Permafrost area monitoring method and device based on digital twinning, computer equipment and medium

Abstract

The application relates to a permafrost region monitoring method and device based on digital twinning, computer equipment and a medium, which comprises the following steps: obtaining a hydrogeological model of a permafrost region and multi-source state data collected from a full section of a road in the permafrost region; obtaining a physical model of the permafrost region by combining the pretreated multi-source state data and the hydrogeological model, and coupling the physical model, a pre-trained data model and a neural network model to obtain an initial digital twinning model; driving the initial digital twinning model to perform virtual mapping to obtain mapping state parameters; performing reliability evaluation on the initial digital twinning model based on the mapping state parameters and first actual state parameters of the permafrost region to obtain an evaluation result; when the evaluation result indicates that the initial digital twinning model is reliable, obtaining a digital twinning model to monitor a plurality of hydrological parameters of the permafrost region by using the digital twinning model. The method can comprehensively and accurately monitor the groundwater in the permafrost region.

Description

Technical Field

[0001] This application relates to the field of permafrost highway engineering technology, and in particular to a monitoring method, device, computer equipment and medium for permafrost regions based on digital twins. Background Technology

[0002] Permafrost, also known as frozen soil, refers to a layer of soil and rock that remains frozen for many years. Permafrost typically consists of two parts: the upper active layer, which thaws in summer and freezes again in winter; and the lower permafrost layer, which remains frozen year-round. Under the influence of freeze-thaw cycles, rainfall infiltration, snowmelt recharge, and surface runoff, permafrost regions possess abundant groundwater resources, mainly including three types: surface water, inter-layer water, and groundwater. Intelligent monitoring of groundwater distribution in permafrost regions not only helps to deepen the understanding of its hydrothermal coupling mechanism but also provides crucial data support and theoretical basis for revealing the gestation mechanisms of geological hazards and engineering problems induced by permafrost degradation.

[0003] Existing groundwater monitoring technologies in permafrost regions include electrical resistivity tomography, ground-penetrating radar, seismic exploration, acoustic wave detection, and magnetic detection. These technologies can only perform passive tasks and are not capable of predicting and assessing the evolution trend of groundwater distribution.

[0004] Invention patent application CN116702047A discloses a method, system and medium for real-time groundwater monitoring. It acquires historical environmental data and historical monitoring data in the groundwater monitoring area and constructs a pollution monitoring model based on digital twins to achieve groundwater pollution monitoring and prediction. However, this method focuses on predicting the state of groundwater pollution. Since the distribution of groundwater is related to many factors, monitoring only the burial depth of groundwater will inevitably cause deviations in the prediction of groundwater development trends, resulting in inaccurate prediction results.

[0005] Invention patent application CN115587542A discloses a groundwater inversion simulation method, system, equipment, and medium based on reinforcement learning. Based on basic groundwater data, it constructs a hydrogeological model based on dynamic digital twins. Through groundwater monitoring stations, it collects water quality data from various monitoring wells in real time, thereby obtaining accurate hydrogeological modeling and pollution source tracing. However, this model focuses on predicting groundwater quality indicators and cannot achieve comprehensive prediction.

[0006] Chinese patent application CN116187758A discloses a method, device, electronic equipment, and storage medium for monitoring hydrological disasters. After collecting precipitation data for each rainfall event, it calculates precipitation runoff based on real-time and forecast precipitation data. This runoff is then used to determine the runoff table for the watershed covered by the real-time precipitation. The runoff evolution data is compared with runoff thresholds to generate a disaster risk propensity report, which is then used to predict flood disasters. However, this method involves geological data and relies solely on precipitation as the data source for predicting water flow within the watershed, resulting in significant uncertainty and making it difficult to accurately predict disaster occurrences.

[0007] Chinese patent application CN116625324A discloses a water conservancy monitoring method based on digital twins. This method collects hydrological and water conservancy data through hardware devices, constructs a digital twin, and uses it to predict operational parameters for future periods. The water conservancy monitoring system cloud platform generates data reports or graphs based on the data collected by the hardware devices in the target area, monitors the target area in real time, and performs early warning analysis on the monitored area based on the monitoring results. However, the monitoring hardware devices primarily monitor the attributes of the monitored object itself, without addressing much of the external influence on the monitored object, thus potentially leading to less than ideal prediction results. Summary of the Invention

[0008] Therefore, it is necessary to provide a method, device, computer equipment, and medium for monitoring permafrost regions based on digital twins, addressing the aforementioned technical problems. This method can comprehensively and accurately monitor groundwater in permafrost regions.

[0009] A monitoring method for permafrost regions based on digital twins, the method comprising:

[0010] S1. Obtain the hydrogeological model of the permafrost region and multi-source state data collected from the full cross-section of the road in the permafrost region, wherein the permafrost region is a soil and rock layer with a freezing time exceeding a preset time.

[0011] S2. Combining the preprocessed multi-source state data and the hydrogeological model, a physical model of the permafrost region is obtained, and the physical model, the pre-trained data model, and the neural network model are coupled to obtain an initial digital twin model. The neural network is trained based on the historical state parameters and current state parameters of the permafrost region.

[0012] S3. Drive the initial digital twin model to perform virtual mapping to obtain mapping state parameters;

[0013] S4. Based on the mapped state parameters and the first actual state parameters of the permafrost region, the reliability of the initial digital twin model is evaluated to obtain the evaluation results;

[0014] S5. When the evaluation results indicate that the initial digital twin model is reliable, a digital twin model is obtained to monitor multiple hydrological parameters of the permafrost region.

[0015] In this application, a physical model of the permafrost region is obtained by acquiring a hydrogeological model of the permafrost region and multi-source state data collected from the full cross-section of roads in the permafrost region. This physical model is then combined with the pre-processed multi-source state data and the hydrogeological model. The physical model, the pre-trained data model, and the neural network model are coupled to obtain an initial digital twin model. The neural network is trained based on the historical and current state parameters of the permafrost region and drives the initial digital twin model to perform virtual mapping to obtain mapped state parameters. Based on the mapped state parameters and the first actual state parameters of the permafrost region, the reliability of the initial digital twin model is evaluated to obtain the evaluation results. When the evaluation results indicate that the initial digital twin model is reliable, the digital twin model is obtained. In this way, the digital twin model can be used to monitor groundwater in the permafrost region in a comprehensive and accurate manner.

[0016] In one embodiment, the process of constructing the hydrogeological model of the permafrost region includes:

[0017] Obtain hydrogeological parameters, roadbed parameters, and pavement parameters in permafrost regions;

[0018] The hydrogeological parameters, the roadbed parameters, and the pavement parameters are preprocessed to obtain the preprocessed hydrogeological parameters, roadbed parameters, and pavement parameters.

[0019] Based on the geological complexity of the permafrost region, the hydrogeological parameters, the roadbed parameters, and the pavement parameters, a target model architecture is established.

[0020] Based on the hydrogeological parameters, the roadbed parameters, the pavement parameters, and the target model architecture, a hydrogeological model of the permafrost region is constructed.

