Anti-freezing and heat preservation system for plateau cold region tunnel and control method thereof

Abstract

The application discloses a highland cold region tunnel anti-freezing and heat preserving system and a control method thereof, relates to the technical field of tunnel heat preservation, and comprises, from inside to outside, a sprayed concrete layer, a first water-proof layer, a heat insulation buffer layer, an electric heating and temperature testing layer, tunnel lining, a reinforced concrete arch ring, a second water-proof layer, secondary lining, a second layer of geotextile, a composite heat preserving layer, a finishing layer and a facing layer which are sequentially arranged on a tunnel rock wall; the electric heating and temperature testing layer comprises electric heating wires and a plurality of temperature sensors; the application further comprises a control box, a neural network controller is arranged in the control box, signal lines of the temperature sensors are connected with input ports of the neural network controller, and control circuits of the electric heating wires are connected with output ports of the neural network controller. The application effectively solves the anti-freezing and heat preserving problems of highland tunnels, improves the durability and safety of tunnel structures, and reduces operation and maintenance costs.

Description

Technical Field

[0001] This invention relates to the field of tunnel insulation technology, specifically to a frost-proof insulation system and control method for tunnels in high-altitude and cold regions. Background Technology

[0002] The construction and planning of railways in plateau regions necessitate tunnel engineering in the harsh environment of the plateau. This harsh environment primarily encompasses both severe geological and climatic conditions. Tunnels in harsh geological environments are subject to long-term freeze-thaw cycles, and construction disturbances can trigger disasters such as avalanches at the tunnel entrance and collapses in fractured zones. The harsh climatic environment results in poor construction conditions for plateau tunnels, with low temperatures, lack of oxygen, and extremely challenging working conditions within the tunnels. These harsh environments present numerous significant challenges in plateau tunnel construction, such as avalanches caused by disturbances at the tunnel entrance, collapses due to insufficient support structure strength in unfavorable geological sections, and structural failures induced by freeze-thaw cycles.

[0003] From a construction perspective, the temperature of the surrounding rock in tunnels is affected by changes in the external environment, and varies depending on the tunnel's depth. Tunnels experience significant temperature differences between winter and summer, with long, harsh winters. This leads to frost heave under the influence of low temperatures, seepage, and groundwater, causing damage to the rock mass and support structures, and greatly increasing the risk of engineering accidents. Currently, with the expansion of highway and railway networks, more and more tunnel projects are being undertaken in high-latitude, frigid regions. However, based on engineering cases from around the world, the construction of high-latitude, frigid tunnels under complex geological conditions presents more technical challenges than tunnels in other regions. Improper handling can result in substantial economic losses. The effectiveness of the insulation layer construction within the tunnel directly affects its normal operation and service life. Therefore, how to adopt effective measures during the construction of these tunnels to achieve the expected results during use is a pressing problem for the engineering community. Summary of the Invention

[0004] To address the aforementioned problems, this invention aims to provide a frost-proof and heat-insulating system and its control method for tunnels in high-altitude and cold regions. This system overcomes the shortcomings of existing technologies and effectively solves the frost-proof and heat-insulating challenges of high-altitude tunnels through innovative structural design, intelligent sensing systems, and intelligent control mechanisms based on neural networks. It also improves the durability and safety of tunnel structures and reduces operation and maintenance costs.

[0005] The main idea of ​​the technical solution adopted in this invention is as follows: By designing a multi-layered composite structure combining a buffer layer, a heating layer, and a composite insulation layer, the insulation and anti-freezing capabilities are enhanced from a physical perspective. For water-rich areas, an integrated drainage and anti-freezing system combining drainage, insulation, and heat tracing is designed. Drainage pipes are wrapped with insulation material, and automatic drainage valves are installed at low points. Combined with electric heat tracing, this prevents accumulated water from freezing and clogging the drainage system. In terms of temperature control, intelligent predictive control based on neural networks is introduced to replace traditional threshold switching control. When the system predicts a sharp drop in future temperature, it preheats the lining with higher power in advance to avoid excessive temperature fluctuations and prevent surface freezing. For temperature prediction, an XGBoost-LSTM fusion model is constructed. Through weighted fusion, the prediction results of both models are combined, resulting in significantly higher accuracy than a single algorithm.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The anti-freezing and heat preservation system for tunnels in high-altitude and cold regions includes, from the inside out, a shotcrete layer, a first waterproof layer, a heat insulation buffer layer, an electric heating and temperature testing layer, a tunnel lining, a reinforced concrete arch, a second waterproof layer, a secondary lining, a second layer of geotextile, a composite insulation layer, and a plastering and finishing layer. The electric heating and temperature testing layer includes an electric heating wire and multiple temperature sensors; It also includes a control box, which contains a neural network controller. The signal line of the temperature sensor is connected to the input port of the neural network controller, and the control circuit of the electric heating wire is connected to the output port of the neural network controller.

[0007] This invention also provides a method for frost protection and insulation construction of tunnels in high-altitude and cold regions, comprising: Shotcrete is sprayed onto the tunnel walls, and after the concrete hardens, it forms a shotcrete layer.

[0008] The first layer of geotextile is fixed to the outside of the shotcrete layer using steel nails, forming the first waterproof layer.

[0009] A heat insulation buffer layer is set on the outside of the first waterproof layer. The heat insulation buffer layer is set on the shallow buried side of the tunnel and is evenly attached to the tunnel wall in the circumferential direction.

