Karst water disaster prediction method and device based on cloud processing

Abstract

The application discloses a karst water disaster prediction method and device based on cloud processing, and relates to the technical field of intelligent detection. The method comprises the following steps: deploying a karst water disaster monitoring Internet of Things in a monitoring area, collecting real-time multi-source heterogeneous monitoring data of the monitoring area based on the karst water disaster monitoring Internet of Things, and transmitting the real-time multi-source heterogeneous monitoring data to a cloud server; cleaning the real-time multi-source heterogeneous monitoring data on the cloud server, and performing dynamic time window alignment based on mutual information analysis to obtain preprocessed real-time standard monitoring data; inputting the preprocessed real-time standard monitoring data into a karst water disaster prediction model constructed based on a deep learning algorithm and a multi-objective optimization algorithm, performing karst water disaster prediction, and obtaining real-time karst water disaster prediction results; and performing graded early warning release according to the real-time karst water disaster prediction results. The application solves the problems of poor multi-source data fusion, difficult time lag alignment, difficult model training under a small sample, and difficult balance between cloud real-time performance and accuracy in the prior art.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology, and in particular to a method and device for predicting karst water hazards based on cloud processing. Background Technology

[0002] Karst water hazards are one of the major geological disasters faced during the construction and operation of tunnels and underground engineering projects. Due to the complexity and concealment of karst geological conditions, water inrush and mudslide accidents are often characterized by their suddenness, large water inflow, and high destructive power, seriously threatening construction safety and causing huge economic losses.

[0003] Existing methods for monitoring karst water hazards mainly rely on threshold alarms from single sensors (such as water level gauges and rain gauges). This method has the following limitations: 1) Low integration of multi-source data: Multi-source heterogeneous data such as hydrogeology, meteorology and environment, and microseismic monitoring are often analyzed independently, making it difficult to reflect the inherent coupling relationship between various physical quantities.

[0004] 2) Improper handling of time lags: The impact of surface meteorological factors such as rainfall on groundwater levels has a lag, and traditional methods are difficult to accurately align time-series data, resulting in a decrease in prediction accuracy.

[0005] 3) Limited sample size and poor model generalization ability: Karst water inrush accidents are low-probability events, and historical label data is scarce, making it difficult to train deep learning-based prediction models and prone to overfitting.

[0006] 4) The issue of balancing real-time performance and computing power in the cloud: When processing massive amounts of real-time data in the cloud, a single fixed prediction model is difficult to dynamically balance prediction accuracy and response speed according to the server load, which may lead to delays in early warning under high load.

[0007] Therefore, there is an urgent need for a cloud-based method for predicting karst water hazards that can integrate multi-source heterogeneous data, solve the problem of sample scarcity, and have adaptive scheduling capabilities. Summary of the Invention

[0008] This invention provides a method and apparatus for predicting karst water hazards based on cloud processing. This invention solves the problems of poor multi-source data fusion, difficulty in time-delay alignment, difficulty in model training with small samples, and the difficulty in balancing real-time performance and accuracy in the cloud.

[0009] In a first aspect, embodiments of the present invention provide a method for predicting karst water hazards based on cloud processing, the method comprising: Deploy a karst water hazard monitoring Internet of Things (IoT) in the monitoring area, collect real-time multi-source heterogeneous monitoring data of the monitoring area based on the karst water hazard monitoring IoT, and transmit it to the cloud server; The real-time multi-source heterogeneous monitoring data is cleaned on the cloud server, and dynamic time window alignment is performed based on mutual information analysis to obtain preprocessed real-time standard monitoring data. The preprocessed real-time standard monitoring data is input into the karst water hazard prediction model constructed based on deep learning algorithm and multi-objective optimization algorithm to predict karst water hazards and obtain real-time karst water hazard prediction results. Based on real-time karst water hazard prediction results, graded early warnings are issued.

[0010] The technical solution provided in this application has at least the following beneficial effects: By deploying an IoT system for karst flood monitoring, comprehensive real-time acquisition and cloud aggregation of multi-source heterogeneous data, including hydrological, meteorological, and microseismic data, were achieved, resulting in rich data dimensions. Utilizing dynamic time window alignment technology based on mutual information analysis, the optimal lag time between meteorological and hydrological data was automatically calculated, resolving the temporal misalignment problem caused by the rainfall infiltration lag effect and improving data consistency. A data augmentation model based on physical constraints was introduced, not only addressing the scarcity of karst flood samples but also ensuring that the generated samples conformed to physical laws by constraining the generated data through Darcy's law, thus improving the training quality of the karst flood prediction model. A multi-objective fruit fly optimization algorithm was used to optimize the hyperparameters of the karst flood prediction model, obtaining a Pareto solution set that balances accuracy and speed. Different models (high-precision, balanced, and ultra-fast) were adaptively scheduled according to cloud load, effectively balancing the accuracy and real-time performance of early warnings. A tiered early warning release mechanism was established, automatically triggering different levels of warnings based on prediction results, achieving closed-loop management from data acquisition and intelligent prediction to emergency response.

[0011] In one alternative implementation, the real-time multi-source heterogeneous monitoring data includes real-time hydrogeological monitoring data, real-time meteorological and environmental monitoring data, and real-time microseismic monitoring data.

