Soil temperature control method based on multi-algorithm fusion of alpine complex disturbance scene

Abstract

This invention provides a multi-algorithm fusion method for temperature control of vegetated soil in complex disturbance scenarios in high-altitude and cold regions, belonging to the field of intelligent temperature control technology. The method includes: S1: collecting multi-source environmental data and obtaining soil suitability parameters; S2: preprocessing the slope of the target area and deploying intelligent temperature control core components; S3: determining the benchmark value and dynamic adjustment range of the core temperature control index based on multi-source environmental data and soil suitability parameters; S4: constructing a fusion algorithm model and designing a deep interactive fusion mechanism; S5: training the fusion algorithm model, optimizing the algorithm parameters, and then adjusting the algorithm interaction parameters through overall testing to complete model optimization; S6: performing temperature control based on the optimized fusion algorithm model. This invention solves the problems of low temperature control accuracy, poor stability, and high energy consumption caused by multi-source disturbances, insufficient prediction, response lag, and parameter solidification in the process of temperature control of vegetated soil in high-altitude and cold regions.

Description

Technical Field

[0001] This invention relates to the field of intelligent temperature control (physical parameter control) technology, and in particular to a multi-algorithm fusion method for controlling the temperature of vegetated soil based on complex disturbance scenarios in high-altitude and cold regions. Background Technology

[0002] Currently, intelligent soil temperature control technology has become a key support for modern agriculture and ecological restoration, especially focusing on precise control methods and system design for physical parameters such as temperature. For example, CN120631085B (Invention title: A Smart Temperature Sensing Control System and Method for Plantation Soil Temperature) discloses a soil temperature control scheme for plantations. This scheme achieves risk analysis and resource optimization control of plantation soil temperature by dividing soil temperature monitoring sub-blocks, determining soil temperature stress factors, and constructing a resource area mapping matrix. It significantly improves the accuracy and efficiency of temperature control in seedling cultivation environments and provides a typical implementation path for soil temperature control technology.

[0003] However, the existing technical solutions still have significant limitations when dealing with the special and complex disturbance scenarios in high-altitude and cold regions: First, they lack the ability to accurately capture multi-source dynamic coupled disturbances such as freeze-thaw cycles, ultraviolet radiation, and gusts in high-altitude and cold regions. Traditional control algorithms mostly rely on feedback from single environmental parameters, making it difficult to adapt to the coupled effects of disturbances in extreme environments. Second, their temperature prediction capabilities are insufficient. Existing technologies mostly rely on real-time data for passive response regulation, failing to form a short-term accurate prediction mechanism that combines historical trends and dynamic disturbances, resulting in a lag in regulation response. Third, temperature control parameters are mostly semi-fixed configurations. Although they can achieve zoned regulation, they lack a parameter dynamic optimization mechanism that is deeply adapted to dynamic environmental changes and vegetation growth stages. In scenarios such as seasonal changes and freeze-thaw cycles in high-altitude and cold regions, they are prone to problems such as decreased temperature control accuracy and high energy consumption. Fourth, they have not formed a closed-loop collaborative system for the entire process of "disturbance observation - temperature prediction - parameter optimization - precise control." The linkage between control algorithms and actuators is insufficient, making it difficult to balance temperature control stability and energy economy under complex disturbances.

[0004] The unique climatic conditions of high-altitude and cold regions make vegetated soil temperatures susceptible to multi-source disturbances such as freeze-thaw cycles, ultraviolet radiation, and gusty winds, resulting in drastic temperature fluctuations that are difficult to stabilize. Traditional vegetated soil temperature control technologies often employ single heating or passive insulation methods, lacking the ability to accurately capture complex disturbances and predict temperature change trends in advance. Furthermore, temperature control parameters are mostly fixed settings, making dynamic adjustment based on the environment difficult, leading to delayed response and insufficient accuracy. In addition, traditional technologies lack a closed-loop system encompassing disturbance observation, temperature prediction, parameter optimization, and precise control, making it difficult to balance energy consumption and control effectiveness during the temperature control process. Moreover, the lack of synergistic adaptation with the vegetated soil structure further affects the stability of the temperature environment required for vegetation growth, thus hindering the efficiency and effectiveness of ecological restoration in high-altitude and cold regions. Summary of the Invention

[0005] This invention provides a multi-algorithm fusion method for vegetated soil temperature control in complex disturbance scenarios in high-altitude and cold regions. The aim is to address the problems of low temperature control accuracy, poor stability, and high energy consumption caused by multi-source disturbances, insufficient prediction, response lag, and parameter rigidity in vegetated soil temperature control in these regions. By constructing an intelligent temperature control system through algorithm fusion, it achieves precise, rapid, and stable regulation of vegetated soil temperature under complex environments, balancing temperature control effectiveness and energy consumption. This enables accurate capture, short-term prediction, dynamic optimization, and closed-loop control of soil temperature in complex disturbance scenarios in high-altitude and cold regions, improving the robustness, response speed, and energy economy of physical parameter control, and enhancing the application capability of intelligent soil temperature regulation technology in extreme environments.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A multi-algorithm fusion method for vegetation soil temperature control based on complex disturbance scenarios in high-altitude and cold regions includes:

[0008] S1: Collect multi-source environmental data of the target area in the high-altitude cold region through a distributed sensor array, while collecting historical meteorological data and conducting soil sample testing to obtain soil suitability parameters;

[0009] S2: After pre-treating the slope of the target area, deploy intelligent temperature control core components, which include a photothermal conversion unit, a phase change energy storage unit, a distributed sensor array, an adaptive irrigation system, and a geothermal compensation device.

[0010] S3: Based on multi-source environmental data and soil suitability parameters, determine the benchmark values ​​and dynamic adjustment ranges of core temperature control indicators;

[0011] S4: Construct a fusion algorithm model that includes a fractional-order cooperative perturbation observer, Gaussian process regression-fuzzy cognitive graph, hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, and improved Lyapunov adaptive sliding mode control algorithm, and design a deep interactive fusion mechanism for the algorithm;

[0012] S5: Train the fusion algorithm model, optimize the algorithm parameters, and then adjust the algorithm interaction parameters through overall testing to complete the model optimization;

[0013] S6: The distributed sensor array is activated to continuously collect real-time data. Based on the core temperature control index benchmark value and dynamic adjustment range, the current target temperature is determined. Based on the optimized fusion algorithm model, multi-source dynamic disturbances are captured by a fractional-order collaborative disturbance observer. Short-term temperature prediction is performed using Gaussian process regression-fuzzy cognitive map. With the current target temperature, temperature prediction value, and disturbance estimate value as input, the optimal heat release rate of the phase change energy storage unit and the optimal output power of the geothermal compensation device are solved by the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm. Based on the improved Lyapunov adaptive sliding mode control algorithm, control commands are generated to drive the intelligent temperature control core components to perform temperature control operations: the photothermal conversion unit absorbs ultraviolet rays and converts them into infrared thermal radiation to provide energy for the phase change energy storage unit to store heat; the phase change energy storage unit adjusts the heat release intensity through temperature control valves according to the optimal heat release rate to achieve heat storage during the day and heat release at night; the geothermal compensation device controls the power supply status of the carbon fiber heating wire through pulse width modulation signals according to the optimal output power, starting or stopping heating, and adjusting the heating power at the same time.