[0010] An electric heating and temperature testing layer, including an electric heating wire and a temperature sensor, is installed on the outside of the heat insulation buffer layer. A control box is also installed inside the tunnel, which contains a neural network controller. The signal line of the temperature sensor is connected to the input port of the neural network controller, and the control circuit of the electric heating wire is connected to the output port of the neural network controller.

[0011] The tunnel lining, reinforced concrete arch, second waterproof layer, and secondary lining structure are constructed sequentially.

[0012] A second layer of geotextile is laid on the surface of the secondary lining, followed by a composite insulation layer. The composite insulation layer includes a nano-aerogel felt layer, a flame-retardant polyurethane insulation layer, and a waterproof and breathable membrane layer arranged from the inside out. The outermost layer is coated with a finishing layer and a finishing layer; A drainage and antifreeze system is installed on the outside of the tunnel lining.

[0013] Further, based on the above technical solutions: the heat insulation buffer layer is a polystyrene foam board with a thickness of 5-8cm, which is adhered to the surface of the first waterproof layer with adhesive and fixed with expansion bolts and pressure plates.

[0014] Further, based on the above technical solution: the electric heating wires are evenly arranged in a serpentine pattern on the outside of the heat insulation buffer layer, with a spacing of 30-50cm; the temperature sensors are arranged at the tunnel arch, arch waist, arch shoulder, invert arch, and arch foot to form a temperature sensor network.

[0015] Furthermore, based on the above technical solutions, the drainage and antifreeze system includes a ring-shaped drainage blind ditch, insulated drainage pipes, and an electric heat tracing system. A ring-shaped drainage blind ditch is set around the bottom of the outside of the tunnel lining, with a perforated HDPE drainage pipe buried inside. The drainage pipe is wrapped with geotextile. The drainage pipe is wrapped with rubber and plastic insulation board, and an automatic drainage valve is installed at the lowest point. The electric heat tracing system includes an electric heat tracing tape wrapped around the insulated drainage pipe, which is connected to a neural network controller.

[0016] This invention also provides a method for controlling frost prevention and heat preservation in tunnels in high-altitude and cold regions, comprising: Temperature data at various monitoring points in the tunnel are collected in real time using temperature sensors. Temperature data is preprocessed to obtain effective temperature data; The effective temperature data is transmitted to the neural network controller; The neural network controller analyzes and processes the effective temperature data based on a pre-trained neural network model, predicts the temperature of the tunnel at the next moment, and outputs the final temperature prediction value. Based on the final temperature prediction, the heating power and heating duration of the electric heating wire and electric heat tracing system are dynamically controlled to adjust the temperature inside the cave to remain within the set range at the next moment.

[0017] Furthermore, based on the above technical solutions, the neural network model includes a first algorithm module, a second algorithm module, and an output module. The first algorithm module outputs the first predicted value based on the XGBoost model; The second algorithm module outputs a second predicted value based on the LSTM model; The output module calculates the final temperature prediction value based on the first and second predicted values ​​and their corresponding weights.

[0018] Furthermore, the weights of the first and second predicted values ​​are adaptively adjusted using the AdaGrad algorithm. The weights are iteratively updated by minimizing the loss function until the loss function converges or the preset number of iterations is reached.

[0019] Based on the above technical solutions, the loss function is further defined as follows:

[0020] in, For loss function, These are the weighting coefficients. This represents the total number of samples, i.e., the number of data points used to calculate the loss. For sample index, representing the first... One sample; For the first The actual temperature value of each sample; For the first The first predicted value for each sample; For the first The second predicted value for each sample.

[0021] The beneficial effects of this invention are: 1. Multi-layer composite insulation structure: It changes the traditional single insulation layer design and adopts multiple protections of buffer layer, heating layer and composite insulation layer.

[0022] Inner layer insulation buffer: Polystyrene foam board is used as an insulation buffer layer, which can both keep the lining warm and buffer the compression damage caused by frost heave deformation of the surrounding rock. Intermediate active heating: An electric heating and temperature testing layer is added, upgrading passive insulation to active antifreeze. When extreme low temperatures occur or the insulation layer is insufficient to maintain the temperature, it can actively provide heat to the lining, ensuring that the lining structure itself does not freeze and guaranteeing structural safety. Outer high-efficiency composite insulation: A combination of nano-aerogel felt, PU polyurethane + waterproof and breathable membrane is used, taking into account both insulation and moisture removal. It can maintain a stable internal temperature in the tunnel even in extreme low-temperature environments, reducing the workload of the heating system.

[0023] 2. Anti-freezing design of the drainage system: For water-rich areas, an integrated anti-freezing drainage system combining drainage, insulation, and heat tracing is designed. Drainage pipes are wrapped with insulation material, and automatic drain valves are installed at low points. Combined with electric heat tracing, this prevents accumulated water from freezing and clogging the drainage system.

[0024] 3. Intelligent Control Method: A neural network-based intelligent predictive control method is introduced to replace traditional threshold switching control. When the system predicts a sharp drop in temperature, it preheats the lining with higher power to avoid excessive temperature fluctuations and prevent surface freezing. The system not only considers the current temperature but also incorporates forecast data for the next 12-24 hours, comprehensively considering equipment start-up and shutdown costs and energy consumption constraints. It continuously optimizes the heating strategy to stabilize the temperature within a safe range, achieving on-demand heating and reducing energy consumption. This overcomes the drawbacks of traditional temperature control systems' "lagging adjustment," avoids large temperature fluctuations, and improves the comfort of the tunnel environment and the safety of the structure.