[0021] In this application, hydrogeological parameters, subgrade parameters, and pavement parameters of permafrost regions are obtained. These parameters are then preprocessed to obtain the preprocessed hydrogeological parameters, subgrade parameters, and pavement parameters. Based on the geological complexity, hydrogeological parameters, subgrade parameters, and pavement parameters of the permafrost regions, a target model architecture is established. Based on the hydrogeological parameters, subgrade parameters, pavement parameters, and target model architecture, a hydrogeological model of the permafrost regions is constructed. This allows the constructed hydrogeological model to more realistically reflect the complex water-thermal-mechanical coupling process in the permafrost regions, improving model accuracy and regional adaptability.

[0022] In one embodiment, the method further includes:

[0023] Simulation analysis is performed using the hydrogeological model to output simulated state parameters and obtain second actual state parameters.

[0024] Select target parameters that meet preset conditions from the second actual state parameters; the preset conditions include that the time to which the second actual state parameter belongs is in the critical period of the freeze-thaw cycle, the space to which the second actual state parameter belongs covers the space where multiple types of permafrost are located, the second actual state parameter is a non-outlier value, and the second actual state parameter has not been used to construct the hydrogeological model.

[0025] When the difference between the simulated state parameters and the target parameters is greater than a preset difference, the key model parameters of the hydrogeological model are optimized to obtain the optimized hydrogeological model, and the physical model of the permafrost region is obtained based on the optimized hydrogeological model.

[0026] In this application, the key model parameters of the hydrogeological model are optimized when the difference between the simulated state parameters and the target parameters is greater than the preset difference. This can, on the one hand, focus on optimizing the core parameters and reduce the calculation cost of irrelevant parameters; on the other hand, improve the efficiency of model calibration and quickly reduce the error between the simulation results and the measured data; and on the other hand, clarify the key points of engineering regulation.

[0027] In one embodiment, step S4 includes:

[0028] Simulation is performed using the initial digital twin model to output the groundwater movement parameters and roadbed mechanical parameters of the permafrost region, and to obtain the actual groundwater movement parameters and actual roadbed mechanical parameters of the permafrost region.

[0029] Based on the groundwater movement parameters, the roadbed mechanical parameters, the actual groundwater movement parameters, and the actual roadbed mechanical parameters, calculate the first performance evaluation index of the initial digital twin model;

[0030] If the first performance evaluation index fails to reach the first preset value, the initial digital twin model is determined to be unreliable.

[0031] When the first performance evaluation index reaches the first preset value, the initial digital twin model is determined to be reliable.

[0032] In this application, an initial digital twin model is used for simulation to output groundwater movement parameters and roadbed mechanical parameters in permafrost areas. The actual groundwater movement parameters and actual roadbed mechanical parameters in permafrost areas are also obtained. Based on the groundwater movement parameters, roadbed mechanical parameters, actual groundwater movement parameters, and actual roadbed mechanical parameters, a first performance evaluation index of the initial digital twin model is calculated. If the first performance evaluation index does not reach a first preset value, the initial digital twin model is determined to be unreliable. If the first performance evaluation index reaches the first preset value, the initial digital twin model is determined to be reliable. In this way, the reliability of the model can be objectively judged through the quantifiable first performance evaluation index, thereby achieving a quantitative evaluation of the model's credibility.

[0033] In one embodiment, the training process of the neural network model includes:

[0034] Obtain the historical state parameters of the permafrost region and the mapped state parameters output by the initial digital twin model;

[0035] Based on the historical state parameters and the mapped state parameters, a neural network model is trained to obtain the trained neural network model.

[0036] The mapping error is calculated based on the current state parameters of the permafrost region and the virtual state parameters output by the neural network model.

[0037] If the mapping error is greater than the second preset value, the weight parameters of the neural network model are optimized based on the mapping error, so as to drive the initial digital twin model to generate virtual state parameters corresponding to the current state parameters through the optimized neural network model, until the mapping error is less than or equal to the second preset value, and a pre-trained neural network model is obtained.

[0038] In this application, historical state parameters of permafrost regions and mapped state parameters output by an initial digital twin model are obtained. Based on these historical and mapped state parameters, a neural network model is trained to obtain a trained neural network model. Based on the current state parameters of the permafrost regions and the virtual state parameters output by the neural network model, the mapping error is calculated. If the mapping error is greater than a second preset value, the weight parameters of the neural network model are optimized based on the mapping error. The optimized neural network model drives the initial digital twin model to generate virtual state parameters corresponding to the current state parameters until the mapping error is less than or equal to the second preset value, thus obtaining a pre-trained neural network model. This enables high-precision, adaptive, and real-time virtual mapping of the infrastructure status in permafrost regions, thereby constructing a "highly consistent, continuously evolving, and reliable" digital twin model.

[0039] In one embodiment, the method further includes:

[0040] The digital twin model is used to predict the hydrological parameters of the permafrost region.

[0041] Based on the hydrological parameters, the risk level of the permafrost region is determined using a risk assessment algorithm.

[0042] In this application, hydrological parameters in permafrost regions are predicted using digital twin models. Based on these parameters, a risk assessment algorithm is used to determine the risk level of the permafrost region, thus enabling a shift from "passive response" to "proactive early warning." This also enhances the scientific rigor and dynamism of risk assessment.

[0043] In one embodiment, step S4 includes:

[0044] Obtain the first actual state parameters of the permafrost region;

[0045] The mapped state parameters are standardized and spatiotemporally aligned with the first actual state parameters to obtain the processed state parameters.

[0046] The state parameters are segmented according to their respective time points to obtain multiple segments of state parameters;

[0047] The cumulative deviation is calculated based on the state parameters of each segment and the corresponding first actual state parameters. The reliability of the initial digital twin model is evaluated based on the cumulative deviation of each segment to obtain the evaluation result of the initial digital twin model.

[0048] In this application, the state parameters are segmented according to their respective time points to obtain multiple segments of state parameters. The cumulative deviation is calculated based on each segment of state parameters and the corresponding first actual state parameter. The reliability of the initial digital twin model is evaluated based on the cumulative deviation of each segment to obtain the evaluation result of the initial digital twin model. This can improve the spatiotemporal accuracy and sensitivity of the model evaluation and avoid the problem of "overall average error masking local failure".

[0049] A monitoring device for permafrost regions based on digital twins, the device comprising:

[0050] The data acquisition module is used to acquire the hydrogeological model of the permafrost region and multi-source state data collected from the full cross-section of the road in the permafrost region, wherein the permafrost region is a soil and rock layer with a freezing time exceeding a preset time.

[0051] The coupling module is used to combine the preprocessed multi-source state data and the hydrogeological model to obtain the physical model of the permafrost region, and to couple the physical model, the pre-trained data model and the neural network model to obtain an initial digital twin model. The neural network is trained based on the historical state parameters and current state parameters of the permafrost region.

[0052] The mapping module is used to drive the initial digital twin model to perform virtual mapping and obtain mapping state parameters;

[0053] The evaluation module is used to evaluate the reliability of the initial digital twin model based on the mapped state parameters and the first actual state parameters of the permafrost region, and obtain the evaluation results.

[0054] A monitoring module is used to obtain a digital twin model when the evaluation results indicate that the initial digital twin model is reliable, so as to use the digital twin model to monitor multiple hydrological parameters of the permafrost region.

[0055] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described above.