[0012] In one alternative implementation, a karst flood hazard monitoring IoT is deployed in the monitoring area. Based on the karst flood hazard monitoring IoT, real-time multi-source heterogeneous monitoring data of the monitoring area is collected and transmitted to a cloud server, including: High-precision piezometers and ultrasonic flow meters were installed in the advanced boreholes at the tunnel face, in the karst fissures exposed in the roadway, and at the location of key aquifers in the monitoring area to collect hydrogeological monitoring data. Rain gauges and temperature and humidity sensors were installed in the surface recharge area and at the entrance of the project in the monitoring area to collect meteorological and environmental monitoring data. An array of microseismic sensors was deployed on the surface of the monitoring area to collect microseismic monitoring data. Connect high-precision piezometers, ultrasonic flow meters, rain gauges, temperature and humidity sensors, and micro-vibration pickup arrays to the IoT gateway within the communication range, and connect the IoT gateway to the cloud server to build an IoT for monitoring karst water hazards. Based on the Internet of Things (IoT) for monitoring karst water hazards, real-time multi-source heterogeneous monitoring data of the monitoring area is collected and aggregated to the IoT gateway. The IoT gateway encrypts and transmits real-time multi-source heterogeneous monitoring data to the cloud server.

[0013] In one optional implementation, real-time multi-source heterogeneous monitoring data is cleaned on a cloud server, and dynamic time window alignment is performed based on mutual information analysis to obtain preprocessed real-time standard monitoring data, including: Using 3 on a cloud server σ The criteria perform preliminary screening of real-time multi-source heterogeneous monitoring data, removing obvious outliers caused by signal interference, and obtaining pre-screened real-time multi-source heterogeneous monitoring data. The five-number summary method was used to identify and process abnormal data in the real-time multi-source heterogeneous monitoring data after preliminary screening, and cleaned real-time multi-source heterogeneous monitoring data was obtained. Based on mutual information analysis, the real-time meteorological and environmental monitoring data and the real-time hydrogeological monitoring data after cleaning are dynamically aligned in time windows to obtain the aligned real-time meteorological and environmental monitoring data. By integrating real-time hydrogeological monitoring data after cleaning, real-time meteorological and environmental monitoring data after alignment, and real-time microseismic monitoring data after cleaning, pre-processed real-time standard monitoring data is obtained.

[0014] In one optional implementation, based on mutual information analysis, dynamic time-window alignment is performed on the cleaned real-time meteorological and environmental monitoring data and the cleaned real-time hydrogeological monitoring data in the cleaned real-time multi-source heterogeneous monitoring data to obtain aligned real-time meteorological and environmental monitoring data, including: Based on the preset sliding window length, the pre-processed historical standard monitoring data sequence, the cleaned historical hydrogeological monitoring data sequence, and the corresponding aligned historical meteorological and environmental monitoring data sequence are extracted from the historical database obtained by manual processing on the cloud server. The cleaned historical hydrogeological monitoring data sequence and the aligned historical meteorological and environmental monitoring data sequence were normalized to obtain the normalized historical hydrogeological monitoring data sequence and the normalized historical meteorological and environmental monitoring data sequence. By iterating through the lag times, the mutual information of the normalized historical hydrogeological monitoring data series and the normalized historical meteorological and environmental monitoring data series is calculated using the following formula:

[0015] In the formula, For mutual information content; This is a normalized historical hydrogeological monitoring data sequence; This is a normalized sequence of historical meteorological and environmental monitoring data; The time lag is the time delay. These are the historical hydrogeological monitoring data points in the normalized historical hydrogeological monitoring data sequence; These are the historical meteorological and environmental monitoring data points in the normalized historical meteorological and environmental monitoring data sequence. For joint probability density; Probability density of historical hydrogeological monitoring data; The probability density of historical meteorological and environmental monitoring data; Find the optimal lag time that maximizes mutual information. ; The cleaned real-time meteorological and environmental monitoring data in the cleaned real-time multi-source heterogeneous monitoring data are shifted backward by the optimal lag time. We obtained aligned real-time meteorological and environmental monitoring data.

[0016] In one optional implementation, the karst water hazard prediction model is constructed based on the CNN-LSTM-MLP algorithm, and the karst water hazard prediction model includes a spatial feature extraction module constructed based on the CNN algorithm, a temporal feature extraction module constructed based on the LSTM algorithm, and a karst water hazard prediction module constructed based on the MLP algorithm.

[0017] In one alternative implementation, a karst flood hazard prediction model is constructed based on deep learning algorithms and multi-objective optimization algorithms, including: Extract a small batch of pre-processed historical standard monitoring data with real labels from the historical database obtained by manual processing on the cloud server. Using a pre-built data augmentation model based on physical constraints, several generated standard monitoring data sets with real labels are generated, and combined with several pre-processed historical standard monitoring data sets, a large number of training sample sets are obtained. An initial karst water hazard prediction model was constructed using the CNN-LSTM-MLP algorithm. Based on the training sample set, the hyperparameters of the initial karst water hazard prediction model are optimized using the multi-objective fruit fly optimization algorithm, resulting in the Pareto solution set of the optimal hyperparameters for different application scenarios and the corresponding final karst water hazard prediction model. The application scenarios include high-precision, balanced, and high-speed models. The model parameters of the final karst water hazard prediction model for each application scenario are packaged into Docker containers and deployed in the inference engine of the cloud server.