[0025] 4. Neural Network Model: An XGBoost-LSTM fusion model was constructed. XGBoost excels at handling the influence of structured features such as tunnel length, slope, inlet / outlet elevation difference, and wind speed on temperature. LSTM excels at capturing the temporal dependence of temperature changes over time. By weighted fusion, the prediction results of the two models are combined, and the weights are dynamically adjusted using the AdaGrad algorithm, achieving a coefficient of determination (R²) of 0.97 and a mean absolute error (MAE) of 0.94, significantly outperforming single algorithms.

[0026] Employing a model predictive control framework, the system not only considers temperature requirements when controlling heating but also continuously optimizes the control sequence, comprehensively taking into account equipment start-up and shutdown costs and energy consumption constraints to achieve on-demand and precise heating. This avoids ineffective energy waste and, compared to traditional constant power heating or simple threshold switching control, significantly saves electricity and reduces the long-term power costs of operating high-altitude tunnels. Attached Figure Description

[0027] Figure 1 This is a cross-sectional schematic diagram of a tunnel structure layer according to a specific embodiment of this application; Figure 2 This is a schematic diagram of the neural network model training process according to a specific embodiment of this application; Figure 3 This is a schematic diagram of a neural network predictive control process according to a specific embodiment of this application; Figure 4 This is a structural diagram of the plateau anti-freezing and heat-insulating tunnel in this invention; Figure 5 The images show the temperature prediction results inside tunnels in cold regions using different algorithms; (a) is the LSTM algorithm, (b) is the XGBoost algorithm, (c) is the fusion algorithm, and (d) is the temperature prediction error inside the tunnel using different algorithms. Figure 6 This is an example of verifying the predicted temperature inside a tunnel in a cold region. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0029] The inventors discovered that, from a construction perspective, the temperature of the surrounding rock in tunnels is affected by changes in the external environment, and varies depending on the tunnel's depth. Tunnels experience significant temperature differences between winter and summer, with long, harsh winters. This leads to frost heave under the influence of low temperatures, seepage, and groundwater, causing damage to the rock mass and support structures, and greatly increasing the risk of engineering accidents. Currently, with the expansion of highway and railway networks, more and more tunnel projects are being undertaken in high-latitude, frigid regions. However, based on engineering cases from around the world, the construction of high-latitude, frigid tunnels under complex geological conditions presents more technical challenges than tunnels in other regions. Improper handling can result in substantial economic losses. The effectiveness of the insulation layer construction within the tunnel directly affects its normal operation and service life. Therefore, how to adopt effective measures during the construction of these tunnels to achieve the expected results during use is a pressing problem for the engineering community.

[0030] Based on the above findings, this application proposes a frost-proof and heat-insulating system and its control method for tunnels in high-altitude and cold regions. By designing a multi-layered composite structure combining a buffer layer, a heating layer, and a composite insulation layer, the system enhances heat insulation and frost protection capabilities from a physical perspective. For water-rich areas, an integrated drainage and frost-proof system combining drainage, insulation, and heat tracing is designed. Drainage pipes are wrapped with insulation material, and automatic drainage valves are installed at low points, combined with electric heat tracing to prevent water accumulation and freezing, which could clog the drainage system. For temperature control, a neural network-based intelligent predictive control is introduced, replacing traditional threshold switching control. When the system predicts a sharp drop in temperature, it preheats the lining with higher power to avoid excessive temperature fluctuations and prevent surface freezing. For temperature prediction, an XGBoost-LSTM fusion model is constructed. Through weighted fusion, the prediction results of both models are combined, resulting in significantly higher accuracy than a single algorithm.

[0031] Example 1: Construction Method for Anti-freezing and Thermal Insulation of Tunnels in High-Altitude and Cold Regions See Figures 1-6 This application discloses a method for frost protection and insulation construction of tunnels in high-altitude and cold regions, including the following steps: S1. First, spray concrete onto the tunnel rock wall. After the concrete solidifies, a spray concrete layer is formed.

[0032] S2. The first layer of geotextile is fixed to the outside of the shotcrete layer with steel nails to form the first waterproof layer.

[0033] S3. A heat insulation buffer layer is set on the outside of the first water-proof layer. The heat insulation buffer layer is set on the shallow buried side of the tunnel and is evenly attached to the tunnel wall in the circumferential direction, with one layer applied every 5m.

[0034] S4. An electric heating and temperature testing layer, including an electric heating wire and a temperature sensor, is installed on the outside of the heat insulation buffer layer. A control box is also installed inside the tunnel, and a neural network controller is installed inside the control box. The signal line of the temperature sensor is connected to the input port of the neural network controller, and the control circuit of the electric heating wire is connected to the output port of the neural network controller.

[0035] S5. Construct the tunnel lining, reinforced concrete arch, second waterproof layer, and secondary lining structure in sequence.

[0036] S6. Lay a second layer of geotextile on the surface of the secondary lining, and then install a composite insulation layer; the composite insulation layer includes a nano aerogel felt layer, a PU flame-retardant polyurethane insulation layer and a waterproof and breathable membrane layer arranged from the inside to the outside. S7. The outermost layer is covered with a finishing layer and a decorative layer.