[0014] This specification also includes a multi-algorithm fusion method for vegetated soil temperature control based on complex disturbance scenarios in high-altitude and cold regions, which further includes:

[0015] S7: The system continuously collects soil moisture data through a distributed sensor array, compares the soil moisture content with the irrigation start-up moisture content benchmark value and the dynamic adjustment range. If the soil moisture content deviates from the suitable range, the system will start the adaptive irrigation system to replenish water. During the water replenishment process, the soil temperature changes are monitored simultaneously. The moisture content change information is input into the fusion algorithm model. The model combines the core temperature control index benchmark value and the dynamic adjustment range to adjust the optimal temperature control parameters, driving the phase change energy storage unit and geothermal compensation device to adapt and adjust the heat release rate and heating power, so as to achieve temperature and humidity coordinated adaptation.

[0016] In this specification, the fractional-order cooperative disturbance observer described in S4 is constructed using fractional-order calculus theory. Utilizing the memory characteristics of fractional-order differential operators, it traces the historical trend of disturbance changes, achieving comprehensive capture of multi-source dynamic coupled disturbances such as freeze-thaw cycles, ultraviolet radiation, and gusts. The output disturbance estimate provides data support for the search range of the constraint parameters of the hyperbolic tangent-Cauchy mutant particle swarm optimization algorithm and the adjustment of the control gain by the improved Lyapunov adaptive sliding mode control algorithm.

[0017] In this specification, the Gaussian process regression-fuzzy cognitive graph described in S4 integrates the probabilistic prediction characteristics of Gaussian process regression with the causal reasoning ability of fuzzy cognitive graph to construct a node system containing key factors such as real-time temperature, number of freeze-thaw cycles, ultraviolet intensity, and disturbance estimates. By dynamically adjusting the strength of causal relationships between nodes, it achieves accurate short-term prediction of soil temperature, providing a forward-looking adaptation basis for the current target temperature determined based on the core temperature control index benchmark value and dynamic adjustment range, and assisting in optimizing the control timing of phase change energy storage units and geothermal compensation devices.

[0018] In this specification, the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm described in S4 dynamically adjusts the inertia weight through the hyperbolic tangent function to balance the global search and local search capabilities of the algorithm. At the same time, the Cauchy mutation mechanism is introduced to perturb the particle position to avoid the algorithm getting trapped in local optima. With the deviation between the current target temperature and the real-time temperature and the system energy consumption as optimization objectives, the algorithm solves for the optimal heat release rate of the phase change energy storage unit and the optimal output power of the geothermal compensation device to meet the core temperature control requirements.

[0019] In this specification, the improved Lyapunov adaptive sliding mode control algorithm described in S4 designs a nonlinear sliding surface, introduces an adaptive law to dynamically adjust the control gain, offsets the effects of disturbance observation errors and model uncertainties, incorporates temperature prediction values ​​as feedforward compensation terms, and uses a saturation function to suppress chattering, ensuring that the generated control commands can accurately drive the temperature control valve of the phase change energy storage unit and the carbon fiber heating wire of the geothermal compensation device, stabilizing the soil temperature within the target range specified by the core temperature control index.

[0020] In this specification, the algorithm deep interaction and fusion mechanism described in S4 is specifically as follows: the perturbation estimate of the fractional-order cooperative perturbation observer is input into the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm and the improved Lyapunov adaptive sliding mode control algorithm, respectively, to constrain the parameter search range and adjust the control gain; the temperature prediction value of Gaussian process regression-fuzzy cognitive map guides the direction of parameter optimization and serves as feedforward compensation; the optimal heat release rate and optimal output power output by the optimization algorithm are fed back to correct the node relationship strength of the prediction model, and at the same time serve as the initial parameters of the control algorithm to shorten the convergence time; the control error feedback of the control algorithm adjusts the observer gain coefficient, forming a closed-loop cooperative system that adapts to the core temperature control index requirements.

[0021] In this manual, the training and optimization of the fusion algorithm model in S5 includes: specifically optimizing the gain coefficient of the fractional-order cooperative perturbation observer, the kernel function parameters and node weights of the Gaussian process regression-fuzzy cognitive graph, the inertial weights and mutation parameters of the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, and the sliding surface parameters and adaptive gain of the improved Lyapunov adaptive sliding mode control algorithm. The optimal parameters of each algorithm are determined through stability analysis and error verification. Then, based on the comprehensive test indicators such as temperature control accuracy, response speed, and robustness required by the core temperature control indicators, the algorithm interaction parameters are adjusted to ensure that the fusion model can accurately drive the core components of intelligent temperature control to adapt to the complex environment of high-altitude and cold regions.

[0022] In this specification, closed-loop feedback control is performed in S6, specifically by comparing the temperature feedback data after temperature control execution with the current target temperature specified by the core temperature control index to obtain the control error. Based on this control error, the gain coefficient of the fractional-order cooperative disturbance observer, the inertial weight of the hyperbolic tangent-Cauchy mutant particle swarm optimization algorithm, and the control gain of the improved Lyapunov adaptive sliding mode control algorithm are dynamically adjusted to optimize the heat release rate of the next phase change energy storage unit and the output power of the geothermal compensation device, ensuring that the temperature remains stable within the dynamic adjustment range of the core temperature control index.

[0023] In this specification, the temperature and humidity coordinated adaptation described in S7 specifically refers to the following: During the water replenishment process of the adaptive irrigation system, the distributed sensor array collects soil temperature data at fixed intervals, calculates the temperature change rate, and inputs it into the fusion algorithm model. The model combines the target temperature range of the core temperature control index and the moisture content change trend to adjust the opening of the temperature control valve of the phase change energy storage unit and the power supply duty cycle of the carbon fiber heating wire of the geothermal compensation device in real time, and adjusts the heat release rate and heating power accordingly to avoid the temperature deviating from the suitable range specified by the core temperature control index due to water replenishment.

[0024] In summary, the present invention has at least the following beneficial effects:

[0025] Temperature control accuracy and stability are significantly improved: Through deep integration of multiple algorithms, the system can accurately identify complex disturbances and make accurate short-term temperature predictions. Temperature control decisions are more forward-looking, effectively reducing temperature fluctuations and keeping the soil temperature stable within the suitable range for vegetation growth.

[0026] Enhanced disturbance resistance: The fractional-order cooperative disturbance observer can comprehensively capture multi-source dynamic disturbances such as freeze-thaw cycles and gusts. Combined with the robust design of the improved sliding mode control algorithm, the temperature control system can still operate stably under extreme environments and complex disturbances, unaffected by external interference.

[0027] Optimized response speed: With the help of temperature prediction and feedforward compensation mechanisms, the temperature control system can start control measures in advance to avoid the temperature deviating from the target range, significantly shorten the temperature control response time, and realize active control instead of passive response.

[0028] Energy consumption and performance balance: By using the particle swarm optimization algorithm to find the global optimal solution for temperature control parameters, the system energy consumption is reduced to the minimum while ensuring temperature control performance, thereby improving the economy and sustainability of the temperature control process.

[0029] Enhanced Adaptability: Relying on incremental training and dynamic parameter adjustment mechanisms, the temperature control system can adapt to environmental dynamics such as seasonal changes and vegetation growth stages in high-altitude and cold regions, continuously optimizing the temperature control strategy without frequent human intervention. Attached Figure Description

[0030] Figure 1 This is a schematic diagram illustrating the steps of the multi-algorithm fusion method for controlling soil temperature in high-altitude cold regions based on complex disturbance scenarios.