[0056] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0057] The beneficial effect of the above-mentioned digital twin-based monitoring methods, devices, computer equipment, and media for permafrost regions is that they enable comprehensive and accurate monitoring of groundwater in permafrost regions. Attached Figure Description

[0058] Figure 1 This is an application environment diagram of a digital twin-based monitoring method for permafrost regions in one embodiment;

[0059] Figure 2 This is a flowchart illustrating a digital twin-based monitoring method for permafrost regions in one embodiment.

[0060] Figure 3 This is a schematic diagram illustrating the construction of a physical model in one embodiment;

[0061] Figure 4 This is a schematic diagram illustrating the information flow loop between the digital twin model and the physical model in one embodiment;

[0062] Figure 5 This is a schematic diagram of a digital twin model in one embodiment;

[0063] Figure 6 This is a diagram of the monitoring platform architecture in one embodiment;

[0064] Figure 7 This is a schematic diagram of the monitoring platform operation in one embodiment;

[0065] Figure 8 This is a structural block diagram of a digital twin-based monitoring device for permafrost regions in one embodiment;

[0066] Figure 9 This is an internal structural diagram of a computer device in one embodiment.

[0067] Reference numerals: 1. Temperature, humidity, and pore water pressure sensor; 2. Asphalt strain sensor; 3. Soil pressure sensor; 4. Concrete strain sensor; 5. Meteorological sensor; 6. Layered settlement sensor; 7. Dynamic weighing sensor; 8. Groundwater level and flow velocity sensor; 9. DDAS dynamic acquisition module; 10. DTMCU static acquisition module; 11. 300M wireless transceiver module; 12. Backup power supply one; 13. Backup power supply two; 14. External power supply interface; 15. Protective enclosure. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0069] The digital twin-based monitoring method for permafrost regions provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 interacts with server 104 via a wired / wireless channel. A data storage system stores the data that server 104 needs to process. The server acquires a hydrogeological model of the permafrost region and multi-source state data collected from the full cross-section of roads in the permafrost region. The server combines the pre-processed multi-source state data and the hydrogeological model to obtain a physical model of the permafrost region, and couples the physical model, a pre-trained data model, and a neural network model to obtain an initial digital twin model. The neural network is trained based on historical and current state parameters of the permafrost region. The server drives the initial digital twin model to perform virtual mapping to obtain mapped state parameters. Based on the mapped state parameters and the first actual state parameters of the permafrost region, the server performs a reliability assessment of the initial digital twin model to obtain an assessment result. When the assessment result indicates that the initial digital twin model is reliable, the server obtains the digital twin model, which is then used to monitor multiple hydrological parameters in the permafrost region. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, etc. Server 104 can be a single server, a server cluster consisting of multiple servers, or a cloud computing center consisting of multiple servers.

[0070] In one embodiment, such as Figure 2 As shown, a monitoring method for permafrost regions based on digital twins is provided, and this method is applied to... Figure 1 Taking server 104 as an example, the following steps are included:

[0071] S1. Obtain the hydrogeological model of the permafrost region and multi-source state data collected from the full cross-section of the road in the permafrost region. The permafrost region is a soil and rock layer with a freezing time exceeding the preset time.

[0072] The hydrogeological model is constructed using a three-dimensional interpolation geostatistical method, based on hydrogeological parameters, roadbed parameters, pavement parameters, and borehole distribution in the permafrost region.

[0073] Multi-source condition data is collected through sensors deployed along the entire road cross-section in permafrost regions. Specifically, real-time information about the entire road cross-section is collected by deploying different types of sensors and uploaded to a database. These sensors include, but are not limited to, temperature, humidity, pore water pressure sensors, asphalt strain sensors, earth pressure sensors, concrete strain sensors, meteorological sensors, stratified settlement sensors, dynamic weighing sensors, groundwater level sensors, and flow velocity sensors. The multi-source condition data is collected in real-time, ensuring that the data loaded into the hydrogeological model is up-to-date.

[0074] S2. Combining preprocessed multi-source state data and hydrogeological models, a physical model of the permafrost region is obtained. The physical model, pre-trained data model, and neural network model are coupled to obtain an initial digital twin model. The neural network is trained based on the historical and current state parameters of the permafrost region.

[0075] The physical model of the permafrost region can be obtained by loading preprocessed multi-source state data into a hydrogeological model. Preprocessing of the multi-source state data can be performed using DDAS (Distributed Data Acquisition System) and DTMCU (Digital Twin Monitoring and Control Unit). The core function of DDAS is to acquire high-frequency dynamic data and perform preliminary data preprocessing, such as filtering and noise reduction. DDAS processes dynamic data related to vehicle loads, such as asphalt strain, earth pressure, and dynamic weighing data. The core function of DTMCU is to acquire low-frequency stable or slowly changing data and perform data storage and format conversion. DTMCU processes environmental and static state data, such as temperature, humidity, groundwater level, and stratified settlement data. A schematic diagram of the physical model construction is shown below. Figure 3 As shown. Figure 3 The sensors and data acquisition and transmission equipment include: temperature, humidity, and pore water pressure sensor 1; asphalt strain sensor 2; soil pressure sensor 3; concrete strain sensor 4; meteorological sensor 5; stratified settlement sensor 6; dynamic weighing sensor 7; groundwater level and flow velocity sensor 8; DDAS dynamic acquisition module 9; DTMCU static acquisition module 10; 300M wireless receiver and transmitter module 11; backup power supply one 12; backup power supply two 13; external power supply interface 14; and protective enclosure 15.

[0076] Data models and neural networks can be used to drive the initial digital twin model to perform simulation / virtual mapping, thereby outputting groundwater movement parameters and roadbed mechanical parameters in permafrost areas, i.e., mapping state parameters.

[0077] Historical state parameters are freeze-thaw parameters of permafrost regions collected historically. These parameters include, but are not limited to, historically collected data on groundwater level changes, water flow velocity, air temperature, and meteorological indices. Current state parameters refer to real-time state parameters of permafrost regions. Examples include real-time groundwater level, water flow velocity, air temperature, and meteorological indices.

[0078] Furthermore, the server controls the sensors to transmit the collected multi-source status data to the data monitoring center via a wireless transmission channel. The data monitoring center then preprocesses the received multi-source status data, loading the preprocessed data into the hydrogeological model to obtain an initial digital twin model. For example, temperature sensor data is used to determine the temperature distribution in permafrost regions, groundwater level and flow velocity sensor data are used to supplement groundwater movement parameters, and pore water pressure sensor data is used to correct the permeability coefficient of the soil and rock mass, thus perfecting the initial parameter system of the hydrogeological model to obtain the initial digital twin model.

[0079] The data preprocessing steps for multi-source state data include:

[0080] 1. Remove outliers and noise from multi-source state data, and interpolate and fill in missing data to ensure data continuity.

[0081] 2. Standardize the data units and formats (such as timestamps and coordinate systems) for the same type of status data to avoid format confusion; associate different sensor data according to the "time-space" dimension. For example, bind temperature sensor data (12℃), humidity sensor data (25%), and pore water pressure sensor data (80kPa) at the same time and location of the borehole.