[0037] Specifically, in S3, a polystyrene foam board, 5-8 cm thick, is installed on the outside of the first waterproof layer as a thermal insulation buffer layer. After the polystyrene foam board is cut to the design dimensions, it is adhered to the surface of the first layer of geotextile using adhesive, ensuring a tight, gapless joint between the foam boards. The foam boards can be fixed using expansion bolts and clamps, with a fixing point spacing of no more than 60 cm, ensuring a tight fit between the thermal insulation buffer layer and the surrounding rock and composite insulation layer.

[0038] This heat insulation buffer layer can further enhance the heat preservation effect on the one hand, and buffer the impact of the freezing and thawing deformation of the surrounding rock on the tunnel lining structure on the other hand, reducing the risk of cracking of the lining structure caused by the deformation of the surrounding rock.

[0039] In section S4, the electric heating and temperature testing layer includes electric heating wires installed outside the insulation buffer layer. These wires are arranged in a serpentine pattern with a spacing of 30–50 cm, serving as the primary heating element to provide heat to the tunnel lining and prevent the lining structure from freezing. The heating wires are connected using high-temperature resistant insulated wires, and the connection points are wrapped with waterproof insulating tape to ensure a secure connection and good insulation. The heating wires can be secured to the lining reinforcement using plastic cable ties.

[0040] Temperature sensors are densely distributed at various locations in the tunnel's surrounding environment, lining structure, and composite insulation layer, specifically at the tunnel's arch crown, arch waist, arch shoulders, invert arch, and arch feet, forming a temperature sensor network. The temperature sensors are high-precision, low-temperature-resistant digital sensors, capable of collecting temperature data from each monitoring point in real time and accurately. The sampling frequency can be set according to actual needs; for example, during periods of frequent freeze-thaw cycles, the sampling frequency can be increased to once per minute. During installation, ensure close contact between the sensor and the monitoring location to guarantee the accuracy of temperature data acquisition. The sensor signal lines are shielded cables and are introduced into the control box inside the tunnel through pre-embedded conduits, with proper waterproofing and insulation measures taken during the connection process.

[0041] A control box is installed at a suitable location inside the tunnel, housing the hardware (such as a server and controller) of the neural network control module. The signal lines of the temperature sensor are connected to the input port of the control module, and the control circuit of the electric heating wire is connected to the output port of the control module. This allows for dynamic and precise control of the heating power and heating duration of the electric heating wire based on temperature data trends and predictions.

[0042] Through programming and debugging, the pre-trained neural network model is deployed into the control module, and on-site testing and optimization are carried out to ensure that the neural network control module can accurately receive temperature data and control the operation of the electric heating wire. For example, when it is predicted that the temperature will drop sharply in the near future, the neural network controls the electric heating wire to preheat at a higher power in advance; when the temperature approaches the set threshold, it achieves stable temperature control by fine-tuning the heating power, thus avoiding energy waste and excessive temperature fluctuations.

[0043] In S5, the second waterproof layer can be a waterproof membrane.

[0044] In S6, the nano-aerogel felt layer is fixed to the second geotextile layer with steel nails during installation. The nano-aerogel felt layer and the second geotextile layer are then gradually bonded together. The nano-aerogel felt layer is 2–3 cm thick. It has an extremely low thermal conductivity, further enhancing the insulation effect, and is lightweight, thus not adding excessive load to the tunnel structure.

[0045] The flame-retardant polyurethane insulation layer uses polyurethane material, which has excellent thermal insulation and waterproof performance. It can effectively prevent low external temperatures from penetrating into the tunnel surrounding rock and prevent groundwater from seeping into the lining structure. Before laying, the lining surface must be cleaned to ensure it is flat and clean. Professional polyurethane spraying equipment is used for on-site spraying to a design thickness of 8-12cm. During spraying, the spraying pressure and speed are controlled to ensure a uniform thickness, smooth surface, and no defects such as sagging or blistering of the polyurethane insulation layer.

[0046] The waterproof and breathable membrane layer prevents external moisture from entering the insulation layer while allowing moisture inside the insulation layer to escape, avoiding a decrease in insulation performance due to moisture accumulation and damage to the insulation layer caused by freezing expansion at low temperatures. During installation, the waterproof and breathable membrane is laid along the tunnel axis, and the membrane materials are connected using hot air welding. The weld width is no less than 10cm, ensuring a firm weld and no leakage. The waterproof and breathable membrane is fixed using plastic expansion bolts and pressure strips, with a fixing point spacing of no more than 50cm to ensure a tight fit between the waterproof and breathable membrane and the insulation layer.

[0047] During tunnel construction and operation, a temperature sensor network collects temperature data from each monitoring point in real time and transmits it to a neural network control module. The control module analyzes and processes the data according to a preset program and neural network model, automatically controlling the operation of the electric heating wires and the electric heat tracing system. Simultaneously, staff can monitor the temperature data and equipment operating status in real time through a monitoring terminal. When abnormalities are detected (such as sudden temperature changes or equipment malfunction alarms), timely analysis and handling are carried out.

[0048] Example 2: Drainage and Antifreeze System Construction sites with large water volumes are equipped with drainage and antifreeze systems, including ring-shaped drainage blind ditches, insulated drainage pipes, and electric heat tracing systems.

[0049] During tunnel excavation, once the design elevation is reached, a ring-shaped trench is excavated around the bottom of the outer side of the tunnel lining. The trench dimensions are determined according to design requirements. This ring-shaped trench serves as a ring-shaped drainage blind ditch, filled with highly permeable graded crushed stone to a thickness of at least 30cm. Perforated HDPE drainage pipes with a diameter of 15-20cm are buried within the ring-shaped drainage blind ditch. The drainage pipes are surrounded by graded crushed stone to the design height to ensure smooth drainage. A layer of geotextile is wrapped around the HDPE drainage pipes, connected by stitching with an overlap width of at least 10cm to prevent silt and other impurities from entering and causing blockages.