[0031] Figure 2 This is a flowchart illustrating the multi-algorithm fusion method for controlling soil temperature in vegetation based on complex disturbance scenarios in high-altitude and cold regions, which is involved in this invention.

[0032] Figure 3 This is a schematic diagram of the multi-algorithm fusion and interaction logic involved in this invention.

[0033] Figure 4 This is a schematic diagram of the multi-algorithm fusion temperature control execution process involved in this invention.

[0034] Figure 5 This is a schematic diagram of the vegetation mat structure involved in this invention (from top to bottom: photothermal conversion layer, microenvironment regulation layer, gradient matrix layer, anchoring layer; the gradient matrix layer includes a seed germination promotion layer, a root growth support layer, and a frozen soil interface improvement layer).

[0035] Figure 6 This is a schematic diagram of the sensor layout and the arrangement of the device on the slope involved in this invention.

[0036] Figure 7 This is a schematic diagram of the photothermal conversion energy storage involved in this invention.

[0037] Figure 8 This is a schematic diagram of data collection and uploading involved in this invention.

[0038] Figure 9 This is a schematic diagram showing the heat release performance data of different graphite addition amounts and heat storage device diameters involved in this invention. Detailed Implementation

[0039] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0040] like Figure 1 As shown, this embodiment provides a multi-algorithm fusion method for vegetated soil temperature control based on complex disturbance scenarios in high-altitude and cold regions, including:

[0041] S1: Collect multi-source environmental data of the target area in the high-altitude cold region through a distributed sensor array, while collecting historical meteorological data and conducting soil sample testing to obtain soil suitability parameters;

[0042] S2: After pre-treating the slope of the target area, deploy intelligent temperature control core components, which include a photothermal conversion unit, a phase change energy storage unit, a distributed sensor array, an adaptive irrigation system, and a geothermal compensation device.

[0043] S3: Based on multi-source environmental data and soil suitability parameters, determine the benchmark values ​​and dynamic adjustment ranges of core temperature control indicators;

[0044] S4: Construct a fusion algorithm model that includes a fractional-order cooperative perturbation observer, Gaussian process regression-fuzzy cognitive graph, hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, and improved Lyapunov adaptive sliding mode control algorithm, and design a deep interactive fusion mechanism for the algorithm;

[0045] S5: Train the fusion algorithm model, optimize the algorithm parameters, and then adjust the algorithm interaction parameters through overall testing to complete the model optimization;

[0046] S6: The distributed sensor array is activated to continuously collect real-time data. Based on the core temperature control index benchmark value and dynamic adjustment range, the current target temperature is determined. Based on the optimized fusion algorithm model, multi-source dynamic disturbances are captured by a fractional-order collaborative disturbance observer. Short-term temperature prediction is performed using Gaussian process regression-fuzzy cognitive map. With the current target temperature, temperature prediction value, and disturbance estimate value as input, the optimal heat release rate of the phase change energy storage unit and the optimal output power of the geothermal compensation device are solved by the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm. Based on the improved Lyapunov adaptive sliding mode control algorithm, control commands are generated to drive the intelligent temperature control core components to perform temperature control operations: the photothermal conversion unit absorbs ultraviolet rays and converts them into infrared thermal radiation to provide energy for the phase change energy storage unit to store heat; the phase change energy storage unit adjusts the heat release intensity through temperature control valves according to the optimal heat release rate to achieve heat storage during the day and heat release at night; the geothermal compensation device controls the power supply status of the carbon fiber heating wire through pulse width modulation signals according to the optimal output power, starting or stopping heating, and adjusting the heating power at the same time.

[0047] In some embodiments, the multi-algorithm fusion method for controlling vegetated soil temperature based on complex disturbance scenarios in high-altitude and cold regions further includes:

[0048] S7: The system continuously collects soil moisture data through a distributed sensor array, compares the soil moisture content with the irrigation start-up moisture content benchmark value and the dynamic adjustment range. If the soil moisture content deviates from the suitable range, the system will start the adaptive irrigation system to replenish water. During the water replenishment process, the soil temperature changes are monitored simultaneously. The moisture content change information is input into the fusion algorithm model. The model combines the core temperature control index benchmark value and the dynamic adjustment range to adjust the optimal temperature control parameters, driving the phase change energy storage unit and geothermal compensation device to adapt and adjust the heat release rate and heating power, so as to achieve temperature and humidity coordinated adaptation.

[0049] In some embodiments, the fractional-order cooperative disturbance observer described in S4 is constructed using fractional-order calculus theory. By utilizing the memory characteristics of fractional-order differential operators, it traces the historical trend of disturbance changes, achieving comprehensive capture of multi-source dynamic coupled disturbances such as freeze-thaw cycles, ultraviolet radiation, and gusts. The output disturbance estimate provides data support for the search range of the constraint parameters of the hyperbolic tangent-Cauchy mutant particle swarm optimization algorithm and the adjustment of the control gain by the improved Lyapunov adaptive sliding mode control algorithm.

[0050] In some embodiments, the Gaussian process regression-fuzzy cognitive graph described in S4 integrates the probabilistic prediction characteristics of Gaussian process regression with the causal reasoning ability of fuzzy cognitive graph to construct a node system containing key factors such as real-time temperature, number of freeze-thaw cycles, ultraviolet intensity, and disturbance estimates. By dynamically adjusting the strength of causal relationships between nodes, it achieves accurate short-term prediction of soil temperature, providing a forward-looking adaptation basis for the current target temperature determined based on the core temperature control index benchmark value and dynamic adjustment range, and assisting in optimizing the control timing of phase change energy storage units and geothermal compensation devices.

[0051] In some embodiments, the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm described in S4 dynamically adjusts the inertia weight through the hyperbolic tangent function to balance the global search and local search capabilities of the algorithm. At the same time, it introduces the Cauchy mutation mechanism to perturb the particle position, avoiding the algorithm from getting trapped in local optima. With the deviation between the current target temperature and the real-time temperature and the system energy consumption as optimization objectives, it solves the optimal heat release rate of the phase change energy storage unit and the optimal output power of the geothermal compensation device to meet the core temperature control index requirements.

[0052] In some embodiments, the improved Lyapunov adaptive sliding mode control algorithm described in S4 designs a nonlinear sliding surface, introduces an adaptive law to dynamically adjust the control gain, offsets the effects of disturbance observation errors and model uncertainties, incorporates temperature prediction values ​​as feedforward compensation terms, and uses a saturation function to suppress chattering, ensuring that the generated control commands can accurately drive the temperature control valve of the phase change energy storage unit and the carbon fiber heating wire of the geothermal compensation device, stabilizing the soil temperature within the target range specified by the core temperature control index.

[0053] In some embodiments, the deep interactive fusion mechanism of the algorithm described in S4 is specifically as follows: the perturbation estimate of the fractional-order cooperative perturbation observer is input into the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm and the improved Lyapunov adaptive sliding mode control algorithm, respectively, to constrain the parameter search range and adjust the control gain; the temperature prediction value of Gaussian process regression-fuzzy cognitive map guides the direction of parameter optimization and serves as feedforward compensation; the optimal heat release rate and optimal output power output by the optimization algorithm are fed back to correct the node relationship strength of the prediction model, and at the same time serve as the initial parameters of the control algorithm to shorten the convergence time; the control error feedback of the control algorithm adjusts the observer gain coefficient, forming a closed-loop cooperative system that adapts to the core temperature control index requirements.