[0082] 3. Compare the status data collected by different sensors to verify the rationality of the status data. Methods for verifying the rationality of status data include: ① Physical law verification, such as a positive correlation between groundwater level sensor data and pore water pressure data, meaning that as the water level rises, the pore pressure also increases. If the water level and pore pressure are negatively correlated, then the water level or pore pressure is considered abnormal; ② Verification using multiple sensors at the same location, such as for temperature and humidity at the same borehole depth, if the humidity is still 0% when the temperature is greater than 0℃, then the humidity data is considered invalid; ③ Spatiotemporal consistency verification, such as the difference between rainfall data collected by two adjacent meteorological sensors should be less than 10%. If the difference is too large, then the terrain (such as whether there is obstruction) should be considered to determine whether the rainfall data collected by the two adjacent meteorological sensors is invalid.

[0083] 4. Select data that matches the requirements of the hydrogeological model. Specifically, determine the list of input parameter types required by the hydrogeological model (such as temperature in permafrost areas, groundwater level, permeability coefficient, and roadbed strain); label the processed multi-source state data in the format of "parameter type-collection location-collection time" (e.g., "temperature-roadbed 1.5m depth-20240610"); based on the labeling of the multi-source state data and the list of input parameter types, select data that meets the target conditions from the multi-source state data. Target conditions include the location being within the entire cross-section of the road, the data type belonging to the list of input parameter types, and the collection time being during the critical period of the freeze-thaw cycle. For example, if the hydrogeological model requires "permafrost temperature data", then select state data labeled as "temperature-active layer / permafrost layer-202406-202408".

[0084] Furthermore, the preprocessed multi-source state data is stored in a data repository, and the multi-source state data is visualized in the form of charts to facilitate subsequent analysis and model invocation.

[0085] S3. Drive the initial digital twin model to perform virtual mapping and obtain the mapping state parameters;

[0086] The mapped state parameters are obtained through the interactive mapping relationships established in the initial digital twin model. These interactive mapping relationships are obtained through a neural network model. Specifically, multi-source state data are input into the neural network model, enabling the model to learn the mapping relationship between "observed changes and how to adjust the internal state of the initial digital twin model," thereby driving the initial digital twin model to perform virtual mapping and obtain the mapped state parameters. For example, the mapped state parameters include the daily mapped values ​​of groundwater level in permafrost regions for the next month (such as daily average water level and rate of water level change), permeability coefficient, meteorological recharge, and freeze-thaw cycle intensity.

[0087] Furthermore, if the geological environment of the permafrost region changes, the interactive mapping relationship of the initial digital twin model will be reconstructed.

[0088] S4. Based on the mapped state parameters and the first actual state parameters of the permafrost region, the reliability of the initial digital twin model is evaluated, and the evaluation results are obtained.

[0089] Among them, there are consistent parameters among the first actual state parameter, the current state parameter, and the second actual state parameter.

[0090] Furthermore, the model performance evaluation index is calculated based on the mapped state parameters and the first actual state parameters of the permafrost region. Thus, the evaluation result of the initial digital twin model is obtained according to the model performance evaluation index and the preset index value.

[0091] S5. When the evaluation results indicate that the initial digital twin model is reliable, a digital twin model is obtained to monitor multiple hydrological parameters in permafrost regions.

[0092] Digital twin models have the following characteristics:

[0093] (1) Visibility. The digital twin model integrates the physical measurement results and simulation calculation results of road groundwater in permafrost areas, which can effectively solve the problem of observability of road groundwater in permafrost areas. It realizes intelligent monitoring of road groundwater in permafrost areas through digital twin in a virtual scene.

[0094] (2) Predictability. The digital twin model has the functions of perception analysis, simulation, iterative optimization and decision control. It realizes intelligent perception, real-time monitoring, accurate positioning and health prediction of the physical entity of road groundwater in permafrost areas. Driven by twin data and model, the monitoring and prediction of the state of road groundwater in permafrost areas can be realized through the iterative operation of the physical model of road groundwater in permafrost areas, the digital twin model and twin data.

[0095] (3) Interpretability. The digital twin model uses deep learning algorithms to describe the evolution mechanism of the digital twin model, and integrates current sensor data, state data and historical data to train and optimize the accuracy and robustness of the digital twin model, so that it has the ability to self-perceive, self-learn, self-adapt and self-optimize, and realize the full data representation of the groundwater status of roads in permafrost areas.

[0096] (4) Interactivity. Through Web service interfaces and data-driven methods, bidirectional mapping, real-time interaction and intelligent collaboration between physical entities and digital twin models are realized.

[0097] Multiple hydrological parameters refer to the property parameters of groundwater in permafrost regions, including but not limited to groundwater level, water flow velocity, and temperature. The information flow and circulation between the digital twin model and the physical model is as follows: Figure 4 As shown in the diagram. The schematic diagram of the resulting digital twin model is as follows. Figure 5 As shown.

[0098] The aforementioned monitoring method for permafrost regions based on digital twins can comprehensively and accurately monitor groundwater in permafrost regions.

[0099] In one embodiment, the process of constructing a hydrogeological model for a permafrost region includes:

[0100] Obtain hydrogeological parameters, roadbed parameters, and pavement parameters in permafrost regions;

[0101] Data preprocessing was performed on hydrogeological parameters, subgrade parameters, and pavement parameters to obtain preprocessed hydrogeological parameters, subgrade parameters, and pavement parameters;

[0102] Based on the geological complexity, hydrogeological parameters, roadbed parameters, and pavement parameters of the permafrost region, a target model framework is established.

[0103] Based on hydrogeological parameters, roadbed parameters, pavement parameters, and target model architecture, a hydrogeological model for permafrost regions is constructed.

[0104] Hydrogeological parameters describe the distribution, movement, and environment of groundwater in permafrost regions. These parameters include basic groundwater information, groundwater movement parameters, water quality and temperature data, and hydrological data of the soil and rock mass. Basic groundwater information includes groundwater level depth, groundwater type, and dynamic changes in groundwater level (the height of the water level relative to a specific elevation datum at different times). Groundwater movement parameters include permeability coefficient, hydraulic conductivity, and release coefficient. Water quality and temperature data include groundwater pH, chloride ion content, sodium ion content, and groundwater temperature at different depths. Hydrological data of the soil and rock mass includes aquifer thickness, aquitard thickness and distribution, porosity, and water content.

[0105] Subgrade parameters are the parameters of the load-bearing structural layer beneath roads in permafrost regions. Subgrade parameters include the physical parameters of the subgrade soil and rock, the mechanical properties of the subgrade, and the parameters related to the subgrade structure and permafrost. The physical parameters of the subgrade soil and rock include the type of subgrade fill material (e.g., silty clay, gravel), dry density (reflecting the degree of compaction), void ratio (measuring the density of the soil), liquid limit, and plastic limit (determining the plastic state of the soil). The mechanical properties of the subgrade include bearing capacity (the subgrade's ability to withstand loads), compression modulus (reflecting the soil's compressibility), and shear strength parameters (e.g., internal friction angle, cohesion). The parameters related to the subgrade structure and permafrost include subgrade height, slope gradient, thickness of the underlying permafrost, ice content, and upper limit depth of permafrost (depth of the top of the permafrost).