[0050] Install the drainage pipes according to the designed route. After installation, wrap the drainage pipes with insulation material. Use rubber-plastic insulation boards with good thermal insulation and low-temperature resistance, with a thickness of at least 5cm. Use specialized adhesive to bond the insulation boards together, ensuring a tight, gapless fit. Simultaneously, install an automatic drain valve at the lowest point of the drainage pipes. The installation of the automatic drain valve should comply with the product manual, ensuring it can open and close normally. When the water accumulation in the pipes reaches a certain level, the automatic drain valve will open to drain the water, preventing it from freezing inside the pipes. After the insulated drainage pipes are installed, the electric heat tracing system is installed. The electric heat tracing tape is wrapped around the drainage pipes, with the wrapping spacing determined according to design requirements, generally 10-15cm. The power supply terminals and the tail end of the electric heat tracing tape are connected using a dedicated junction box, ensuring secure wiring and good insulation. The electric heat tracing system is also connected to a neural network control module. A temperature sensor network collects temperature data from the drainage pipes, and the neural network controls the start, stop, and heating power of the electric heat tracing system based on the data, ensuring that the water in the drainage pipes can drain smoothly and preventing freezing.

[0051] Example 3: Method for Constructing a Temperature Prediction Model for Tunnels in High-Altitude and Cold Regions Compared with the prior art, the embodiments of the present invention adopt a neural network model. This model takes a large amount of temperature data collected by a temperature sensor network as input, and learns the complex laws and characteristics of temperature changes in plateau tunnels through training. It can predict the temperature inside the tunnel and control the temperature inside the tunnel within a set range, thus solving the problem of antifreeze and heat preservation in tunnels in cold plateau regions.

[0052] The following is a detailed description of the implementation details of the antifreeze and heat preservation control method for tunnels in high-altitude and cold regions according to this embodiment. The following content is only for the convenience of understanding and is not necessary for implementing this solution.

[0053] This invention also provides a method for controlling frost prevention and heat preservation in tunnels in high-altitude and cold regions, comprising: Step 1: Collect temperature data at each monitoring point in the tunnel in real time using temperature sensors; Step 2: Preprocess the temperature data to obtain effective temperature data; transmit the effective temperature data to the neural network controller; Step 3: The neural network controller analyzes and processes the effective temperature data based on the pre-trained neural network model, predicts the temperature of the tunnel at the next moment, and outputs the final temperature prediction value. Based on the final temperature prediction, the heating power and heating duration of the electric heating wire and electric heat tracing system are dynamically controlled to adjust the temperature inside the cave to remain within the set range at the next moment.

[0054] Specifically, in step one, temperature data from various temperature monitoring points in the tunnel is collected. Multidimensional time-series and spatial-series data are collected through a temperature sensor network, including: real-time temperature values ​​at different depths in the tunnel (10-15 monitoring points), ambient temperature, wind speed, meteorological data such as the elevation difference between the entrance and exit, and seasonal characteristic codes (such as month and day-night cycle).

[0055] Several existing through tunnels were selected as test tunnels, and the cold-region where the test tunnels are located was designated as the study area. This study area comprises high-altitude cold regions and provinces with high-altitude cold climates. In addition to the data collected above, temperature monitoring data from these tunnels can also be used as supplementary data.

[0056] In step two, the collected tunnel data is used to pre-train the model, and then fine-tuned for specific projects. A historical monitoring dataset of at least 12 months is established, including temperature changes and control responses under different seasons and weather conditions; the raw data is processed by a sliding window, with the window size set to 48 hours (considering the diurnal temperature range cycle on the plateau); the input features are normalized using a standardization method, and random noise is added to simulate sensor measurement errors to enhance the robustness of the model. The loss function is shown in formula (5).

[0057] In step three, the neural network model employs a hybrid architecture combining a Long Short-Term Memory (LSTM) network and a gradient boosting algorithm (XGBoost) to optimize the temporal and nonlinear characteristics of temperature control in plateau tunnels. To combine the advantages of both algorithms and improve the accuracy and practicality of the prediction results, this invention leverages the strengths of both the XGBoost algorithm and the LSTM model, proposing a gradient descent-based XGBoost-LSTM fusion algorithm by weightedly fusing the prediction results of XGBoost and LSTM.

[0058] The fusion algorithm incorporates factors that significantly affect tunnel temperature, including: tunnel length, tunnel slope, entrance / exit elevation difference, ambient temperature, wind speed, time, and depth variation. For the specific neural network model construction process, please refer to [link / reference]. Figure 1 The specific algorithm structure is as follows: 1) The first algorithm module uses the XGBoost model to predict the temperature inside the cave, with the input feature vector being X=[l, d, T]. env [t, v, θ, h], where each eigenvector represents: tunnel length Tunnel depth d, ambient temperature T env Given time t, wind speed v, slope direction θ, and inlet / outlet elevation difference h, and considering that XGBoost consists of K decision trees, the predicted temperature T inside the cave by XGBoost is... XGB for: (1); (2); In the formula: the predicted value of the k-th decision tree for the input feature X is denoted as f. k (x), T XGB This is the first predicted value.

[0059] For the input gate at time The output value ranges from [0,1].

[0060] The Sigmoid activation function compresses the output to the [0,1] interval.