[0018] In one alternative implementation, based on the training sample set, a multi-objective fruit fly optimization algorithm is used to optimize the hyperparameters of the initial karst flood prediction model, obtaining a Pareto solution set of optimal hyperparameters for different application scenarios and the corresponding final karst flood prediction model, including: Define the multi-objective optimization function of the multi-objective fruit fly optimization algorithm, and encode the hyperparameters of the initial karst water hazard prediction model into individual vectors of the multi-objective fruit fly optimization algorithm; The chaotic sequence is generated using the Logistic mapping, and then mapped to the solution space of individual fruit flies to obtain the initial fruit fly population. Calculate the initial distance of each initial fruit fly individual in the initial fruit fly population, and obtain the initial taste concentration judgment value corresponding to the initial distance; The initial taste concentration judgment value is converted into a candidate hyperparameter through a linear mapping function, and the candidate karst water hazard prediction model is constructed by training with a training sample set. The optimization objective value of the candidate karst water hazard prediction model is calculated using a multi-objective optimization function. Based on the optimization target value, the initial fruit fly population is subjected to rapid non-dominated sorting to obtain different frontier levels; The initial fruit fly population is iteratively updated to obtain an updated fruit fly population; Using a multi-objective optimization function, the optimization objective value for each updated fruit fly individual in the updated fruit fly population is calculated, and the frontier level is updated based on the optimization objective value. Until the number of iterations reaches the iteration threshold, the first frontier level of the last iteration, which includes several optimal hyperparameters, is taken as the Pareto solution set, and the final karst water hazard prediction model corresponding to each solution in the Pareto solution set is saved.

[0019] In one optional implementation, preprocessed real-time standard monitoring data is input into a karst water hazard prediction model constructed based on deep learning algorithms and multi-objective optimization algorithms to predict karst water hazards, obtaining real-time karst water hazard prediction results, including: Collect real-time operational monitoring data of the cloud server; the real-time operational monitoring data includes the real-time CPU utilization and the real-time concurrent request queue length of the cloud server; Based on real-time operation monitoring data, the karst water hazard prediction model, constructed using deep learning algorithms and multi-objective optimization algorithms, is adaptively scheduled for the corresponding application scenarios. The preprocessed real-time standard monitoring data is input into the scheduled karst water hazard prediction model to perform karst water hazard prediction and obtain real-time karst water hazard prediction results; the real-time karst water hazard prediction results include real-time water inrush probability and real-time estimated water inflow.

[0020] Secondly, embodiments of the present invention provide a cloud-based karst water hazard prediction device for implementing a karst water hazard prediction method. The device includes: The data acquisition unit is used to deploy a karst water hazard monitoring Internet of Things (IoT) in the monitoring area. Based on the karst water hazard monitoring IoT, it collects real-time multi-source heterogeneous monitoring data of the monitoring area and transmits it to the cloud server. The data preprocessing unit is used to clean real-time multi-source heterogeneous monitoring data on the cloud server and perform dynamic time window alignment based on mutual information analysis to obtain preprocessed real-time standard monitoring data. The karst water hazard prediction unit is used to input preprocessed real-time standard monitoring data into the karst water hazard prediction model constructed based on deep learning algorithm and multi-objective optimization algorithm to predict karst water hazards and obtain real-time karst water hazard prediction results. The tiered early warning release unit is used to release tiered early warnings based on real-time karst water hazard prediction results.

[0021] A third aspect of this invention provides an electronic device, which includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, such that the at least one processor can perform the method proposed in the first aspect of the present invention.

[0022] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the steps of a cloud-based karst water hazard prediction method provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the functional units of a cloud-based karst water hazard prediction device provided in an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0025] The present invention will be further described below with reference to the accompanying drawings.

[0026] Reference Figure 1 , Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention.

[0027] like Figure 1 As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0028] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0029] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and an electronic program for a cloud-based karst flood hazard prediction device.

[0030] exist Figure 1 In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device. The electronic device calls the electronic program of the cloud-based karst water hazard prediction device stored in the memory 1005 through the processor 1001, and executes the cloud-based karst water hazard prediction method provided in the embodiment of the present invention.

[0031] Reference Figure 2 The present invention provides a cloud-based method for predicting karst water hazards, the method comprising: S201: Deploy a karst water hazard monitoring IoT in the monitoring area, collect real-time multi-source heterogeneous monitoring data of the monitoring area based on the karst water hazard monitoring IoT, and transmit it to the cloud server; S202: Clean the real-time multi-source heterogeneous monitoring data on the cloud server, and perform dynamic time window alignment based on mutual information analysis to obtain pre-processed real-time standard monitoring data. S203: Input the preprocessed real-time standard monitoring data into the karst water hazard prediction model constructed based on deep learning algorithm and multi-objective optimization algorithm to predict karst water hazards and obtain real-time karst water hazard prediction results; S204: Based on real-time karst water hazard prediction results, graded early warnings will be issued.

[0032] The technical solution provided in this application has at least the following beneficial effects: By deploying an IoT system for karst flood monitoring, comprehensive real-time acquisition and cloud aggregation of multi-source heterogeneous data, including hydrological, meteorological, and microseismic data, were achieved, resulting in rich data dimensions. Utilizing dynamic time window alignment technology based on mutual information analysis, the optimal lag time between meteorological and hydrological data was automatically calculated, resolving the temporal misalignment problem caused by the rainfall infiltration lag effect and improving data consistency. A data augmentation model based on physical constraints was introduced, not only addressing the scarcity of karst flood samples but also ensuring that the generated samples conformed to physical laws by constraining the generated data through Darcy's law, thus improving the training quality of the karst flood prediction model. A multi-objective fruit fly optimization algorithm was used to optimize the hyperparameters of the karst flood prediction model, obtaining a Pareto solution set that balances accuracy and speed. Different models (high-precision, balanced, and ultra-fast) were adaptively scheduled according to cloud load, effectively balancing the accuracy and real-time performance of early warnings. A tiered early warning release mechanism was established, automatically triggering different levels of warnings based on prediction results, achieving closed-loop management from data acquisition and intelligent prediction to emergency response.