[0054] In some embodiments, the training and optimization of the fusion algorithm model in S5 includes: performing specific optimizations on the gain coefficient of the fractional-order cooperative perturbation observer, the kernel function parameters and node weights of the Gaussian process regression-fuzzy cognitive graph, the inertial weights and mutation parameters of the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, and the sliding surface parameters and adaptive gain of the improved Lyapunov adaptive sliding mode control algorithm. The optimal parameters of each algorithm are determined through stability analysis and error verification. Then, the algorithm interaction parameters are adjusted based on the comprehensive test indicators such as temperature control accuracy, response speed, and robustness required by the core temperature control indicators to ensure that the fusion model can accurately drive the core components of intelligent temperature control to adapt to the complex environment of high-altitude and cold regions.

[0055] In some embodiments, closed-loop feedback control is performed in S6, specifically by comparing the temperature feedback data after temperature control execution with the current target temperature specified by the core temperature control index to obtain the control error. Based on this control error, the gain coefficient of the fractional-order cooperative disturbance observer, the inertial weight of the hyperbolic tangent-Cauchy mutant particle swarm optimization algorithm, and the control gain of the improved Lyapunov adaptive sliding mode control algorithm are dynamically adjusted to optimize the heat release rate of the next phase change energy storage unit and the output power of the geothermal compensation device, ensuring that the temperature remains stable within the dynamic adjustment range of the core temperature control index.

[0056] In some embodiments, the temperature and humidity coordinated adaptation in S7 specifically refers to: during the water replenishment process of the adaptive irrigation system, the distributed sensor array collects soil temperature data at fixed intervals, calculates the temperature change rate and inputs it into the fusion algorithm model. The model combines the target temperature range of the core temperature control index and the moisture content change trend to adjust the opening of the temperature control valve of the phase change energy storage unit and the power supply duty cycle of the carbon fiber heating wire of the geothermal compensation device in real time, and adjusts the heat release rate and heating power accordingly to avoid the temperature deviating from the suitable range specified by the core temperature control index due to water replenishment.

[0057] The technical concept of this invention is as follows:

[0058] This solution uses intelligent temperature control as its core, constructing a complete technical system encompassing data acquisition, algorithm fusion, closed-loop regulation, and maintenance optimization. For example... Figure 1 and Figure 2 As shown, firstly, multi-source environmental data such as soil temperature, freeze-thaw cycle count, ultraviolet intensity, and wind speed are collected through a distributed sensor array. Then, a temperature control decision model is constructed, deeply integrating a fractional-order collaborative perturbation observer, a Gaussian process regression-fuzzy cognitive graph, a hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, and an improved Lyapunov adaptive sliding mode control algorithm. This achieves accurate perturbation observation, short-term temperature prediction, global optimization of temperature control parameters, and precise control. A closed-loop feedback mechanism dynamically adjusts algorithm parameters and temperature control commands, driving the phase change energy storage unit, geothermal compensation device, and other actuators to work collaboratively. Simultaneously, a modular composite structure can be used to enhance the synergistic effect of temperature control and structure. Through regular maintenance and model iteration optimization, it continuously adapts to environmental changes, providing a stable and suitable temperature environment for vegetation soils in high-altitude and cold regions.

[0059] S1: Preliminary Preparations and Data Collection

[0060] A comprehensive field survey of the target area in the high-altitude and cold region was conducted, combined with the collection of historical meteorological data, to construct a complete environmental baseline dataset. During the field survey, high-precision soil thermometers were used to measure soil temperature in three typical areas of the target area: uphill, middle slope, and downhill. Five monitoring points were set up in each area, and continuous monitoring was conducted for 72 hours, with data recorded every hour. A portable anemometer (model TESTO410-2) was used to simultaneously measure instantaneous wind speed in the same area, with data recorded every 30 minutes. Soil samples from the 0-50cm soil layer were collected using a soil sampler, with three samples collected from each area for subsequent testing. Atmospheric pressure data were measured using a digital barometer (model GM100), with measurements taken twice daily at fixed times (10:00 AM and 4:00 PM). Slope gradient was measured using a slope meter (model NK-2), with three measurements taken in each area. Slope flatness was measured using a 2-meter straightedge and feeler gauge, with five measurements taken in each area. Historical meteorological data were obtained from provincial meteorological stations and ecological environment monitoring departments in the target area, covering 12 key data points for the past 5 years, including annual average temperature, extreme minimum temperature, extreme maximum temperature, annual freeze-thaw cycle count, active layer thickness, ultraviolet radiation intensity, annual precipitation, sunshine duration, initial soil temperature, initial soil moisture content, soil bulk density, and soil pH. All parameters were measured in triplicate in different regions, and the arithmetic mean was taken as the official data to ensure the representativeness and accuracy of the data.

[0061] The factory prefabrication process for the core components of the intelligent temperature control system was initiated to ensure that the component performance is adapted to the cold environment. The photothermal conversion unit uses a PLA film doped with 5% nano-silicon carbide, prepared by a casting method, and then edge-sealed with PVC tape to prevent brittleness at low temperatures. The phase change energy storage unit uses a paraffin / expanded graphite composite phase change material as its core, prepared by mixing them at a mass ratio of 95:5 (95% paraffin, 5% expanded graphite). This ratio has been experimentally verified to balance heat storage capacity and thermal conductivity. The composite phase change material is encapsulated in a PVC film bag using a sheet-like modular design. Each encapsulated unit is 50cm × 50cm and 2cm thick, optimizing heat release efficiency by increasing the heat dissipation area. The heat release performance of the paraffin / expanded graphite composite phase change energy storage unit was verified based on experimental data (refer to Table 1 and...). Figure 9 The phase change temperature of this composite material is slightly lower than that of pure paraffin (with no significant functional impact), and the latent heat of phase change is reduced by 10.14% compared to pure paraffin, but the thermal conductivity is increased by 214.30%, effectively solving the problems of poor thermal conductivity and slow heat release rate of pure paraffin. At -20℃, the latent heat of phase change decay rate is ≤3%, which can stably maintain the thermal storage performance and avoid excessive reliance on geothermal compensation devices at extreme low temperatures. The phase change energy storage unit with a sheet-like encapsulation design underwent equivalent testing (referring to the linear relationship between the diameter of the thermal storage unit and the heat release time, converting the equivalent diameter of the sheet-like structure's heat dissipation area): After simulating solar radiation thermal storage for 4 hours during the day, it can sustain heat release for ≥8 hours at 25℃ at night, with a stable heat release power of 15-20W / m². 2 It fully meets the nighttime soil insulation needs of vegetation in high-altitude and cold regions (target temperature 5-15℃); at extreme low temperatures (-15℃), the heat release time is shortened to 6 hours, at which point it works in conjunction with a geothermal compensation device (geothermal compensation power supplement 5-10W / m²). 2 It can achieve 24-hour continuous temperature control to avoid excessive energy consumption. After 50 freeze-thaw cycles (simulating the four seasons in a cold region), the latent heat of phase change of this composite phase change material retains ≥97%, and the encapsulation film is undamaged and leak-free, proving its long-term stability in the complex freeze-thaw environment of a cold region and solving the problem of latent heat decay at low temperatures.

[0062] Table 1. Heat release performance data for different graphite addition amounts and thermal storage device diameters

[0063] ;

[0064] The phase change energy storage unit of this invention adopts a sheet-like design (equivalent diameter 35mm) and uses a formula of 95% paraffin wax + 5% expanded graphite. Its heat release time is shortened by 39.5% compared with pure paraffin wax, and the thermal conductivity is significantly improved. It can meet the temperature control requirements of daytime heat storage and nighttime heat release in high-altitude and cold regions. Through structural optimization in low-temperature environments, the heat release time is extended to more than 8 hours, eliminating the need for excessive reliance on geothermal compensation devices and reducing system energy consumption.