[0061] For input The weight matrix to the input gate.

[0062] The hidden state of the previous moment The weight matrix to the input gate.

[0063] For a moment The input vector.

[0064] For a moment The hidden state vector.

[0065] This is the bias term for the input gate.

[0066] Forget the gate at any time The output value ranges from [0,1].

[0067] For input The weight matrix to the forget gate.

[0068] The state that was activated at the previous moment The weight matrix to the forget gate.

[0069] This is the bias term for the forget gate.

[0070] For a moment The candidate cell states have values ​​ranging from [-1, 1].

[0071] The hyperbolic tangent activation function compresses the output to the interval [-1, 1].

[0072] For input The weight matrix for candidate cell states.

[0073] The hidden state of the previous moment The weight matrix for candidate cell states.

[0074] This is a bias term for the candidate cell state.

[0075] For the output gate at time The output value ranges from [0,1].

[0076] For input The weight matrix of the output gate.

[0077] The hidden state of the previous moment The weight matrix of the output gate.

[0078] This is the bias term for the output gate.

[0079] For a moment The cellular state.

[0080] For a moment The cellular state.

[0081] This is the Hadamard product (element-wise multiplication).

[0082] For a moment The hidden state serves as the input for the next time step and the output for the current time step.

[0083] The cell state is compressed into the range [-1,1][-1,1] for output.

[0084] 2) Second algorithm module: Let T be the temperature inside the cave over the past n time steps. t-n T t-n+1 ... T t-1 Ambient temperature time series T env,t-n T env,t-n+1 ... T env,t-1 Wind speed time sequence v t-n v t-n+1 ..., v t-1 Position and timing p t-n p t-n+1 ... p t-1 The time series data are concatenated according to the practical section to form a feature matrix X. t-n+t-1 row j X t-n+t-1 =[T t-n+j-1 ,T env,t-n+j-1 ,v t-n+j-1 ,p t-n+j-1 ], j=1,2,…,n. The feature vector x at the current time step. t =[d,T env ,t,v]. And use equation (2) to calculate h. t Finally, through a fully connected layer, the LSTM's predicted value T for the temperature inside the hole is obtained.LSTM for: (3); In the formula: T LSTM This is the second predicted value. This is the weight matrix of the fully connected layer, used to weight the hidden states. Linear transformation to temperature prediction space. This is a bias term for the fully connected layer, used to adjust the intercept of the linear transformation and improve the model's fitting ability.

[0085] 3) By weighted fusion of the prediction results from XGBoost and LSTM, the final predicted temperature T inside the cave is: (4); In the formula: α is the weighting coefficient, which needs to be continuously updated and iterated using the AdaGrad algorithm. The specific process is as follows: ①Assuming historical data is The weights of the XGBoost predictions are initialized to α, and the weights of the LSTM predictions are initialized to β, satisfying α + β = 1. Typically, α and β are initialized to 0.5, assuming that the predictions from the two models are of equal importance.

[0086] ② Calculate the fusion prediction result based on the initial weights: .

[0087] ③ Define a loss function to measure the error between the predicted result and the true value. The formula for calculating the loss function is: (5); in, For; loss function, These are the weighting coefficients. This represents the total number of samples, i.e., the number of data points used to calculate the loss. For sample index, representing the first... One sample; For the first The actual temperature value of each sample; For the first The first predicted value for each sample; For the first The second predicted value for each sample.

[0088] ④ To update the weights, it is necessary to calculate the gradient with respect to the loss function: (6); ⑤ Initialize the cumulative sum of squared gradients G0=0, initial weights α0, learning rate η, and a very small positive number ϵ. At the t-th iteration, the current weights and the cumulative sum of squared gradients are respectively: (7); ⑥ Repeat steps ① to ⑤ until the loss function converges or the number of prediction iterations is reached.

[0089] In the adaptive weight adjustment gradient descent process of the XGBoost-LSTM fusion algorithm, the hyperparameters are optimized using five-fold cross-validation and grid search. The optimization range of the hyperparameters is as follows: the learning rate (η) is set to [0.001, 0.01, 0.1], and the optimal value is finally determined to be 0.01 through verification. This value can balance the stability of weight updates and the convergence efficiency under a sample size of 960 groups; the optimization range of the number of iterations is [200, 400, 600]. After the convergence judgment of the loss function (mean square error decreases by <0.001 after 50 consecutive iterations), the optimal number of iterations is selected as 400; the initial weights α and β are kept at the default value of 0.5, and the minimum positive number ε is taken as 1×10-8. The regularization parameter λ of XGBoost (optimization range [0.01, 0.1, 1]), tree depth (max_depth, optimization range [3, 5, 7]), and timestep of LSTM (timestep, optimization range [6, 12, 18]) were optimized. The optimal combination was finally determined to be: λ=0.1, max_depth=5, timestep=12. This combination made the coefficient of determination R2 of the fusion algorithm reach 0.98 and the mean absolute error MAE reduced to 0.91.

[0090] Online learning and adaptive optimization mechanism: Rolling updates, collecting new operational data every quarter to incrementally train the neural network model; Anomaly detection and compensation, based on the principle of statistical process control (SPC), setting a temperature change threshold to detect sensor anomalies, and when an abnormal data point is detected, starting Kalman filtering for data fusion and state estimation.

