynamic risk map into an intelligent decision support system to generate emergency response suggestions and obtain a customized emergency response plan; An update module 206, configured to perform system evaluation and knowledge accumulation based on the execution data of the emergency response plan, and perform parameter update on the environmental risk multimodal recognition model to obtain an optimized risk monitoring system.

[0063] Through the collaborative cooperation of the above-mentioned components, by deploying a multi-source sensor network in and around the coal gangue yard, a comprehensive, continuous, and real-time monitoring of environmental risks has been achieved. Compared with the traditional single-parameter and low-frequency monitoring methods, the monitoring range and depth have been significantly expanded; by inputting the original environmental risk monitoring data into the edge computing node for local processing, the data transmission delay problem caused by the traditional centralized architecture has been solved, the response time has been shortened by 85%, and the reaction speed of the monitoring system to sudden risk events has been greatly improved; the environmental risk multimodal recognition model constructed based on the preprocessed data integrates the recognition functions of pollution risk and geological disaster risk, overcomes the limitations of the separated design of risk monitoring in the traditional system, and improves the recognition ability of compound risks. Among them, the convolutional neural network is used to extract features and classify pollution risk data, and the long short-term memory network is used to analyze the time series pattern of geological disaster risk data. These two algorithms are optimized for spatial features and time series features respectively, fully adapting to the multi-dimensional characteristics of the environmental risk data in the coal gangue yard; the hierarchical early warning information and the dynamic risk map generated according to the risk recognition result provide a spatial visual expression of the risk, intuitively showing the risk distribution and evolution trend, which is convenient for managers to quickly grasp the risk situation; the customized emergency response plan generated by the intelligent decision support system based on the early warning information and the risk map provides precise guidance for risk response. Among them, the multi-objective optimization algorithm and the Monte Carlo simulation algorithm comprehensively consider multiple dimensions such as response timeliness, resource utilization efficiency, and risk control effect. Through the analysis and knowledge accumulation of the emergency response execution data, the system realizes self-adaptive optimization and continuously improves the risk recognition and prevention and control capabilities.

[0064] Above Figure 2From the perspective of modular functional entities, the real-time monitoring system for the environmental risks of coal gangue yards based on edge computing in the embodiments of the present invention is described in detail. Next, the real-time monitoring device for the environmental risks of coal gangue yards based on edge computing in the embodiments of the present invention is described in detail from the perspective of hardware processing.

[0065] Figure 3 FIG. 4 is a schematic structural diagram of a real-time monitoring device for the environmental risks of coal gangue yards based on edge computing provided by an embodiment of the present invention. The real-time monitoring device 300 for the environmental risks of coal gangue yards based on edge computing may vary greatly due to different configurations or performances, and may include one or more processors (central processing units, CPUs) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 for storing application programs 333 or data 332 (for example, one or more mass storage device terminals). Among them, the memory 320 and the storage media 330 may be transient storage or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the real-time monitoring device 300 for the environmental risks of coal gangue yards based on edge computing. Further, the processor 310 may be configured to communicate with the storage media 330 and execute a series of instruction operations in the storage media 330 on the real-time monitoring device 300 for the environmental risks of coal gangue yards based on edge computing to implement the steps of the above-mentioned real-time monitoring method for the environmental risks of coal gangue yards based on edge computing.

[0066] The real-time monitoring device 300 for the environmental risks of coal gangue yards based on edge computing may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, and so on. Those skilled in the art can understand that Figure 3 The shown structural diagram of the real-time monitoring device for the environmental risks of coal gangue yards based on edge computing does not limit the real-time monitoring device for the environmental risks of coal gangue yards based on edge computing provided by the present invention, and may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.

[0067] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions run on a computer, the computer is caused to execute the steps of the method for real-time monitoring of the environmental risk of a coal gangue yard based on edge computing.