[0065] The distributed sensor array uses a high-precision soil temperature and humidity sensor (model SHT30) with a measurement accuracy of ±0.5℃. The number of sensors is calculated based on the construction area, with one sensor deployed for every 10 square meters to ensure comprehensive monitoring coverage. The adaptive irrigation system is equipped with atomizing sprinklers (model SP-01) and a 0.5kW booster pump (model 12V-50W), with a sprinkler rated flow rate of 0.5L / m³. 2 •min; The carbon fiber heating wire of the geothermal compensation device has a diameter of 0.5mm and is connected in parallel. The insulation material of the wire is polytetrafluoroethylene to ensure that the damage of a single wire will not affect the overall operation. After all components are prefabricated, a low-temperature environment adaptability test is carried out. The components are placed in a -20℃ environment and left to stand for 48 hours. Only after checking for no damage and normal function can they be shipped from the factory.

[0066] Specialized laboratory tests were conducted on the collected soil samples to obtain a set of soil suitability parameters. Soil samples were divided into multiple groups and placed in constant temperature incubators with five temperature gradients (5℃, 8℃, 10℃, 12℃, and 15℃) to conduct germination experiments with common vegetation seeds from high-altitude and cold regions (Leymus chinensis, Stipa purpurea, Leymus chinensis, Agropyron cristatum, and Poa annua). The temperature range at which the seed germination rate was highest was recorded to determine the target temperature range for the substrate layer. A soil tensiometer (model TDR-300) was used to measure the water retention capacity of the soil at five moisture content gradients (10%, 15%, 20%, 22%, and 25%). Based on the needs of seed germination and root growth, the target moisture content range was determined. The EC value of the soil samples was directly measured using a portable EC meter (model HI98311). Soil culture experiments were conducted with five EC value gradients (0.8 mS / cm, 1.0 mS / cm, 1.5 mS / cm, 2.0 mS / cm, and 2.2 mS / cm) to observe vegetation growth and ultimately determine the suitable soil EC value standard for vegetation growth. All laboratory tests were conducted in three parallel sets, and the optimal range of the test results was taken as the soil suitability parameter.

[0067] During data acquisition, multiple types of sensors are used simultaneously to monitor parameters across all dimensions, such as... Figure 8 As shown, soil temperature, soil moisture, soil EC value, and soil pH value are collected by soil temperature sensor, soil moisture sensor, portable EC meter, and soil pH meter, respectively. After preliminary processing, the collected data is simultaneously uploaded to a local database (model MySQL8.0) and a cloud server (Alibaba Cloud ECS server). Dual encryption technology is used for backup to ensure secure data storage and convenient access.

[0068] S2: Slope Pretreatment and Component Deployment

[0069] Standardized pretreatment of the target area slope was implemented to lay the foundation for component installation. First, manual labor combined with a small excavator (model PC18MR-3) was used to clear debris such as gravel, weeds, deadwood, humus clumps, and exposed roots from the slope. For areas with a slope greater than 30°, slope trimming was performed to ensure the final slope was controlled within 30°. Next, a grader (model GR165) was used to level the slope. During leveling, the slope was monitored in real time with a slope meter, and the flatness was checked with a 2-meter straightedge to ensure the slope undulation did not exceed 5cm. Finally, a high-pressure sprayer (model 3WBD-16) was used to spray liquid mulch film (acrylate emulsion). During spraying, the nozzle was kept 50cm away from the slope, moved at a uniform speed, and the spray rate was controlled at 0.5kg / m². 2 This forms a uniform impermeable layer. After the impermeable layer is laid, 10 1m sections are randomly selected. 2 For each sample, the film thickness was measured using a film thickness measuring instrument (model CH-101) to ensure that the film thickness reached more than 0.1 mm and that there were no missed sprays or damage.

[0070] Intelligent temperature control components are deployed according to scientific layout principles to ensure efficient system operation. Combined with... Figure 6 The distributed sensor array is arranged in a quincunx pattern with a 5-meter spacing between sensors. The sensors are vertically embedded into the soil layer at a depth of 0-50cm, and the surrounding gaps are compacted with soil after implantation to prevent insufficient contact between the sensors and the soil. The adaptive irrigation system's atomizing nozzles are installed at 3-meter intervals, with the nozzles 1.5 meters above the slope, spraying in a fan shape to ensure all areas are covered by water mist, eliminating irrigation blind spots. The water pipes are made of PE material and secured to anchoring nails with cable ties. The carbon fiber heating wires of the geothermal compensation device are laid parallel to the slope with a 20cm spacing. After laying, they are fixed to anchoring nails to prevent displacement during freeze-thaw cycles. The wire connections are sealed with waterproof insulating tape. The photothermal conversion unit and phase change energy storage unit are laid sequentially from top to bottom, covering the entire slope with aligned edges and no gaps. The units are fixed together using snap-fit ​​connections.

[0071] In some embodiments, components may be deployed using a modular composite structure, such as Figure 5As shown, the composite structure consists of a photothermal conversion layer, a microenvironment control layer, a gradient matrix layer, and an anchoring layer from top to bottom. Each layer synergistically enhances temperature control and structural stability. The photothermal conversion layer, with a thickness of 0.5 mm, is made of PLA film doped with 5% nano-silicon carbide and prepared by a casting method: polylactic acid particles are vacuum dried at 80°C for 4 hours, dissolved in chloroform to prepare an 8% (w / v) solution, and 5% nano-silicon carbide powder and 0.5% KH-550 silane coupling agent (based on PLA dry weight) are added. The mixture is ultrasonically dispersed for 1 hour to form a casting solution, cast onto a glass plate to evaporate the solvent at 40°C for 12 hours, and then thoroughly dried in a vacuum oven at 60°C. The microenvironment control layer is formed by hot pressing phase change microcapsules and a polymer binder. The phase change microcapsules account for 70%~85% of the mass, and the polymer binder accounts for 30%~15%. The phase change microcapsules use a paraffin / expanded graphite composite as the core material. The wall material is melamine-formaldehyde resin (9:1 mass ratio), with paraffin phase transition temperature of 5-10℃ and expanded graphite particle size of 80-120 mesh; the gradient matrix layer has a total thickness of 30cm, the upper layer is a 0-5cm seed germination promoting layer (humus: biochar: water retention agent = 2:1:0.3), the middle layer is a 5-15cm root growth support layer (vermiculite: desulfurized gypsum: slow-release fertilizer = 3:1:0.5), and the lower layer is a 15-30cm frozen soil interface improvement layer (imported soil: frozen soil conditioner = 95:5, in the frozen soil conditioner polyacrylamide: ferrous sulfate = 1:1); the anchoring layer is a three-dimensional coconut fiber mesh with a length of 40cm and a pull-out force ≥5kN basalt fiber nails, combined with shape memory alloy (Ni-Ti) anchor rods.