[0091] Edge computing nodes: Industrial-grade edge servers are deployed on-site in the tunnel, configured with NVIDIA Jetson AGXOrin, and feature: real-time data acquisition and preprocessing capabilities (10ms response time); model inference capabilities (single-step prediction latency <50ms); edge caching and breakpoint resume functionality to ensure no data loss during network interruptions. Predictive control algorithm: A model predictive control framework is adopted, which uses neural networks to predict temperature changes over the next 12 to 24 hours. The control sequence is continuously optimized, and safety boundary constraints are set considering equipment start-up and shutdown costs and energy consumption to ensure that the lining temperature is always maintained within the safe range of -5℃ to 10℃. Through the above specific implementation methods, the anti-freezing and heat preservation design method for plateau tunnels proposed in this invention can be effectively implemented, realizing intelligent and precise control of anti-freezing and heat preservation in plateau tunnels, and ensuring the anti-freezing and heat preservation performance and structural safety of tunnels in harsh environments.

[0092] Through the above specific implementation methods, the anti-freezing and heat preservation design method for plateau tunnels proposed in this invention can be effectively implemented, realizing intelligent and precise control of anti-freezing and heat preservation in plateau tunnels, and ensuring the anti-freezing and heat preservation performance and structural safety of tunnels in harsh environments.

[0093] This application also relates to an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the high-altitude cold-region tunnel anti-freezing and heat preservation control method in the above embodiments.

[0094] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0095] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0096] Example 4: Application Case of Temperature Prediction Model for Tunnels in High-Altitude and Cold Regions (1) Analysis of prediction results After the code computation was completed, the evaluation metrics for the LSTM algorithm, XGBoost algorithm, and gradient descent-based XGBoost-LSTM fusion algorithm were output, as shown in Table 1. The coefficient of determination for each individual algorithm was close to 0.9, while the coefficient of determination for the fusion algorithm reached 0.97, indicating that the temperature predictions of the fusion algorithm fit the actual values ​​better. The mean absolute errors of the different algorithms were all small, with the fusion algorithm having the smallest mean absolute error, indicating that the average difference between the predicted and actual values ​​of the fusion algorithm was the smallest. The root mean square error of the two individual algorithms both exceeded 2, while that of the fusion algorithm was only 1.25, indicating that the average difference between the prediction results and the actual values ​​of the fusion algorithm was significantly optimized compared to the other two.

[0097] Table 1 Evaluation index values ​​of different algorithms ; The comparison results between the predicted values ​​of the temperature inside the tunnel in cold regions and the sample values ​​of different algorithms are as follows: Figure 5 As shown. Comparison Figure 5 As can be seen from (a), (b), and (c), the predicted values ​​of the cave temperature by each algorithm are relatively close to the sample values, while the prediction results of the fusion algorithm are more stable and almost coincide with the sample value line, showing better prediction performance.

[0098] Figure 5 (d) shows the error between the predicted values ​​and the sample values ​​for different algorithms. As can be seen from the figure, the prediction results of the XGBoost-LSTM fusion algorithm, the XGBoost algorithm, and the LSTM algorithm all showed varying degrees of error compared to the sample values. However, the XGBoost and LSTM algorithms exhibited higher dispersion than the fusion algorithm. Error analysis of the three algorithms revealed that the average errors of the fusion algorithm, the XGBoost algorithm, and the LSTM algorithm were 1.26, 2.47, and 2.63, respectively. Furthermore, Figure 5 (d) There are two points of maximum relative error. At the first point of maximum error, the errors of the three algorithms are 14.0, 22.5, and 25.7, respectively; at the second point of maximum error, the errors of the three algorithms are 14.1, 20.5, and 22.2, respectively. This further proves that the fusion algorithm has higher prediction accuracy.

[0099] (2) Engineering example verification Due to the significant geographical variations in cold-region tunnels, on-site monitoring is challenging, and temperature characteristics vary complexly. Therefore, this section verifies the reliability and accuracy of the predicted temperature inside cold-region tunnels. Verification and comparison are based on typical engineering measured data (monitoring of the Que'ershan Tunnel, which was not monitored for a full year, and the Yuxi Molegai Tunnel, which was monitored for a full year). Specific prediction results can be found in [link to relevant documentation]. Figure 6 .

[0100] Figure 6 (a) shows that the maximum temperature prediction errors at depths of 280m and 650m in the Queershan Tunnel are 4.1℃ and 2.3℃, respectively, and the average prediction errors are 1.79℃ and 1.25℃, respectively. Figure 6 (b) The maximum error of temperature prediction at a depth of 20m from the entrance of the Yuxi Molegai Tunnel in the first and second years was 2.2℃ and 12.8℃, respectively, and the average prediction error was 1.33℃ and 6.29℃, respectively.

[0101] Comprehensive analysis revealed that the average absolute error of the predicted temperatures for the two tunnels under various operating conditions was approximately 2.7℃, increasing to 4-6℃ when the tunnels were in extreme temperatures. This analysis indicates that the fusion algorithm's prediction of tunnel interior temperatures in cold regions has relatively small errors, and to a certain extent, it can meet the accuracy requirements for tunnel interior temperature analysis and tunnel design and construction.