[0033] In one optional implementation, the real-time multi-source heterogeneous monitoring data includes real-time hydrogeological monitoring data, real-time meteorological and environmental monitoring data, and real-time microseismic monitoring data; In this embodiment, real-time hydrogeological monitoring data includes groundwater level, water pressure (pore water pressure), water inflow, water temperature, and water turbidity in the advanced boreholes at the tunnel face, the karst fissures exposed in the roadway, and key aquifer locations. Real-time meteorological and environmental monitoring data include rainfall (rain intensity), evaporation, temperature, atmospheric pressure, and relative humidity in the surface recharge area and at the project entrance; Real-time microseismic monitoring data includes microseismic event frequency, microseismic energy, magnitude, waveform data, and source coordinates.

[0034] In one alternative implementation, a karst flood hazard monitoring IoT is deployed in the monitoring area. Based on the karst flood hazard monitoring IoT, real-time multi-source heterogeneous monitoring data of the monitoring area is collected and transmitted to a cloud server, including: S2011: High-precision piezometers and ultrasonic flow meters are installed in the advanced boreholes at the tunnel face, in the karst fissures exposed in the roadway, and at key aquifer locations in the monitoring area to collect hydrogeological monitoring data. S2012: Rain gauges and temperature and humidity sensors are installed in the surface recharge area and at the entrance of the project in the monitoring area to collect meteorological and environmental monitoring data. S2013: Deploy an array of microseismic sensors on the surface of the monitoring area to collect microseismic monitoring data; S2014: Connect high-precision piezometers, ultrasonic flow meters, rain gauges, temperature and humidity sensors, and micro-vibration pickup arrays to the IoT gateway within the communication range, and connect the IoT gateway to the cloud server to build an IoT for monitoring karst water hazards. S2015: Based on the Internet of Things for monitoring karst water hazards, collect real-time multi-source heterogeneous monitoring data of the monitoring area, and aggregate the real-time multi-source heterogeneous monitoring data to the Internet of Things gateway through LoRa or ZigBee networking. S2016: In the Internet of Things (IoT) gateway, real-time multi-source heterogeneous monitoring data is encrypted and transmitted to the cloud server; In this embodiment, the IoT gateway uses the Message Queuing Telemetry Transport (MQTT) protocol to push encrypted data packets (real-time multi-source heterogeneous monitoring data) to a cloud message queue (such as Apache Kafka) via a 4G / 5G cellular network, thereby achieving breakpoint resume and load balancing.

[0035] In one optional implementation, real-time multi-source heterogeneous monitoring data is cleaned on a cloud server, and dynamic time window alignment is performed based on mutual information analysis to obtain preprocessed real-time standard monitoring data, including: S2021: Using 3 on cloud servers σ The criteria perform preliminary screening of real-time multi-source heterogeneous monitoring data, removing obvious outliers caused by signal interference, and obtaining pre-screened real-time multi-source heterogeneous monitoring data. S2022: Use the five-number summary method to identify and process abnormal data in the real-time multi-source heterogeneous monitoring data after preliminary screening, and obtain cleaned real-time multi-source heterogeneous monitoring data; S2023: Based on mutual information analysis, the real-time meteorological and environmental monitoring data and the real-time hydrogeological monitoring data after cleaning are dynamically aligned by time window to obtain the aligned real-time meteorological and environmental monitoring data. S2024: Integrate real-time hydrogeological monitoring data after cleaning, real-time meteorological and environmental monitoring data after alignment, and real-time microseismic monitoring data after cleaning to obtain pre-processed real-time standard monitoring data.

[0036] In one optional implementation, based on mutual information analysis, dynamic time-window alignment is performed on the cleaned real-time meteorological and environmental monitoring data and the cleaned real-time hydrogeological monitoring data in the cleaned real-time multi-source heterogeneous monitoring data to obtain aligned real-time meteorological and environmental monitoring data, including: S20231: Based on the preset sliding window length, extract the pre-processed historical standard monitoring data sequence, the cleaned historical hydrogeological monitoring data sequence, and the corresponding aligned historical meteorological and environmental monitoring data sequence from the historical database obtained by manual processing on the cloud server. S20232: Normalize the cleaned historical hydrogeological monitoring data sequence and the aligned historical meteorological and environmental monitoring data sequence to obtain the normalized historical hydrogeological monitoring data sequence and the normalized historical meteorological and environmental monitoring data sequence. S20233: Traverse the lag time to calculate the mutual information between the normalized historical hydrogeological monitoring data sequence and the normalized historical meteorological and environmental monitoring data sequence. The formula is:

[0037] In the formula, For mutual information content; This is a normalized historical hydrogeological monitoring data sequence; This is a normalized sequence of historical meteorological and environmental monitoring data; The time lag is the time delay. These are the historical hydrogeological monitoring data points in the normalized historical hydrogeological monitoring data sequence; These are the historical meteorological and environmental monitoring data points in the normalized historical meteorological and environmental monitoring data sequence. For joint probability density; Probability density of historical hydrogeological monitoring data; The probability density of historical meteorological and environmental monitoring data; S0234: Obtain the optimal lag time that maximizes mutual information. ; S0235: Shift the cleaned real-time meteorological and environmental monitoring data in the cleaned real-time multi-source heterogeneous monitoring data backward by the optimal lag time. We obtained aligned real-time meteorological and environmental monitoring data.