[0106] Pavement parameters refer to the structural layer parameters of the roadway in permafrost regions. These parameters primarily reflect the material properties and structural performance of the pavement. Pavement parameters include the physical properties of pavement materials, pavement mechanical properties, and pavement structural parameters. The physical properties of pavement materials include the density, porosity, and moisture content of each structural layer material (such as asphalt mixtures and cement concrete). The pavement mechanical properties include the compressive strength, flexural strength, and modulus of elasticity of the asphalt surface layer, as well as the flexural strength, compressive strength, and modulus of elasticity of the cement concrete surface layer. Pavement structural parameters include the thickness of each structural layer (such as the thickness of the asphalt surface layer, base course, and subbase course), the number of structural layers, and the bonding state between layers (such as continuity and the presence of interlayers).

[0107] Data preprocessing for hydrogeological parameters, subgrade parameters, and pavement parameters includes: cleaning the hydrogeological parameters, subgrade parameters, and pavement parameters to remove erroneous data and outliers; converting the hydrogeological parameters, subgrade parameters, and pavement parameters to ensure that the format and units of data of the same type are consistent; and using interpolation methods to fill in missing data.

[0108] The constructed target model architecture aligns with the modeling objectives of this application. The geological complexity of permafrost regions refers to the complexity of the permafrost medium. Furthermore, if the geological complexity of the permafrost region exceeds a complexity threshold and the description of physical processes is critical, a mechanistic model is selected. Further, if the data volume of hydrogeological parameters, subgrade parameters, and pavement parameters exceeds a third preset value, and there is correlation between various types of parameters, a statistical model is selected.

[0109] Furthermore, based on the seepage theory and heat transfer theory of porous media, an equation describing the multiphase flow and heat transfer process of groundwater in permafrost regions is selected to establish the target model framework. For example, Darcy's law is used to describe the flow of groundwater in permafrost regions, and the heat conduction equation is used to describe the heat transfer in permafrost regions.

[0110] Furthermore, the impact of freeze-thaw cycles on pore water phase, soil porosity, and permeability in permafrost regions is characterized by soil moisture freeze-thaw curves or constitutive relations based on physical processes (such as the Clausius-Clapyeron equation), thus obtaining the target model architecture.

[0111] By substituting hydrogeological parameters, roadbed parameters, and pavement parameters into the target model framework, a hydrogeological model of the permafrost region is obtained.

[0112] Furthermore, methods such as least squares estimation, Bayesian estimation, and maximum likelihood estimation are used to estimate some unknown or uncertain parameters in the hydrogeological model in order to optimize the hydrogeological model.

[0113] In this embodiment, hydrogeological parameters, subgrade parameters, and pavement parameters of the permafrost region are acquired, and the data of these parameters are preprocessed to obtain preprocessed hydrogeological parameters, subgrade parameters, and pavement parameters. Based on the preset modeling purpose, the geological complexity of the permafrost region, and the hydrogeological, subgrade, and pavement parameters, a target model architecture is established. Based on the hydrogeological parameters, subgrade parameters, pavement parameters, and target model architecture, a hydrogeological model of the permafrost region is constructed. This allows the constructed hydrogeological model to more realistically reflect the complex water-thermal-mechanical coupling process of the permafrost region, improving the model's accuracy and regional adaptability.

[0114] In one embodiment, the method further includes:

[0115] Simulation analysis is performed using a hydrogeological model to output simulated state parameters and obtain second actual state parameters.

[0116] Select target parameters that meet preset conditions from the second actual state parameters; the preset conditions include that the time to which the second actual state parameter belongs is in the critical period of the freeze-thaw cycle, the space to which the second actual state parameter belongs covers the space where multiple types of permafrost are located, the second actual state parameter is a non-outlier value, and the second actual state parameter has not been used to construct a hydrogeological model.

[0117] When the difference between the simulated state parameters and the target parameters is greater than the preset difference, the key model parameters of the hydrogeological model are optimized to obtain the optimized hydrogeological model, and the physical model of the permafrost region is obtained based on the optimized hydrogeological model.

[0118] The simulated state parameters include, but are not limited to, simulated groundwater parameters, simulated permafrost parameters, and simulated engineering parameters. Simulated groundwater parameters include groundwater level, groundwater flow velocity and direction, and the distribution range of different types of groundwater (above-freeze water, interlayer water, and below-freeze water). Simulated permafrost parameters include the upper limit depth of permafrost, the thawing depth of the active layer, and data on pore water phase changes under freeze-thaw cycles. Simulated engineering parameters include the distribution of subgrade moisture content and pore water pressure.

[0119] The second set of actual state parameters includes actual groundwater parameters, actual frozen soil parameters, and actual engineering parameters in permafrost regions.

[0120] The fact that the second actual state parameter belongs to a critical period of the freeze-thaw cycle means that the time when the second actual state parameter is generated falls within a critical period of the freeze-thaw cycle. The critical period of the freeze-thaw cycle includes summer and winter, with thawing in summer and freezing in winter. The fact that the second actual state parameter spatially covers the spaces of various permafrost types means that the spatial location of the second actual state parameter is within the space occupied by any type of permafrost.

[0121] The second actual state parameter not being used to construct the hydrogeological model means that the specific values ​​of the second actual state parameter are inconsistent with those of the hydrogeological parameters, subgrade parameters, and pavement parameters, but the data types can be consistent. This avoids data overfitting. For example, if the hydrogeological parameters, subgrade parameters, and pavement parameters are data from boreholes Z1 to Z5, then the second actual state parameter is data from boreholes Z6 and Z7.

[0122] When comparing the difference between simulated state parameters and target parameters, simulated state parameters of the same type, location, and time are compared with target parameters.

[0123] Key model parameters refer to parameters that significantly influence the simulated state parameters output by a hydrogeological model. The process of determining key model parameters involves analyzing the impact of different model parameters on the simulated state parameters output by the hydrogeological model using one or more of the following methods: single-factor analysis, multivariate statistical analysis, and Monte Carlo simulation. Based on the degree of influence of each model parameter, key model parameters are determined. For example, by fixing other parameters and changing only one model parameter (e.g., permeability coefficient ±20%), the magnitude of change in the model output (e.g., groundwater level) is observed. If there is a significant change, this model parameter is considered a key model parameter. Optimizing key model parameters in a hydrogeological model allows for focused optimization of core parameters, reducing the computational cost of irrelevant parameters; improving model calibration efficiency and quickly narrowing the error between simulation results and measured data; and clarifying key engineering control points. For instance, if the permeability coefficient is a key model parameter, drainage measures can be used to change the permeability coefficient, thereby controlling the groundwater level.

[0124] Furthermore, one or more of the following methods—gradient descent, genetic algorithm, and particle swarm optimization—are used to optimize the key model parameters, resulting in an optimized hydrogeological model.

[0125] Furthermore, a second performance evaluation index for the hydrogeological model is calculated based on the simulated state parameters and target parameters. This index is used to determine the accuracy of the hydrogeological model, and when the accuracy indicated by the second performance evaluation index does not meet the accuracy requirements, the key model parameters are optimized. The second performance evaluation index includes, but is not limited to, mean square error and coefficient of determination. The accuracy requirements include the mean square error reaching a preset value and the coefficient of determination reaching a preset value.