[0102] Combination Figure 6 The prediction results show that the maximum error point always occurs at extreme temperatures (the lowest or highest temperature), and the error value exceeds the average error range. This is mainly because the prediction model is built based on ambient temperature data and does not consider the direct influence of other factors such as geothermal gradient and latent heat of water-ice phase change on the temperature inside the cave, resulting in a large error in the predicted temperature inside the cave at extreme temperatures. Furthermore, the impact of construction or other auxiliary facilities (ventilation, air curtains, etc.) is not considered. Figure 6 (b) Due to factors such as construction disturbances, the predicted temperature (for the same month in different years) of the Yuxi Molegai Tunnel varies greatly, resulting in generally large errors in the prediction results. At the same time, since the tunnel temperature in the original dataset is often monitored on a one-year basis, the prediction model is unable to reflect temperature changes when making predictions for consecutive years, resulting in errors that do not meet the accuracy requirements.

[0103] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for preventing freezing and maintaining heat in tunnels in high-altitude and cold regions, characterized in that, include: Temperature data at various monitoring points in the tunnel are collected in real time using temperature sensors. Temperature data is preprocessed to obtain effective temperature data; The effective temperature data is transmitted to the neural network controller; The neural network controller analyzes and processes the effective temperature data based on a pre-trained neural network model, predicts the temperature of the tunnel at the next moment, and outputs the final temperature prediction value. The neural network model includes a first algorithm module, a second algorithm module, and an output module. The first algorithm module outputs the first predicted value based on the XGBoost model; The second algorithm module outputs a second predicted value based on the LSTM model; The AdaGrad algorithm is used to adaptively adjust the weights of the first and second predicted values. The weights are iteratively updated by minimizing the loss function until the loss function converges or the preset number of iterations is reached. The output module calculates the final temperature prediction value based on the first and second predicted values ​​and their corresponding weights. Based on the final temperature prediction, the heating power and heating duration of the electric heating wire and electric heat tracing system are dynamically controlled to adjust the temperature inside the cave to remain within the set range at the next moment.

2. The method for preventing freezing and maintaining heat in tunnels in high-altitude cold regions according to claim 1, characterized in that, The loss function is: ； in, For loss function, These are the weighting coefficients. This represents the total number of samples, i.e., the number of data points used to calculate the loss. For sample index, representing the first... One sample; For the first The actual temperature value of each sample; For the first The first predicted value for each sample; For the first The second predicted value for each sample.

3. A high-altitude, cold-region tunnel anti-freezing and insulation system, characterized in that, It includes, from the inside out, a shotcrete layer, a first waterproof layer, a heat insulation buffer layer, an electric heating and temperature testing layer, a tunnel lining, a reinforced concrete arch, a second waterproof layer, a secondary lining, a second layer of geotextile, a composite insulation layer, and a plastering and finishing layer, all set on the tunnel rock wall. The electric heating and temperature testing layer includes an electric heating wire and multiple temperature sensors; It also includes a control box, which contains a neural network controller. The signal line of the temperature sensor is connected to the input port of the neural network controller, and the control circuit of the electric heating wire is connected to the output port of the neural network controller.

4. A method for frost protection and insulation construction of tunnels in high-altitude and cold regions, characterized in that: include: Shot concrete is sprayed onto the tunnel rock wall, and after the concrete hardens, a shot concrete layer is formed. The first layer of geotextile is fixed to the outside of the shotcrete layer using steel nails to form the first waterproof layer. A heat insulation buffer layer is set on the outside of the first waterproof layer. The heat insulation buffer layer is set on the shallow buried side of the tunnel and is evenly attached to the tunnel wall in the circumferential direction. An electric heating and temperature testing layer, including an electric heating wire and a temperature sensor, is installed on the outside of the heat insulation buffer layer. A control box is also installed inside the tunnel, which contains a neural network controller. The signal line of the temperature sensor is connected to the input port of the neural network controller, and the control circuit of the electric heating wire is connected to the output port of the neural network controller. The tunnel lining, reinforced concrete arch, second waterproof layer, and secondary lining structure are constructed sequentially. A second layer of geotextile is laid on the surface of the secondary lining, followed by a composite insulation layer. The composite insulation layer includes a nano-aerogel felt layer, a flame-retardant polyurethane insulation layer, and a waterproof and breathable membrane layer arranged from the inside out. The outermost layer is coated with a finishing layer and a finishing layer; A drainage and antifreeze system is installed on the outside of the tunnel lining.

5. The method for frost protection and insulation construction of tunnels in high-altitude cold regions according to claim 4, characterized in that: The heat insulation buffer layer is made of polystyrene foam board with a thickness of 5-8cm. It is bonded to the surface of the first waterproof layer with adhesive and fixed with expansion bolts and pressure plates.

6. The method for anti-freezing and heat preservation construction of tunnels in high-altitude cold regions according to claim 5, characterized in that: The electric heating wires are evenly arranged in a serpentine pattern on the outside of the heat insulation buffer layer, with a spacing of 30-50cm; temperature sensors are arranged at the tunnel arch, arch waist, arch shoulder, invert arch, and arch foot to form a temperature sensor network.

7. The method for frost protection and insulation construction of tunnels in high-altitude cold regions according to claim 6, characterized in that: The drainage and antifreeze system includes a ring-shaped drainage blind ditch, insulated drainage pipes, and an electric heat tracing system; A ring-shaped drainage blind ditch is set around the bottom of the outside of the tunnel lining, with a perforated HDPE drainage pipe buried inside. The drainage pipe is wrapped with geotextile. The drainage pipe is wrapped with rubber and plastic insulation board, and an automatic drainage valve is installed at the lowest point. The electric heat tracing system includes an electric heat tracing tape wrapped around the insulated drainage pipe, which is connected to a neural network controller.

8. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the antifreeze and heat preservation control method for tunnels in high-altitude cold regions as described in any one of claims 1-2.

Images