[0072] After all components (including composite structure components) are installed, connect them to the central control system (PLC-S7-200), ensuring secure wiring connections and intact insulation. Initiate a comprehensive component self-test procedure to ensure the system's initial state is stable. During the self-test, first test the data transmission performance of the distributed sensor array, checking that the data transmission delay of each sensor does not exceed 1 second and that the data accuracy reaches over 99%. Second, test the adaptive irrigation system, observing the sprinkler atomization effect after starting the water pump, checking that the system response time does not exceed 2 seconds and that the spray flow rate is stable at 0.5 L / m³. 2 • min; then test the geothermal compensation device. After powering on, use an infrared thermometer (model FLIR TG165) to detect the surface temperature of the heating wire to ensure that the power density reaches 50W / m. 2 Finally, check the installation firmness of the photothermal conversion unit, phase change energy storage unit, and other layers of the composite structure. Gently pull the edge of the unit by hand; there should be no looseness. After the self-inspection is completed, record the results of each test item to form a detailed component deployment self-inspection report. Replace or adjust any problematic components in a timely manner.

[0073] S3: Temperature control parameter threshold calibration

[0074] Based on the previously collected environmental baseline dataset and soil suitability parameter set, the benchmark values ​​of core temperature control indicators were accurately determined. Combining laboratory soil suitability test results and referencing the growth characteristics of alpine vegetation such as *Leymus chinensis*, *Stipa purpurea*, *Leymus chinensis*, *Agropyron cristatum*, and *Poa annua*, the target temperature range for the substrate layer was determined to be 5℃ to 15℃. Based on the extreme minimum temperatures in the environmental baseline dataset (typically -30℃ to -20℃), and considering the heating efficiency and energy balance of the geothermal compensation device, the start-up temperature threshold for the geothermal compensation device was determined to be -15℃. According to soil water retention capacity test results and the water requirements for seed germination, the start-up moisture content threshold for the irrigation system was determined to be 15%, and the target moisture content was determined to be 22%. Based on soil EC value test results, the soil EC value control threshold was determined to be 1.0 mS / cm - 2.0 mS / cm. Referring to soil pH value test results, the soil pH value control threshold was determined to be 6.5 - 8.5.

[0075] Considering the distinct four seasons characteristic of high-altitude cold regions, dynamic adjustment ranges for parameter thresholds are set to ensure that the thresholds can adapt to environmental changes in real time. In winter (November to February of the following year), temperatures are extremely low with frequent freeze-thaw cycles. Therefore, the dynamic adjustment range for the target temperature range of the substrate layer is set to ±2℃, the adjustment range for the geothermal compensation device's activation temperature threshold is set to ±1℃, the adjustment range for the irrigation system's activation moisture content threshold is set to ±1%, the adjustment range for the soil EC value control threshold is set to ±0.2mS / cm, and the adjustment range for the soil pH value control threshold is set to ±0.3. In spring (March to May) and autumn (September to October), temperature fluctuations are larger, and the target temperature range adjustment range is set to ±1.5℃. The activation threshold is adjusted within ±0.8℃, the irrigation activation moisture content threshold within ±0.8%, the soil EC value control threshold within ±0.15mS / cm, and the soil pH value control threshold within ±0.2. In summer (June to August), when temperatures are relatively high, the target temperature range is adjusted within ±1℃, the geothermal compensation activation threshold within ±0.5℃, the irrigation activation moisture content threshold within ±0.5%, the soil EC value control threshold within ±0.1mS / cm, and the soil pH value control threshold within ±0.1. This dynamic adjustment range ensures temperature stability for vegetation growth while avoiding excessive energy consumption by the system.

[0076] The calibrated core temperature control benchmark values ​​and dynamic adjustment ranges are entered into the central control system to form a standardized parameter threshold configuration file. The configuration file is stored in AES-256 encrypted format and simultaneously backed up to a local database (MySQL 8.0) and a cloud server (Alibaba Cloud ECS server). Figure 8As shown, cloud backup employs dual encryption technology to ensure data security. After data entry, three threshold verification tests are conducted, simulating three temperature conditions (-20℃), room temperature (10℃), and high temperature (20℃), three moisture content conditions (low moisture content (12%), suitable moisture content (22%), and high moisture content (28%), three EC value conditions (low EC value (0.8mS / cm), suitable EC value (1.5mS / cm), and high EC value (2.2mS / cm), and three pH value conditions (low pH value (6.2), suitable pH value (7.5), and high pH value (8.8)). The tests aim to verify whether the system can accurately trigger the corresponding operation based on the thresholds. After successful verification, the parameter threshold configuration file officially takes effect.

[0077] S4: Construction of Multi-Algorithm Fusion Model

[0078] like Figure 3 As shown, this step focuses on building a temperature control decision model that deeply integrates four algorithms. Through bidirectional interaction and real-time parameter feedback between the algorithms, it addresses the problem of insufficient temperature control accuracy in complex environments of high-altitude and cold regions. The four algorithms are: fractional-order cooperative perturbation observer, hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, improved Lyapunov adaptive sliding mode control algorithm, and Gaussian process regression-fuzzy cognitive graph algorithm. Each algorithm performs its own function while also deeply collaborating to build a complete temperature control decision system. The specific construction process is as follows:

[0079] The model input variable set is defined as follows ,in This represents the real-time soil temperature collected by a distributed sensor array. This represents the cumulative number of freeze-thaw cycles since the start of construction. Represents real-time ultraviolet radiation intensity. Represents instantaneous wind speed; the model output variable set is defined as follows: ,in Represents the heat release rate of the phase change energy storage unit. This represents the output power of the geothermal compensation device. The input variables directly reflect the dynamic changes in the temperature control environment in high-altitude and cold regions, while the output variables directly determine the operating status of the temperature control actuator. The selection of variables ensures that the model comprehensively represents the temperature control process.

[0080] A fractional-order cooperative perturbation observer is constructed to accurately capture complex environmental disturbances. Disturbances such as freeze-thaw cycles, ultraviolet radiation, and gusts in high-altitude cold regions exhibit dynamic and coupled characteristics. Traditional integer-order observers struggle to capture the memory characteristics and dynamic changes of these disturbances; therefore, fractional-order calculus theory is used to construct the observer. The observer's state equation is designed as follows:

[0081] ;

[0082] in For fractional differential operators, The fractional order is set to 0.6. This value was determined through extensive simulation experiments to ensure both the observer's response speed and its ability to effectively capture the memory characteristics of disturbances. This is an estimate of soil temperature. This is an estimate of the rate of temperature change. This is the estimated total disturbance. To control the input amount, The observer gain coefficients are initially set to 5, 3, and 2, respectively. By co-designing the gain coefficients, the observer asymptotically converges the errors between the temperature estimate and the real-time value, and the errors between the disturbance estimate and the actual value, to zero. This enables accurate observation of multi-source coupled disturbances and provides reliable disturbance input data for subsequent optimization and control algorithms.

[0083] A Gaussian process regression-fuzzy cognitive graph is constructed to achieve accurate short-term soil temperature prediction. To address the challenge of traditional prediction algorithms failing to balance nonlinearity and causal reasoning capabilities, the accurate prediction characteristics of Gaussian process regression are combined with the causal reasoning advantages of fuzzy cognitive graphs. First, the node system of the fuzzy cognitive graph is determined, with six core nodes: real-time temperature, cumulative freeze-thaw cycles, UV intensity, perturbation estimate, phase change heat release rate, and geothermal compensation power. Edges between nodes represent the strength of causal relationships, and the initial weights are set from -1 to 1, with positive weights indicating facilitating effects and negative weights indicating inhibiting effects. Then, Gaussian process regression is embedded into the reasoning process of the fuzzy cognitive graph, using a hybrid kernel function of the squared exponential kernel and the Matern kernel to improve the generalization ability of the prediction. The kernel function expression is as follows:

[0084] ;

[0085] in For any two samples in the training sample set, For signal variance, For length scale, The noise variance is used. Historical temperature data and perturbation estimates are used as training samples. The kernel function parameters are optimized through Gaussian process regression, while the causal relationship strength of the nodes in the fuzzy cognitive graph is dynamically adjusted. Finally, the predicted soil temperature for the next 30 minutes is output. This provides advance notice for temperature control decisions, enabling a proactive temperature control mode based on prediction and regulation.