[0038] In one optional implementation, the karst water hazard prediction model is constructed based on the Convolutional Neural Networks (CNN)-Long Short-Term Memory (LSTM)-Multi-Layer Perceptron (MLP) algorithm, and the karst water hazard prediction model includes a spatial feature extraction module based on the CNN algorithm, a temporal feature extraction module based on the LSTM algorithm, and a karst water hazard prediction module based on the MLP algorithm. In this embodiment, the spatial feature extraction module uses a one-dimensional convolutional layer (1D-CNN) to slide along the time dimension to extract the local correlation between different sensor data (for example, the enhancement of microseismic signals often precedes the sudden increase in water pressure). The convolutional kernel size is 3, the number is 64, and the activation function is ReLU. The temporal feature extraction module inputs the feature sequences extracted by CNN into a two-layer LSTM network. The LSTM captures long-term groundwater dynamic trends through gating mechanisms (forget gate, input gate, output gate), solves the gradient vanishing problem, and remembers historical hydrological evolution patterns. The karst water hazard prediction module flattens the output of the LSTM and maps it to two output nodes: node 1 outputs the probability of water inrush (Sigmoid activation) and node 2 outputs the expected water inrush volume (Linear activation).

[0039] In one alternative implementation, a karst flood hazard prediction model is constructed based on deep learning algorithms and multi-objective optimization algorithms, including: A101: Extract a small batch of pre-processed historical standard monitoring data with real labels from the historical database obtained by manual processing on the cloud server. A102: Using a pre-built data augmentation model based on physical constraints, generate several generated standard monitoring data sets with real labels, and combine them with several pre-processed historical standard monitoring data sets to obtain a large number of training sample sets. In this embodiment, a data augmentation model based on physical constraints is used to enhance small sample data of karst water inrush. The discriminator loss function of the data augmentation model includes a physical constraint term based on Darcy's law to ensure that the generated enhanced sample conforms to hydrogeological and physical laws. A103: Using the CNN-LSTM-MLP algorithm, an initial karst water hazard prediction model was constructed; A104: Based on the training sample set, the multi-objective fruit fly optimization algorithm is used to optimize the hyperparameters of the initial karst water hazard prediction model, and obtain the Pareto solution set of the optimal hyperparameters for different application scenarios and the corresponding final karst water hazard prediction model; the application scenarios include high-precision, balanced and high-speed types. A105: Package the model parameters of the final karst water hazard prediction model for each application scenario into a Docker container and deploy it in the inference engine of the cloud server.

[0040] In one alternative implementation, the data augmentation model is built based on the Physically Constrained Generative Adversarial Networks (PC-GAN) algorithm, and the data augmentation model includes: generator G The system employs a fully connected layer combined with a transposed convolutional layer structure. The input consists of a 100-dimensional Gaussian noise vector and a conditional vector (preprocessed historical standard monitoring data). The output is a simulated time-series waveform of "water level-water pressure-microseismic energy", which generates standard monitoring data. Discriminator D The discriminator employs a convolutional neural network (CNN) structure to distinguish between real and generated samples; its key feature is that a physical constraint term is introduced into the loss function of the discriminator. The formula for the total loss function is:

[0041] In the formula, This represents the total loss value. For generator G and discriminator D The loss value; This represents the physical loss value. The regularization coefficient is used.

[0042] In the formula, Virtual water inflow data for hydrogeological monitoring data generated by the generator; The water level change data generated by the generator from hydrogeological monitoring data; This represents the straight-line distance of the groundwater flow path; This is the hydraulic gradient, which is the head loss per unit length in the direction of water flow; It is an indicator of the permeability of rock or soil, representing the rate at which fluid seeps under a unit hydraulic gradient; This represents the cross-sectional area through which the groundwater flows. A small batch of preprocessed historical standard monitoring data with real labels is input into the initial data augmentation model for training until the total loss value is reached. If the total loss value is less than the threshold, the final data augmentation model is obtained, which is used to generate several standard monitoring data.

[0043] In one alternative implementation, based on the training sample set, a multi-objective fruit fly optimization algorithm is used to optimize the hyperparameters of the initial karst flood prediction model, obtaining a Pareto solution set of optimal hyperparameters for different application scenarios and the corresponding final karst flood prediction model, including: A1041: Define the multi-objective optimization function of the multi-objective fruit fly optimization algorithm, and encode the hyperparameters of the initial karst water hazard prediction model into individual vectors of the multi-objective fruit fly optimization algorithm. ,in, For the first i Hyperparameters of an individual fruit fly i For individual indicators; →Number of CNN convolutional kernels (range 32-128); →Number of hidden cells in LSTM (range 50-300); →Dropout drop rate (range 0.1-0.5); →Initial learning rate (range 0.0001-0.01). The multi-objective optimization function includes minimizing the reciprocal of the accuracy function and minimizing the reciprocal of the velocity function, as shown in the formula:

[0044] In the formula, For individual fruit flies The first optimization objective value; The root mean square error on the validation set; To find the minimum value, to prevent the denominator from being zero;