[0126] Furthermore, when the difference between the simulated state parameters and the target parameters exceeds a preset difference, the key model parameters and / or model structure of the hydrogeological model are optimized to obtain an optimized hydrogeological model. Model structure optimization includes mesh reconstruction, equation correction, and boundary condition optimization. For example, the model mesh in areas with drastic groundwater changes is densified; the influence coefficient of freeze-thaw cycles on permeability is adjusted; and the groundwater recharge boundary is corrected based on meteorological sensor data.

[0127] The optimized hydrogeological model can be used to simulate hydrogeological changes in permafrost regions, such as predicting groundwater level changes, analyzing groundwater flow direction and velocity, and studying the impact of freeze-thaw cycles on groundwater. Furthermore, by changing the model's input parameters, simulations under different scenarios can be conducted, providing a scientific basis for the design, construction, and maintenance of road engineering projects in permafrost regions.

[0128] In this embodiment, when the difference between the simulated state parameters and the target parameters is greater than a preset difference, the key model parameters of the hydrogeological model are optimized. This can, on the one hand, focus on optimizing core parameters and reduce the computational cost of irrelevant parameters; on the other hand, improve the efficiency of model calibration and quickly reduce the error between simulation results and measured data; and on the other hand, clarify the key points of engineering regulation.

[0129] In one embodiment, step S4 includes:

[0130] Simulation was performed using an initial digital twin model to output groundwater movement parameters and roadbed mechanical parameters in the permafrost region, and to obtain the actual groundwater movement parameters and actual roadbed mechanical parameters in the permafrost region.

[0131] Based on groundwater movement parameters, roadbed mechanical parameters, actual groundwater movement parameters, and actual roadbed mechanical parameters, the first performance evaluation index of the initial digital twin model is calculated.

[0132] If the first performance evaluation index fails to reach the first preset value, the initial digital twin model is determined to be unreliable.

[0133] When the first performance evaluation index reaches the first preset value, the reliability of the initial digital twin model is determined.

[0134] The initial digital twin model has basic state mapping capabilities. For example, it can simulate the real-time temperature distribution of the roadbed.

[0135] Groundwater movement parameters refer to the simulated groundwater movement parameters in permafrost regions. Examples include groundwater level changes and groundwater flow velocity. Subgrade mechanical parameters refer to the simulated road condition parameters in permafrost regions. Examples include subgrade settlement data. Actual groundwater movement parameters refer to the actual groundwater movement parameters in permafrost regions. Actual subgrade mechanical parameters refer to the actual road condition parameters in permafrost regions. The first set of actual condition parameters includes both actual groundwater movement parameters and actual subgrade mechanical parameters; the mapped condition parameters include both simulated groundwater movement parameters and subgrade mechanical parameters.

[0136] The primary performance evaluation metric is used to assess the accuracy of the initial digital twin model. The primary performance evaluation metric includes, but is not limited to, mean squared error and coefficient of determination.

[0137] When there are multiple first performance evaluation metrics, each first performance evaluation metric corresponds to a first preset value. A first performance evaluation metric failing to reach its first preset value means that one or more first performance evaluation metrics have failed to reach their corresponding first preset values. For example, a determination coefficient greater than or equal to the first preset value indicates that the accuracy of the initial digital twin model meets the standard; a determination coefficient less than the first preset value indicates that the accuracy of the initial digital twin model does not meet the standard.

[0138] Furthermore, the initial digital twin model is updated in real time using real-time monitoring data. For example, precipitation data collected by meteorological sensors can be input into the initial digital twin model in real time to correct the groundwater recharge of the initial digital twin model; data from stratified settlement sensors can reflect roadbed deformation and adjust the mechanical parameters of the soil and rock mass in the initial digital twin model, so that the initial digital twin model can dynamically reflect the actual changes in groundwater in roads in permafrost areas.

[0139] In this embodiment, an initial digital twin model is used for simulation to output groundwater movement parameters and roadbed mechanical parameters in the permafrost region. The actual groundwater movement parameters and actual roadbed mechanical parameters in the permafrost region are also obtained. Based on the groundwater movement parameters, roadbed mechanical parameters, actual groundwater movement parameters, and actual roadbed mechanical parameters, a first performance evaluation index of the initial digital twin model is calculated. If the first performance evaluation index does not reach a first preset value, the initial digital twin model is determined to be unreliable. If the first performance evaluation index reaches the first preset value, the initial digital twin model is determined to be reliable. In this way, the reliability of the model can be objectively judged through the quantifiable first performance evaluation index, thereby achieving a quantitative evaluation of the model's credibility.

[0140] In one embodiment, the optimization steps of the initial digital twin model include:

[0141] Step 1: Standardize the data format of monitoring variables: First, collect various key variables involved in groundwater monitoring in permafrost areas, including temperature, humidity, pore water pressure, and water flow velocity. Standardize the data format of these variables, with the following specific requirements: ① Timestamp alignment: All data collected by sensors should be recorded in the format of "year-month-day hour:minute:second" (e.g., 2024-06-10 14:30:00) to ensure consistency in the time base of data from different sources; ② Unit standardization: Use a unified unit for data of the same type, such as using degrees Celsius (°C) for temperature and kilopascals (kPa) for pressure; ③ Data type standardization: Store all data in the format of "value + unit," such as "12°C" and "80kPa," to avoid ambiguity; ④ Metadata annotation: Each data entry should include metadata such as the acquisition device number and acquisition location (e.g., "sensor 1-Z1 borehole - 2m depth") to ensure data traceability and facilitate subsequent verification and correlation analysis.

[0142] Step 2: Extracting Evaluation Indicators: Based on the various monitoring variables in the unified format described above, evaluation indicators that reflect the model's operational status and the actual situation of the monitored objects are extracted through statistical analysis. These evaluation indicators are a comprehensive reflection of the monitoring variables, and their values ​​are jointly determined by one or more related monitoring variables, quantifying the degree of fit between the model and the physical entity.

[0143] Step 3: Define the Target Optimization Direction: Based on the extracted evaluation indicators, define the target optimization direction of the model. The core of target optimization is to adjust relevant system parameters based on feedback from the evaluation indicators, ultimately enabling the model to more accurately reflect the actual state of groundwater in permafrost regions. In other words, the target optimization effect is reflected through changes in various evaluation indicators; when the evaluation indicators reach the preset ideal range, the model optimization achieves the expected goal.

[0144] Step 4: Solving with a Multi-Objective Optimization Algorithm: The optimization of the initial digital twin model is treated as a multi-objective optimization problem involving multiple agents. A Pareto-based multi-objective optimization algorithm is used to solve it. During the solution process, the following core elements should be noted: ① Objective Functions: Multiple objective functions are set, each corresponding to an objective to be achieved by the model optimization, thus forming the objective system for model optimization; ② Constraint Functions: Corresponding constraint functions are set to limit the value range of various variables during the optimization process (e.g., the penetration coefficient must not exceed a certain reasonable upper limit), ensuring that the optimization process conforms to the actual engineering scenario and physical laws; ③ Variable Set: A valid set of optimization variables is defined. These variables cover the key parameters of the digital twin (such as the penetration coefficient, neural network weights, model time step, etc.), and all variables are within a reasonable value range in a multi-dimensional real space. Through algorithmic solving, the optimal solution that satisfies all constraints and simultaneously achieves a relatively good state for multiple objective functions is found, completing the optimization of the initial digital twin model.