[0086] A hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm is constructed to achieve the global optimum solution for temperature control parameters. Traditional particle swarm optimization algorithms are prone to getting trapped in local optima. To address this deficiency, a hyperbolic tangent function is introduced to dynamically adjust the inertia weights, balancing the algorithm's global and local search capabilities. Cauchy mutation is introduced to perturb the particle positions, preventing premature convergence. The particle inertia weight update formula is designed as follows:

[0087] ;

[0088] in For the first The inertial weight of the generation, The maximum inertia weight is set to 0.9. The minimum inertia weight is set to 0.4. The adjustment factor is set to 5. The maximum number of iterations is set to 100. The Cauchy mutation formula is designed as follows:

[0089] ;

[0090] in For the first The generation The positions of the particles, with a particle dimension of 2, correspond to the phase change heat release rate and the geothermal compensation power, respectively. To adjust the asynchronous length, set it to 0.1. The random numbers are from a standard Cauchy distribution. The fitness function of the algorithm is defined as a weighted sum of temperature deviation and energy consumption, i.e. ,in For the target temperature, The weight for temperature deviation is set to 0.7. The energy consumption weight is set to 0.3. This fitness function is used to optimize the balance between temperature control accuracy and energy consumption, outputting the optimal parameters for phase change heat release rate and geothermal compensation power. , .

[0091] An improved Lyapunov adaptive sliding mode control algorithm is constructed to enhance the robustness and response speed of the temperature control system. To address the impact of disturbance uncertainties and model errors on temperature control accuracy, a nonlinear sliding surface is designed, and an adaptive law is introduced to dynamically adjust the control gain, offsetting the effects of disturbance observation errors and model uncertainties. The sliding surface is defined as:

[0092] ;

[0093] in For temperature deviation, The rate of change of temperature deviation. The coefficient for the integral term is set to 2. The coefficient for the differential term is set to 1. The design of the nonlinear sliding surface ensures both a fast system response and reduces steady-state error. The adaptive law is designed as follows:

[0094] ;

[0095] in The rate of change of the estimated value of the uncertain parameter. The adaptive gain is set to 0.1. To avoid the chattering problem of traditional sliding mode control, a saturation function is used instead of the sign function, and the control law formula is designed as follows:

[0096] ;

[0097] in These are estimates of uncertain parameters. The scaling factor is set to 3. It is a saturation function. The saturation function parameter is set to 0.1. The predicted temperature from the Gaussian process regression-fuzzy cognitive map is incorporated into the control law as a feedforward compensation term to further improve control accuracy and ensure that the system can still operate stably under complex disturbances.

[0098] A deep interactive algorithm fusion mechanism is designed to enable the collaborative operation of four algorithms. The four algorithms form a closed-loop fusion system through parameter feedback and result interaction. The specific interaction logic is as follows: First, the total perturbation estimate output by the fractional-order collaborative perturbation observer... On the one hand, it serves as a constraint on the fitness function of the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, limiting the particle search range and preventing the optimization parameters from deviating from actual working conditions; on the other hand, it is input into the adaptive law of the improved Lyapunov adaptive sliding mode control algorithm to dynamically adjust the control gain. This enables the control algorithm to counteract disturbances in real time. Secondly, the temperature prediction value output by the Gaussian process regression-fuzzy cognitive map... On the one hand, it serves as a reference for the optimization objective in the hyperbolic tangent-Cauchy mutant particle swarm optimization algorithm, guiding particles to search for the optimal parameters required for future temperature. On the other hand, it serves as a feedforward compensation term in the improved Lyapunov adaptive sliding mode control algorithm, adjusting the control quantity in advance to reduce temperature fluctuations. Third, the optimal parameters output by the hyperbolic tangent-Cauchy mutant particle swarm optimization algorithm... , Feedback is fed into the Gaussian process regression-fuzzy cognitive graph to correct the strength of the causal relationships between nodes, enabling the prediction model to optimize prediction results by combining actual control effects; simultaneously, it serves as the initial control parameters for the improved Lyapunov adaptive sliding mode control algorithm, shortening the convergence time of the control algorithm. Fourth, the control error of the improved Lyapunov adaptive sliding mode control algorithm... (in The actual control input is fed back to the fractional-order cooperative disturbance observer, and the observer gain coefficient is dynamically adjusted. This improves the accuracy of disturbance observation. The specific formula for algorithm interaction is as follows:

[0099] ;

[0100] in The initial gain coefficients for the observer are set to 5, 3, and 2 respectively. The gain adjustment coefficients are set to 0.8, 0.5, and 0.3 respectively. The inertia weight correction factor is set to 0.05. The parameter adjustment coefficient is set to 0.2. Through the above interaction mechanism, the four algorithms form an organic whole, realizing the full-process coordination of disturbance observation, temperature prediction, parameter optimization, and precise control.

[0101] In terms of data support, a comprehensive dataset collected in the early stages was used, including data from laboratory simulations of extremely cold environments (temperature -20℃ to 20℃, wind speed 0m / s to 24.5m / s, ultraviolet radiation intensity 80W / m). 2 Up to 150W / m 2 The 1,000 sets of temperature control data obtained (including precipitation from 0 mm to 50 mm and sunshine duration from 4 h to 12 h) and the 4,000 sets of historical temperature control data collected in the field, covering the correspondence between different temperatures, disturbances, control quantities, and adjustment effects, provide a sufficient data foundation for model construction.

[0102] S5: Fusion Model Training and Optimization

[0103] Data preprocessing and partitioning were performed to provide high-quality data support for model training. The collected 5000 sets of data were divided into training and testing sets in a 7:3 ratio. The training set contained 3500 sets of data for model parameter optimization, while the testing set contained 1500 sets of data for model performance validation. Data preprocessing employed Z-score normalization to eliminate the influence of differences in parameter magnitudes. The normalization formula is as follows: ,in The mean of the parameters, Let be the standard deviation of the parameter. After preprocessing, outlier removal is performed using the 3σ criterion, which removes data that deviates from the mean by more than three times the standard deviation, ensuring the reliability of the training data.

[0104] Specialized training was conducted on a fractional-order cooperative perturbation observer to optimize the observer gain coefficient. The mean square error between the actual perturbation values ​​on the test set and the estimated perturbation values ​​output by the observer was used as the training objective. The mean square error formula is as follows: ,in The number of samples in the test set. For the first The perturbation estimate for each sample, For the first The actual perturbation value of each sample. The gain coefficient is optimized using the gradient descent method. The learning rate is set to 0.001, and the training is iterated 500 times. The mean square error is calculated every 10 iterations. Training is stopped when the mean square error is less than 0.01, and the optimal gain coefficient at this time is saved.