[0068] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, system, and unit can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0069] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to cause a device for real-time monitoring of the environmental risk of a coal gangue yard based on edge computing (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A real-time monitoring method for environmental risks in a coal gangue yard based on edge computing, characterized in that, The method includes: Deploying a multi-source sensor network for data collection on the coal gangue yard and its surrounding environment to obtain original environmental risk monitoring data; Inputting the original environmental risk monitoring data into an edge computing node for data verification, compensation correction, normalization, and feature extraction processing to obtain preprocessed data; Constructing and training an environmental risk multi-modal recognition model based on the preprocessed data to obtain pollution risk recognition results and geological disaster risk recognition results, including: constructing a pollution risk recognition layer based on the preprocessed data, inputting soil heavy metal detection data, water quality parameter data, and gas concentration data into a convolutional neural network for feature extraction and classification to obtain preliminary pollution risk recognition results; constructing a geological disaster risk recognition layer based on the preprocessed data, inputting surface deformation monitoring data, soil humidity data, and meteorological parameter data into a long short-term memory network for time series pattern analysis to obtain preliminary geological disaster risk recognition results; inputting the preliminary pollution risk recognition results and the preliminary geological disaster risk recognition results into a comprehensive risk assessment layer, performing correlation analysis and weight allocation on the two types of risk results through a weighted fusion algorithm for multi-source data fusion to obtain a fused risk assessment result; performing matching analysis between the fused risk assessment result and a historical risk event database for risk type discrimination and level assessment to obtain a risk matching result; performing a semi-supervised learning algorithm and an incremental learning algorithm on the risk matching result for model parameter update to obtain an optimized risk recognition model; inputting the real-time monitored preprocessed data into the optimized risk recognition model for risk status calculation to obtain the pollution risk recognition result and the geological disaster risk recognition result; Generating hierarchical early warning information based on the pollution risk recognition result and the geological disaster risk recognition result and constructing a dynamic risk map to obtain a visual expression of risk distribution; Inputting the hierarchical early warning information and the dynamic risk map into an intelligent decision support system to generate emergency response suggestions to obtain a customized emergency response plan; Performing system evaluation and knowledge accumulation based on the execution data of the emergency response plan, and performing parameter update on the environmental risk multi-modal recognition model to obtain an optimized risk monitoring system.

2. The real-time monitoring method for the environmental risk of a coal gangue yard based on edge computing according to claim 1, wherein The deploying a multi-source sensor network for data collection on the coal gangue yard and its surrounding environment to obtain original environmental risk monitoring data includes: Installing soil temperature sensors, soil humidity sensors, and gas concentration sensors at different depths inside the coal gangue yard to collect internal parameters of the yard to obtain temperature distribution data, water content change data, and harmful gas concentration data; Deploying rapid heavy metal content detection sensors on the surface of the coal gangue yard to detect element contents in the surface soil to obtain lead, cadmium, and arsenic heavy metal element content data; Deploying surface deformation monitoring devices around the coal gangue yard to monitor surface changes in real time to obtain surface settlement data, landslide omen data, and debris flow omen data; Water quality monitoring sensors are deployed in the water body downstream of the coal gangue yard to detect the physical and chemical properties of the water body, and pH value data, conductivity data, and dissolved oxygen data are obtained; Meteorological parameter collection devices are deployed around the coal gangue yard to record the environmental meteorological conditions, and rainfall data, wind speed data, and air pressure data are obtained; The temperature distribution data, water content change data, harmful gas concentration data, heavy metal element content data, ground settlement data, landslide omen data, debris flow omen data, pH value data, conductivity data, dissolved oxygen data, rainfall data, wind speed data, and air pressure data are summarized and integrated to obtain the original environmental risk monitoring data.