[0145] In one embodiment, the training process of a neural network model includes:

[0146] Obtain historical state parameters of permafrost regions and mapped state parameters output from the initial digital twin model;

[0147] Based on historical state parameters and mapped state parameters, a neural network model is trained to obtain the trained neural network model.

[0148] The mapping error is calculated based on the current state parameters of the permafrost region and the virtual state parameters output by the neural network model.

[0149] If the mapping error is greater than the second preset value, the weight parameters of the neural network model are optimized based on the mapping error. The optimized neural network model drives the initial digital twin model to generate virtual state parameters corresponding to the current state parameters until the mapping error is less than or equal to the second preset value, thus obtaining a pre-trained neural network model.

[0150] Among them, historical state parameters are freeze-thaw parameters of permafrost regions collected historically. Historical state parameters include, but are not limited to, historically collected groundwater level changes, water flow velocity, air temperature, specific location of data collection, and meteorological indices.

[0151] The initial digital twin model outputs mapping state parameters including groundwater movement parameters and roadbed mechanical parameters.

[0152] The historical state parameters and the mapped state parameters form the training set for the neural network model. The neural network model is used to output the virtual state parameters of permafrost regions. For example, the neural network model is a backpropagation (BP) neural network.

[0153] The current state parameters of permafrost regions refer to the real-time state parameters of the permafrost region. These include, for example, the real-time groundwater level, permeability coefficient, meteorological recharge, and freeze-thaw cycle intensity. The virtual state parameters output by the neural network model are those output by the initial digital twin model. For example, if the groundwater level in the permafrost region has risen by 0.5m, then the virtual groundwater level in the initial digital twin model should also rise by 0.5m simultaneously.

[0154] Mapping error refers to the difference between the current state parameters and the virtual state parameters.

[0155] Furthermore, the gradient descent method is used to optimize the weight parameters of the neural network model.

[0156] Furthermore, data transmission can be performed through a Web Service communication interface.

[0157] In this embodiment, by acquiring historical state parameters of the permafrost region and the mapped state parameters output by the initial digital twin model, a neural network model is trained based on the historical state parameters and the mapped state parameters to obtain the trained neural network model. Based on the current state parameters of the permafrost region and the virtual state parameters output by the neural network model, the mapping error is calculated. If the mapping error is greater than a second preset value, the weight parameters of the neural network model are optimized based on the mapping error, so that the optimized neural network model drives the initial digital twin model to generate virtual state parameters corresponding to the current state parameters, until the mapping error is less than or equal to the second preset value, and a pre-trained neural network model is obtained. In this way, a high-precision, adaptive, and real-time virtual mapping of the infrastructure state in the permafrost region can be achieved, thereby constructing a "highly consistent, continuously evolving, and reliable" digital twin model.

[0158] In one embodiment, the method further includes:

[0159] Predicting hydrological parameters in permafrost regions using digital twin models;

[0160] Based on hydrological parameters, the risk level of permafrost regions is determined using a risk assessment algorithm.

[0161] The hydrological parameters include, but are not limited to, groundwater level, flow velocity, and temperature in permafrost regions. A pre-trained neural network model is coupled into the digital twin model, which drives the digital twin model to make predictions, thereby obtaining the hydrological parameters of permafrost regions.

[0162] Furthermore, if the risk level matches the actual risk level in the permafrost region, it indicates that the predicted hydrological parameters are effective in supporting engineering decisions. If the risk level does not match the actual risk level in the permafrost region, the parameters in the risk assessment algorithm or the digital twin model should be adjusted to ensure that the risk level output by the risk assessment algorithm matches the actual risk level in the permafrost region. For example, if the risk assessment algorithm outputs a "road distress risk level" based on "water level rise of 1m", and the road distress risk level is "minor settlement risk", and there is indeed 2mm settlement in the permafrost region with no further expansion trend, it indicates that the prediction results are effective in supporting engineering decisions. If the road distress risk level represents "no risk", but settlement occurs in the permafrost region, then the parameters in the risk assessment algorithm or the digital twin model need to be adjusted.

[0163] Furthermore, if the error between the hydrological parameters predicted by the digital twin model and the actual hydrological parameters is greater than the fourth preset value, the neural network model is trained using the latest state parameters of the permafrost region so that the hydrological parameters predicted by the digital twin model are consistent with the actual hydrological parameters.

[0164] In this embodiment, hydrological parameters of permafrost regions are predicted using a digital twin model. Based on these parameters, a risk assessment algorithm is used to determine the risk level of the permafrost region, thus enabling a shift from "passive response" to "proactive early warning." This also enhances the scientific rigor and dynamism of risk assessment.

[0165] In one embodiment, step S4 includes:

[0166] Obtain the first actual state parameters of the permafrost region;

[0167] The mapped state parameters and the first actual state parameters are standardized and spatiotemporally aligned to obtain the processed state parameters.

[0168] The state parameters are segmented according to their respective time points to obtain multiple segments of state parameters;

[0169] The cumulative deviation is calculated based on the state parameters of each segment and the corresponding first actual state parameters. The reliability of the initial digital twin model is evaluated based on the cumulative deviation of each segment, and the evaluation result of the initial digital twin model is obtained.

[0170] Among them, the mapped state parameters are the various state parameters of the permafrost region mapped by the pre-trained neural network model driving the initial digital twin model.

[0171] Standardization includes unifying data formats and units.

[0172] Spatiotemporal alignment refers to aligning the time series of the mapped state parameters with the time series of the first actual state parameters, and aligning the positional sequence of the mapped state parameters with the positional sequence of the first actual state parameters.

[0173] The state parameters include the processed mapped state parameters and the first actual state parameters.

[0174] In a specific application, the prediction of "water level rising by 1m in the next month" is broken down into weekly segments. At the end of each week, the cumulative deviation between the predicted water level and the actual monitored water level is compared. If the weekly deviation is less than 0.05m and the total monthly deviation is less than 0.1m, it indicates that the initial digital twin model has high accuracy in water level prediction and the prediction trend is reliable. If the deviation suddenly increases in a certain week, it is necessary to investigate whether the initial digital twin model has missed key dynamic factors, such as sudden rainfall not being updated to meteorological data input in time, or temporary changes in permeability coefficient caused by freeze-thaw cycles not being captured by the logical model.

[0175] Furthermore, the rationality of the prediction mechanism of the digital twin model is verified. For example, if the hydrological parameters predicted by the digital twin model already include the correlation that "for every 0.3m rise in water level, the roadbed settlement increases by 1mm", and this correlation is consistent with the trend of on-site settlement data, then the prediction mechanism of the digital twin model is reliable.

[0176] In this embodiment, the state parameters are segmented according to their respective time points to obtain multiple segments of state parameters. The cumulative deviation is calculated based on each segment of state parameters and the corresponding first actual state parameter. The reliability of the initial digital twin model is evaluated based on the cumulative deviation of each segment to obtain the evaluation result of the initial digital twin model. This can improve the spatiotemporal accuracy and sensitivity of the model evaluation and avoid the problem of "overall average error masking local failure".