[0045] In the formula, For individual fruit flies The second optimization objective value; The average time (in milliseconds) for the model to perform one inference on the cloud server; A1042: Use Logistic mapping to generate chaotic sequences and map the chaotic sequences to the solution space of individual fruit flies to obtain the initial fruit fly population; The formula is:

[0046] In the formula, For the first n+ 1. n One chaotic variable; The stability coefficient is typically 4. This sequence is ergodic and random, ensuring that the initial population is uniformly distributed in the solution space, avoiding getting trapped in local optima, which is superior to traditional random initialization. n For chaotic variable indicators;

[0047] In the formula, The first in the initial fruit fly population i The initial fruit fly individual exist a shaft and b The position of the axis; For the first i One chaotic variable; For search space a The upper and lower bounds of the axis; For individual indicators; For search space b The upper and lower bounds of the axis; A1043: Calculate the initial distance of each initial fruit fly individual in the initial fruit fly population, and obtain the initial taste concentration judgment value corresponding to the initial distance; The formula is:

[0048] In the formula, For the first i Individual fruit flies The distance; For the first i Individual fruit flies exist a shaft and b The position of the axis;

[0049] In the formula, For the first i Taste concentration determination value for an individual fruit fly; A1044: The initial taste concentration judgment values ​​are transformed into candidate hyperparameters through a linear mapping function. Training is then performed using a training sample set to construct candidate karst water hazard prediction models. Finally, a multi-objective optimization function is used to calculate the optimization objective value of the candidate karst water hazard prediction models. ; A1045: Based on the optimization target value, the initial fruit fly population is subjected to rapid non-dominated sorting to obtain different frontier levels; A1046: Iteratively update the initial fruit fly population to obtain an updated fruit fly population;

[0050] In the formula, For the first t+ The first iteration i Individual fruit flies exist a shaft and b The position of the axis; for Levy Distribute random numbers; for Levy Step length, and ∈[1,2]; For the first t The optimal fruit fly individual in the next iteration exist a shaft and b The position of the axis; A random number between (0, 1); For the first t The convergence factor of the next iteration; t This is an indicator of the number of iterations.

[0051] In the formula, These are the maximum and minimum values ​​of the convergence factor; This represents the maximum number of iterations. , To adjust the parameters; It is the hyperbolic tangent function; A1047: Using a multi-objective optimization function, calculate the optimization objective value for each updated fruit fly individual in the updated fruit fly population, and update the frontier level based on the optimization objective value; A1048: Until the number of iterations reaches the iteration threshold, the first frontier level of the last iteration, which includes several optimal hyperparameters, is taken as the Pareto solution set, and the final karst water hazard prediction model corresponding to each solution in the Pareto solution set is saved.

[0052] In one optional implementation, preprocessed real-time standard monitoring data is input into a karst water hazard prediction model constructed based on deep learning algorithms and multi-objective optimization algorithms to predict karst water hazards, obtaining real-time karst water hazard prediction results, including: S2031: Collect real-time operation monitoring data from the cloud server; the real-time operation monitoring data includes the real-time CPU utilization of the cloud server. and the length of the real-time concurrent request queue ; S2032: A karst water hazard prediction model based on deep learning algorithms and multi-objective optimization algorithms, which adaptively schedules the corresponding application scenarios based on real-time operation monitoring data. In this embodiment, Strategy A (High Load): When real-time CPU utilization... CPU utilization threshold and / or real-time concurrent request queue length Length threshold It automatically switches to the "ultra-fast" karst water hazard prediction model, sacrificing a small amount of prediction accuracy to prioritize a response time of <50ms and ensure timely delivery of early warning information. Strategy B (Low Load): When real-time CPU utilization is low CPU utilization threshold and the length of the real-time concurrent request queue Length threshold A "high-precision" karst water hazard prediction model was adopted, making full use of idle computing power to conduct high-precision trend extrapolation and provide data support for long-term prevention and control. Strategy C (Start or Default): By starting or defaulting, maintain the "balanced" karst water hazard prediction model; S2033: Input the preprocessed real-time standard monitoring data into the scheduled karst water hazard prediction model to perform karst water hazard prediction and obtain real-time karst water hazard prediction results; the real-time karst water hazard prediction results include real-time water inrush probability and real-time estimated water inflow.

[0053] In one optional implementation, tiered early warnings are issued based on real-time karst flood hazard prediction results, including: S2041: If the real-time probability of water inrush in the real-time karst water hazard prediction results are... satisfy This generates blue / yellow alert levels; S2042: If the real-time probability of water inrush is... satisfy This generates an orange alert level warning signal; S2043: If the real-time probability of water inrush is... satisfy This generates a red emergency level warning signal; S2044: The cloud server calls the SMS service, APP push interface and on-site sound and light alarm through the API gateway to simultaneously release early warning signals and the corresponding real-time predicted water inflow of karst water hazard prediction results.

[0054] This invention also provides a cloud-based karst flood hazard prediction device, referring to... Figure 3The device may include the following units: The data acquisition unit 301 is used to deploy a karst water hazard monitoring Internet of Things (IoT) in the monitoring area, collect real-time multi-source heterogeneous monitoring data of the monitoring area based on the karst water hazard monitoring IoT, and transmit it to the cloud server. The data preprocessing unit 302 is used to clean real-time multi-source heterogeneous monitoring data on a cloud server and perform dynamic time window alignment based on mutual information analysis to obtain preprocessed real-time standard monitoring data. The karst water hazard prediction unit 303 is used to input the preprocessed real-time standard monitoring data into the karst water hazard prediction model constructed based on deep learning algorithm and multi-objective optimization algorithm to predict karst water hazards and obtain real-time karst water hazard prediction results. The graded early warning release unit 304 is used to release graded early warnings based on real-time karst water hazard prediction results.