3. The real-time monitoring method for the environmental risk of a coal gangue yard based on edge computing according to claim 1, wherein, The original environmental risk monitoring data is input into the edge computing node for data verification, compensation and correction, standardization, and feature extraction processing to obtain preprocessed data, including: Set threshold ranges and change rate inspection rules for the original environmental risk monitoring data, perform data verification processing, and obtain verified data after removing outliers; Perform compensation and correction calculations on the verified data according to the sensor drift compensation formula and temperature influence correction model to obtain corrected monitoring data; Perform dimension conversion and unit unification operations on the corrected monitoring data for standardization processing to obtain uniformly formatted standard data; Execute the timestamp calibration algorithm on the standard data for time synchronization processing to obtain synchronized data with consistent time series; Input the synchronized data into the wavelet transform filtering algorithm and the moving average filtering algorithm for data denoising processing to obtain denoised smooth data; Calculate statistical features such as mean, variance, and change rate of the smooth data and extract frequency domain features through fast Fourier transform for feature extraction processing to obtain the preprocessed data.

4. The real-time monitoring method for environmental risks of coal gangue yards based on edge computing according to claim 1, wherein Generate hierarchical warning information based on the pollution risk identification results and the geological disaster risk identification results and construct a dynamic risk map to obtain a visual expression of the risk distribution, including: Execute risk probability calculation and impact range assessment algorithms on the pollution risk identification results and the geological disaster risk identification results for risk level division to obtain four-level risk warning data; Generate a data structure containing risk type descriptions, affected areas, warning levels, development trends, and response suggestions based on the four-level risk warning data for warning information encapsulation to obtain structured warning information; Perform spatial registration of the structured warning information with the base map of the geographic information system for geographic coordinate mapping to obtain geographically coordinated risk information; Execute color coding algorithms and heat map generation algorithms on the geographically coordinated risk information for risk visualization rendering to obtain a risk distribution heat map; Perform time series superposition processing on the risk distribution heat map for spatio-temporal evolution analysis to obtain a spatio-temporal dynamic risk evolution sequence; Import the risk evolution sequence into a multi-scale display engine for view generation and interactive interface construction to obtain the visual expression of the risk distribution.