[0177] In some embodiments, a digital twin-based platform for monitoring and predicting groundwater levels along roads in permafrost regions is provided, and the platform's architecture diagram is shown below. Figure 6 As shown, the platform architecture diagram is as follows: Figure 7 As shown.

[0178] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0179] Based on the same inventive concept, this application also provides a digital twin-based permafrost monitoring device for implementing the aforementioned digital twin-based permafrost monitoring method. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations of one or more digital twin-based permafrost monitoring device embodiments provided below can be found in the limitations of the digital twin-based permafrost monitoring method described above, and will not be repeated here.

[0180] In one embodiment, such as Figure 8 As shown, a monitoring device for permafrost regions based on digital twins is provided, comprising:

[0181] The data acquisition module is used to acquire the hydrogeological model of the permafrost region and multi-source state data collected from the full cross-section of the road in the permafrost region, wherein the permafrost region is a soil and rock layer with a freezing time exceeding a preset time.

[0182] The coupling module is used to combine the preprocessed multi-source state data and the hydrogeological model to obtain the physical model of the permafrost region, and to couple the physical model, the pre-trained data model and the neural network model to obtain an initial digital twin model. The neural network is trained based on the historical state parameters and current state parameters of the permafrost region.

[0183] The mapping module is used to drive the initial digital twin model to perform virtual mapping and obtain mapping state parameters;

[0184] The evaluation module is used to evaluate the reliability of the initial digital twin model based on the mapped state parameters and the first actual state parameters of the permafrost region, and obtain the evaluation results.

[0185] A monitoring module is used to obtain a digital twin model when the evaluation results indicate that the initial digital twin model is reliable, so as to use the digital twin model to monitor multiple hydrological parameters of the permafrost region.

[0186] The modules in the aforementioned digital twin-based permafrost monitoring device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0187] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores various parameters. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a digital twin-based monitoring method for permafrost regions.

[0188] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0189] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0190] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0191] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0192] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0193] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0194] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A monitoring method for permafrost regions based on digital twins, characterized in that, The method includes: S1. Obtain the hydrogeological model of the permafrost region and multi-source state data collected from the full cross-section of the road in the permafrost region, wherein the permafrost region is a soil and rock layer with a freezing time exceeding a preset time. S2. Combining the preprocessed multi-source state data and the hydrogeological model, a physical model of the permafrost region is obtained, and the physical model, the pre-trained data model, and the neural network model are coupled to obtain an initial digital twin model. The neural network is trained based on the historical state parameters and current state parameters of the permafrost region. S3. Drive the initial digital twin model to perform virtual mapping to obtain mapping state parameters; S4. Obtain the first actual state parameters of the permafrost region; standardize and align the mapped state parameters with the first actual state parameters to obtain the processed state parameters; segment the state parameters according to their respective time points to obtain multiple segments of state parameters; calculate the cumulative deviation based on each segment of state parameters and the corresponding first actual state parameters; evaluate the reliability of the initial digital twin model based on the cumulative deviation of each segment to obtain the evaluation result of the initial digital twin model. S5. When the evaluation results indicate that the initial digital twin model is reliable, a digital twin model is obtained to monitor multiple hydrological parameters of the permafrost region.

2. The method according to claim 1, characterized in that, The process of constructing the hydrogeological model of the permafrost region includes: Obtain hydrogeological parameters, roadbed parameters, and pavement parameters in permafrost regions; The hydrogeological parameters, the roadbed parameters, and the pavement parameters are preprocessed to obtain the preprocessed hydrogeological parameters, roadbed parameters, and pavement parameters. Based on the geological complexity of the permafrost region, the hydrogeological parameters, the roadbed parameters, and the pavement parameters, a target model architecture is established. Based on the hydrogeological parameters, the roadbed parameters, the pavement parameters, and the target model architecture, a hydrogeological model of the permafrost region is constructed.

3. The method according to claim 1 or 2, characterized in that, The method further includes: Simulation analysis is performed using the hydrogeological model to output simulated state parameters and obtain second actual state parameters. Select target parameters that meet preset conditions from the second actual state parameters; the preset conditions include that the time to which the second actual state parameter belongs is in the critical period of the freeze-thaw cycle, the space to which the second actual state parameter belongs covers the space where multiple types of permafrost are located, the second actual state parameter is a non-outlier value, and the second actual state parameter has not been used to construct the hydrogeological model. When the difference between the simulated state parameters and the target parameters is greater than a preset difference, the key model parameters of the hydrogeological model are optimized to obtain the optimized hydrogeological model, and the physical model of the permafrost region is obtained based on the optimized hydrogeological model.

4. The method according to claim 1, characterized in that, Step S4 includes: Simulation is performed using the initial digital twin model to output the groundwater movement parameters and roadbed mechanical parameters of the permafrost region, and to obtain the actual groundwater movement parameters and actual roadbed mechanical parameters of the permafrost region. Based on the groundwater movement parameters, the roadbed mechanical parameters, the actual groundwater movement parameters, and the actual roadbed mechanical parameters, calculate the first performance evaluation index of the initial digital twin model; If the first performance evaluation index fails to reach the first preset value, the initial digital twin model is determined to be unreliable. When the first performance evaluation index reaches the first preset value, the initial digital twin model is determined to be reliable.

5. The method according to claim 1, characterized in that, The training process of the neural network model includes: Obtain the historical state parameters of the permafrost region and the mapped state parameters output by the initial digital twin model; Based on the historical state parameters and the mapped state parameters, a neural network model is trained to obtain the trained neural network model. The mapping error is calculated based on the current state parameters of the permafrost region and the virtual state parameters output by the neural network model. If the mapping error is greater than the second preset value, the weight parameters of the neural network model are optimized based on the mapping error, so as to drive the initial digital twin model to generate virtual state parameters corresponding to the current state parameters through the optimized neural network model, until the mapping error is less than or equal to the second preset value, and a pre-trained neural network model is obtained.

6. The method according to claim 1, characterized in that, The method further includes: The digital twin model is used to predict the hydrological parameters of the permafrost region. Based on the hydrological parameters, the risk level of the permafrost region is determined using a risk assessment algorithm.

7. A monitoring device for permafrost regions based on digital twins, used to perform the method according to any one of claims 1-6, characterized in that, The device includes: The data acquisition module is used to acquire the hydrogeological model of the permafrost region and multi-source state data collected from the full cross-section of the road in the permafrost region, wherein the permafrost region is a soil and rock layer with a freezing time exceeding a preset time. The coupling module is used to combine the preprocessed multi-source state data and the hydrogeological model to obtain the physical model of the permafrost region, and to couple the physical model, the pre-trained data model and the neural network model to obtain an initial digital twin model. The neural network is trained based on the historical state parameters and current state parameters of the permafrost region. The mapping module is used to drive the initial digital twin model to perform virtual mapping and obtain mapping state parameters; The evaluation module is used to evaluate the reliability of the initial digital twin model based on the mapped state parameters and the first actual state parameters of the permafrost region, and obtain the evaluation results. A monitoring module is used to obtain a digital twin model when the evaluation results indicate that the initial digital twin model is reliable, so as to use the digital twin model to monitor multiple hydrological parameters of the permafrost region.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1 to 6.

Images