[0055] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. Memory, used to store computer programs; The processor, when executing the program stored in the memory, implements the cloud-based karst water hazard prediction method of the present invention.

[0056] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EI) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned terminal and other devices. The memory can include Random Access Memory (RAM), or non-volatile memory, such as at least one disk storage device. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor.

[0057] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0058] Furthermore, to achieve the above objectives, embodiments of the present invention also propose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the cloud-based karst water hazard prediction method of the present invention.

[0059] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable hardware devices (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0061] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0062] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0063] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. "And / or" indicates that either one or both can be chosen. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0064] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting karst water hazards based on cloud processing, characterized in that, The method includes: Deploy a karst water hazard monitoring Internet of Things (IoT) in the monitoring area, collect real-time multi-source heterogeneous monitoring data of the monitoring area based on the karst water hazard monitoring IoT, and transmit it to the cloud server; The real-time multi-source heterogeneous monitoring data is cleaned on the cloud server, and dynamic time window alignment is performed based on mutual information analysis to obtain preprocessed real-time standard monitoring data. The preprocessed real-time standard monitoring data is input into the karst water hazard prediction model constructed based on deep learning algorithm and multi-objective optimization algorithm to predict karst water hazards and obtain real-time karst water hazard prediction results. Based on real-time karst water hazard prediction results, graded early warnings are issued.

2. The karst water hazard prediction method based on cloud processing according to claim 1, characterized in that, The real-time multi-source heterogeneous monitoring data includes real-time hydrogeological monitoring data, real-time meteorological and environmental monitoring data, and real-time microseismic monitoring data.

3. The karst water hazard prediction method based on cloud processing according to claim 2, characterized in that, Deploy an IoT (Internet of Things) for karst flood hazard monitoring in the monitoring area. Based on this IoT, collect real-time multi-source heterogeneous monitoring data of the monitoring area and transmit it to a cloud server, including: High-precision piezometers and ultrasonic flow meters were installed in the advanced boreholes at the tunnel face, in the karst fissures exposed in the roadway, and at the location of key aquifers in the monitoring area to collect hydrogeological monitoring data. Rain gauges and temperature and humidity sensors were installed in the surface recharge area and at the entrance of the project in the monitoring area to collect meteorological and environmental monitoring data. An array of microseismic sensors was deployed on the surface of the monitoring area to collect microseismic monitoring data. Connect high-precision piezometers, ultrasonic flow meters, rain gauges, temperature and humidity sensors, and micro-vibration pickup arrays to the IoT gateway within the communication range, and connect the IoT gateway to the cloud server to build an IoT for monitoring karst water hazards. Based on the Internet of Things (IoT) for monitoring karst water hazards, real-time multi-source heterogeneous monitoring data of the monitoring area is collected and aggregated to the IoT gateway. The IoT gateway encrypts and transmits real-time multi-source heterogeneous monitoring data to the cloud server.

4. The karst water hazard prediction method based on cloud processing according to claim 3, characterized in that, Real-time multi-source heterogeneous monitoring data is cleaned on a cloud server, and dynamic time window alignment is performed based on mutual information analysis to obtain preprocessed real-time standard monitoring data, including: Using 3 on a cloud server σ The criteria perform preliminary screening of real-time multi-source heterogeneous monitoring data, removing obvious outliers caused by signal interference, and obtaining pre-screened real-time multi-source heterogeneous monitoring data. The five-number summary method was used to identify and process abnormal data in the real-time multi-source heterogeneous monitoring data after preliminary screening, and cleaned real-time multi-source heterogeneous monitoring data was obtained. Based on mutual information analysis, the real-time meteorological and environmental monitoring data and the real-time hydrogeological monitoring data after cleaning are dynamically aligned in time windows to obtain the aligned real-time meteorological and environmental monitoring data. By integrating real-time hydrogeological monitoring data after cleaning, real-time meteorological and environmental monitoring data after alignment, and real-time microseismic monitoring data after cleaning, pre-processed real-time standard monitoring data is obtained.

5. The karst water hazard prediction method based on cloud processing according to claim 4, characterized in that, Based on mutual information analysis, dynamic time window alignment is performed on the cleaned real-time meteorological and environmental monitoring data and the cleaned real-time hydrogeological monitoring data in the cleaned real-time multi-source heterogeneous monitoring data to obtain aligned real-time meteorological and environmental monitoring data, including: Based on the preset sliding window length, the pre-processed historical standard monitoring data sequence, the cleaned historical hydrogeological monitoring data sequence, and the corresponding aligned historical meteorological and environmental monitoring data sequence are extracted from the historical database obtained by manual processing on the cloud server. The cleaned historical hydrogeological monitoring data sequence and the aligned historical meteorological and environmental monitoring data sequence were normalized to obtain the normalized historical hydrogeological monitoring data sequence and the normalized historical meteorological and environmental monitoring data sequence. By iterating through the lag times, the mutual information of the normalized historical hydrogeological monitoring data series and the normalized historical meteorological and environmental monitoring data series is calculated using the following formula: In the formula, For mutual information content; This is a normalized historical hydrogeological monitoring data sequence; This is a normalized sequence of historical meteorological and environmental monitoring data; The time lag is the time delay. These are the historical hydrogeological monitoring data points in the normalized historical hydrogeological monitoring data sequence; These are the historical meteorological and environmental monitoring data points in the normalized historical meteorological and environmental monitoring data sequence. For joint probability density; Probability density of historical hydrogeological monitoring data; The probability density of historical meteorological and environmental monitoring data; Find the optimal lag time that maximizes mutual information. ; The cleaned real-time meteorological and environmental monitoring data in the cleaned real-time multi-source heterogeneous monitoring data are shifted backward by the optimal lag time. We obtained aligned real-time meteorological and environmental monitoring data.