5. The real-time monitoring method for environmental risks of coal gangue yards based on edge computing according to claim 1, wherein, Input the hierarchical warning information and the dynamic risk map into the intelligent decision support system to generate emergency response suggestions to obtain a customized emergency response plan, including: Match and retrieve the classified early warning information and the dynamic risk map with the knowledge base for coping with the environmental risks of coal gangue yards, conduct a preliminary screening of coping plans, and obtain a set of candidate coping plans; Perform multi-objective optimization calculations on the set of candidate coping plans in combination with the current yard conditions, available resources, and risk development trends, conduct plan evaluation and ranking, and obtain the ranked coping plans; Construct a multi-scenario decision tree based on the ranked coping plans, calculate the expected utility value for each decision node by using the weighted summation method according to the expected risk reduction degree, implementation success probability, and time benefit of the coping measures. At the same time, calculate the resource consumption value by accumulating the human input cost, equipment usage cost, material consumption cost, and time cost. Evaluate the advantages and disadvantages of each decision path through the ratio of the expected utility value to the resource consumption value, generate decision paths, and obtain a decision branch network; Integrate and analyze the decision branch network with meteorological forecast data, engineering activity plans, and sensitive target distribution information, conduct future scenario deduction, and obtain a risk development prediction model; Execute the Monte Carlo simulation algorithm for each decision branch based on the risk development prediction model, conduct an evaluation of the risk coping effect, and obtain the optimized result of the coping plan; Convert the optimized result of the coping plan into a structured document including risk description, coping objectives, technical measures, resource allocation, personnel division of labor, and time nodes, generate a plan document, and obtain the customized emergency response plan; 6. The real-time monitoring method for the environmental risk of a coal gangue yard based on edge computing according to claim 1, characterized in that Conduct system evaluation and knowledge accumulation based on the execution data of the emergency response plan, and update the parameters of the environmental risk multi-modal recognition model to obtain an optimized risk monitoring system, including: Compare the environmental monitoring data before and after the execution of the emergency response plan, calculate the response timeliness, risk control effectiveness, resource utilization efficiency, and secondary risk prevention and control indicators, conduct an effect evaluation, and obtain the risk coping performance evaluation result; Identify the performance deficiencies of each module of the system based on the risk coping performance evaluation result, generate improvement strategies, and obtain optimization suggestions including sensor layout adjustment, sampling frequency optimization, algorithm improvement, and threshold calibration; Store the complete data of risk events, coping processes, and effect evaluation results in the historical case database, conduct case induction, and obtain structured risk coping experience data; Execute the association rule mining algorithm and the sequential pattern mining algorithm on the structured risk coping experience data, conduct knowledge extraction, and obtain risk pattern features and the best coping strategies; Based on the risk pattern features and the best coping strategies, adjust the parameters and optimize the structure of the environmental risk multi-modal recognition model, conduct model update, and obtain an optimized risk recognition model; Integrate the optimized risk recognition model with the optimized sensor network layout, data processing algorithm, and decision support system, conduct system integration, and obtain the optimized risk monitoring system; 7. A real-time monitoring system for environmental risks in a coal gangue yard based on edge computing, characterized in that, For implementing the real-time monitoring method for environmental risks of coal gangue yards based on edge computing as described in any one of claims 1-6, the real-time monitoring system for environmental risks of coal gangue yards based on edge computing includes: The acquisition module is used to deploy a multi-source sensor network for the coal gangue yard and its surrounding environment to collect data, and obtain the original environmental risk monitoring data; The verification module is used to input the original environmental risk monitoring data into the edge computing node for data verification, compensation and correction, standardization and feature extraction processing, and obtain the preprocessed data; The training module is used to construct an environmental risk multi-modal recognition model based on the preprocessed data and perform training to obtain the pollution risk recognition result and the geological disaster risk recognition result, including: constructing a pollution risk recognition layer based on the preprocessed data, inputting the soil heavy metal detection data, water quality parameter data and gas concentration data into a convolutional neural network for feature extraction and classification to obtain a preliminary pollution risk recognition result; constructing a geological disaster risk recognition layer based on the preprocessed data, inputting the surface deformation monitoring data, soil humidity data and meteorological parameter data into a long short-term memory network for time series pattern analysis to obtain a preliminary geological disaster risk recognition result; inputting the preliminary pollution risk recognition result and the preliminary geological disaster risk recognition result into the comprehensive risk assessment layer, performing correlation analysis and weight allocation on the two types of risk results through a weighted fusion algorithm, performing multi-source data fusion to obtain a fusion risk assessment result; performing matching analysis on the fusion risk assessment result and the historical risk event database, performing risk type discrimination and level assessment to obtain a risk matching result; performing a semi-supervised learning algorithm and an incremental learning algorithm on the risk matching result to update the model parameters to obtain an optimized risk recognition model; inputting the preprocessed data of real-time monitoring into the optimized risk recognition model to calculate the risk status and obtain the pollution risk recognition result and the geological disaster risk recognition result; The generation module is used to generate hierarchical early warning information according to the pollution risk recognition result and the geological disaster risk recognition result and construct a dynamic risk map to obtain a visual expression of risk distribution; The input module is used to input the hierarchical early warning information and the dynamic risk map into the intelligent decision support system to generate emergency response suggestions and obtain a customized emergency response plan; The update module is used to perform system evaluation and knowledge accumulation based on the execution data of the emergency response plan, and update the parameters of the environmental risk multi-modal recognition model to obtain an optimized risk monitoring system.

8. A real-time monitoring device for the environmental risks of a coal gangue yard based on edge computing, characterized in that, It includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the method for real-time monitoring of the environmental risk of the coal gangue yard based on edge computing according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, the processor is caused to execute the method for real-time monitoring of the environmental risk of the coal gangue yard based on edge computing according to any one of claims 1 to 6.

Images