6. The karst water hazard prediction method based on cloud processing according to claim 5, characterized in that, The karst water hazard prediction model is constructed based on the CNN-LSTM-MLP algorithm, and includes a spatial feature extraction module based on the CNN algorithm, a temporal feature extraction module based on the LSTM algorithm, and a karst water hazard prediction module based on the MLP algorithm.

7. The karst water hazard prediction method based on cloud processing according to claim 6, characterized in that, A karst flood hazard prediction model is constructed based on deep learning algorithms and multi-objective optimization algorithms, including: Extract a small batch of pre-processed historical standard monitoring data with real labels from the historical database obtained by manual processing on the cloud server. Using a pre-built data augmentation model based on physical constraints, several generated standard monitoring data sets with real labels are generated, and combined with several pre-processed historical standard monitoring data sets, a large number of training sample sets are obtained. An initial karst water hazard prediction model was constructed using the CNN-LSTM-MLP algorithm. Based on the training sample set, the hyperparameters of the initial karst water hazard prediction model are optimized using the multi-objective fruit fly optimization algorithm, resulting in the Pareto solution set of the optimal hyperparameters for different application scenarios and the corresponding final karst water hazard prediction model. The application scenarios include high-precision, balanced, and high-speed models. The model parameters of the final karst water hazard prediction model for each application scenario are packaged into Docker containers and deployed in the inference engine of the cloud server.

8. The karst water hazard prediction method based on cloud processing according to claim 7, characterized in that, Based on the training sample set, the hyperparameters of the initial karst flood hazard prediction model are optimized using the multi-objective fruit fly optimization algorithm. This yields the Pareto solution set of optimal hyperparameters for different application scenarios and the corresponding final karst flood hazard prediction model, including: Define the multi-objective optimization function of the multi-objective fruit fly optimization algorithm, and encode the hyperparameters of the initial karst water hazard prediction model into individual vectors of the multi-objective fruit fly optimization algorithm; The chaotic sequence is generated using the Logistic mapping, and then mapped to the solution space of individual fruit flies to obtain the initial fruit fly population. Calculate the initial distance of each initial fruit fly individual in the initial fruit fly population, and obtain the initial taste concentration judgment value corresponding to the initial distance; The initial taste concentration judgment value is converted into a candidate hyperparameter through a linear mapping function, and the candidate karst water hazard prediction model is constructed by training with a training sample set. The optimization objective value of the candidate karst water hazard prediction model is calculated using a multi-objective optimization function. Based on the optimization target value, the initial fruit fly population is subjected to rapid non-dominated sorting to obtain different frontier levels; The initial fruit fly population is iteratively updated to obtain an updated fruit fly population; Using a multi-objective optimization function, the optimization objective value for each updated fruit fly individual in the updated fruit fly population is calculated, and the frontier level is updated based on the optimization objective value. Until the number of iterations reaches the iteration threshold, the first frontier level of the last iteration, which includes several optimal hyperparameters, is taken as the Pareto solution set, and the final karst water hazard prediction model corresponding to each solution in the Pareto solution set is saved.

9. The karst water hazard prediction method based on cloud processing according to claim 8, characterized in that, The preprocessed real-time standard monitoring data is input into a karst water hazard prediction model constructed based on deep learning and multi-objective optimization algorithms to predict karst water hazards, obtaining real-time karst water hazard prediction results, including: Collect real-time operational monitoring data of the cloud server; the real-time operational monitoring data includes the real-time CPU utilization and the real-time concurrent request queue length of the cloud server; Based on real-time operation monitoring data, the karst water hazard prediction model, constructed using deep learning algorithms and multi-objective optimization algorithms, is adaptively scheduled for the corresponding application scenarios. The preprocessed real-time standard monitoring data is input into the scheduled karst water hazard prediction model to perform karst water hazard prediction and obtain real-time karst water hazard prediction results; the real-time karst water hazard prediction results include real-time water inrush probability and real-time estimated water inflow.

10. A cloud-based karst flood hazard prediction device, used to implement the karst flood hazard prediction method as described in any one of claims 1-9, characterized in that, The device includes: The data acquisition unit is used to deploy a karst water hazard monitoring Internet of Things (IoT) in the monitoring area. Based on the karst water hazard monitoring IoT, it collects real-time multi-source heterogeneous monitoring data of the monitoring area and transmits it to the cloud server. The data preprocessing unit is used to clean real-time multi-source heterogeneous monitoring data on the cloud server and perform dynamic time window alignment based on mutual information analysis to obtain preprocessed real-time standard monitoring data. The karst water hazard prediction unit is used to input preprocessed real-time standard monitoring data into the karst water hazard prediction model constructed based on deep learning algorithm and multi-objective optimization algorithm to predict karst water hazards and obtain real-time karst water hazard prediction results. The tiered early warning release unit is used to release tiered early warnings based on real-time karst water hazard prediction results.

Images