[0105] Joint training of Gaussian process regression and fuzzy cognitive graph is performed to improve temperature prediction accuracy. First, the node causal relationship strength matrix of the fuzzy cognitive graph is initialized using the training set data, and the initial weights are determined using a combination of expert experience and random initialization. Then, Gaussian process regression is used to optimize the kernel function parameters. The optimal parameters are solved using maximum likelihood estimation, while overfitting is avoided through 5-fold cross-validation. During training, the prediction error is calculated every 20 iterations, and the prediction error formula is as follows: ,in The number of samples in the training set. For the first Predicted temperature value for each sample For the first The actual temperature value of each sample. Training stops when the prediction error is less than 0.3℃, and the optimal kernel function parameters and node causal relationship strength matrix are saved.

[0106] The hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm was trained to optimize its search performance. The particle swarm size was set to 30, the maximum number of iterations to 100, and the particle positions were set to values ​​within the phase transition heat release rate of 0 W / m³. 2 Up to 50W / m 2 Geothermal compensation power 0W / m 2 Up to 50W / m 2 Using the temperature deviation and energy consumption of the training set as optimization objectives, the fitness value of each particle is calculated by substituting them into the fitness function. The particle position and velocity are updated according to the update rules of the particle swarm optimization algorithm, and a Cauchy mutation operation is performed every 20 iterations. When the change in the fitness function value is less than 0.0001 for 10 consecutive generations, the algorithm is considered to have converged, and the optimal inertia weight adjustment strategy, variable inertia length, and optimal particle position at this time are saved.

[0107] An improved Lyapunov adaptive sliding mode control algorithm was trained to ensure system stability. Based on the temperature deviation data from the training set, the sliding surface parameters were optimized. and adaptive gain The stability of the control algorithm is verified using Lyapunov stability analysis, and a Lyapunov function is constructed. Taking the derivative with respect to time, we get By adjusting the parameters to make This ensures that all signals in the closed-loop system are consistent and bounded. During training, the system response time and steady-state error under different parameter combinations are recorded, and the parameter combination with a response time of less than 10 minutes and a steady-state error of less than ±0.2℃ is selected as the optimal parameter.

[0108] The overall optimization and performance verification of the fusion model were carried out to ensure that the model meets the needs of practical applications. The trained individual algorithms were integrated according to a preset interaction mechanism, and the whole model was tested using test set data. Test metrics included temperature control accuracy, response time, energy consumption, and robustness. The algorithm interaction parameters were adjusted. The optimization algorithm is optimized to achieve the following: when the model's temperature control accuracy error is less than ±0.5℃, the response time is less than 10 minutes, the energy consumption is reduced by 30% compared to traditional temperature control systems (such as simple electric heating and passive insulation), and the model can still run stably under extreme disturbances (wind speed 24.5m / s, temperature -20℃, precipitation 50mm), the model is considered to be optimized. The optimized fusion model parameters are saved to form the final trained model.

[0109] S6: Multi-algorithm fusion temperature control execution

[0110] like Figure 4 As shown, the data acquisition and transmission process of the distributed sensor array is initiated to provide real-time data support for temperature control. The sensor array continuously collects real-time soil temperature data at a frequency of 5 minutes per acquisition. Cumulative number of freeze-thaw cycles Real-time ultraviolet intensity Instantaneous wind speed Soil moisture content Soil EC value Soil pH value The collected data is preferentially transmitted to the central control system via a low-temperature-adaptive LoRa wireless communication protocol (with signal transmission power optimized to 17dBm for environments below -20℃). A wired backup communication link is also added (using an RS485 bus, deployed on the backbone between the sensor array and the central control system). AES-128 encryption is used during transmission to ensure data integrity. For high-altitude, cold environments, a real-time communication link status monitoring module is added. When the LoRa link signal strength drops below -110dBm, the system automatically switches to the wired backup link, with a switching response time of ≤2 seconds. After receiving the data, the control system first performs data verification, eliminating abnormal data, and using linear interpolation to complete missing data, ensuring the continuity and reliability of the input model data. Data acquisition and transmission rely entirely on… Figure 8 The architecture shown enables real-time parameter monitoring, encrypted transmission, and dual backup.

[0111] A fractional-order cooperative disturbance observer is invoked to accurately capture complex disturbances. The control system feeds the calibrated real-time data into the trained fractional-order cooperative disturbance observer, and uses the observer's state equation, combined with the optimized gain coefficient, to achieve accurate capture of complex disturbances. The total disturbance estimate is calculated. During this process, the observer utilizes the memory properties of fractional differential operators to comprehensively analyze current and historical data, accurately identifying multi-source disturbances such as sudden changes in soil temperature caused by freeze-thaw cycles, heat input fluctuations caused by changes in ultraviolet radiation, and heat loss caused by gusts, providing precise disturbance information for subsequent regulation.

[0112] Run Gaussian process regression-fuzzy cognitive graph to complete short-term soil temperature prediction. Real-time temperature... Disturbance estimates Inputting Gaussian process regression-fuzzy cognitive graph, the model, based on pre-trained kernel function parameters and node causal relationship strength matrix, calculates the predicted soil temperature for the next 30 minutes through probabilistic inference of Gaussian process regression and causal transmission of fuzzy cognitive graph. The prediction results not only provide temperature values ​​but also output prediction confidence levels. When the confidence level is below 90%, the system automatically increases the sensor sampling frequency to once every 2 minutes to improve prediction accuracy.

[0113] The hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm is initiated to solve for the optimal temperature control parameters. The target temperature is used as the starting point. Temperature forecast Disturbance estimates Using this as input, the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm is initiated. The algorithm updates the formula, Cauchy mutation formula, and fitness function according to preset inertia weights, iteratively searching for the optimal values ​​of the phase change heat release rate and geothermal compensation power. During the iteration process, the algorithm outputs intermediate optimization results every 10 generations. The control system monitors the optimization process in real time, and outputs the optimal parameters when the fitness function converges. , This ensures that the parameters meet temperature control requirements while minimizing energy consumption.

[0114] The improved Lyapunov adaptive sliding mode control algorithm is executed to generate and issue control commands. The optimal parameters are then... , Disturbance estimates Temperature forecast The input is an improved Lyapunov adaptive sliding mode control algorithm. The algorithm calculates the temperature deviation and the rate of change of deviation using the sliding surface equation, dynamically adjusts the control gain using an adaptive law, and generates the final control command. Control commands are transmitted to the phase change energy storage unit and geothermal compensation device via RS485 wired communication, ensuring stable and delay-free command transmission.

[0115] The phase change energy storage unit and geothermal compensation device execute temperature control operations according to control commands to achieve precise temperature regulation. Combined with... Figure 7 The photothermal conversion layer absorbs light energy and converts it into heat energy, which is then transferred to the phase change microcapsules (phase change energy storage units) in the microenvironment regulation layer for storage. The phase change energy storage units operate at the optimal heat release rate. During operation, the system absorbs infrared thermal radiation converted from ultraviolet light through a photothermal conversion unit, storing the heat in a paraffin / expanded graphite composite. At night, it releases heat according to control commands, regulating the heat release rate through a built-in temperature control valve. The valve opening is proportional to the heat release rate, with an opening range of 0° to 90°, corresponding to a heat release rate of 0 W / m³. 2 Up to 50W / m 2 The geothermal compensation device operates at its optimal output power. During operation, the power supply to the carbon fiber heating wire is adjusted via a pulse width modulation (PWM) signal, with a duty cycle ranging from 0% to 100%, corresponding to an output power of 0W / m. 2 Up to 50W / m 2 When the control command requires the output power to be 0, the geothermal compensation device automatically shuts down, and the phase change energy storage unit independently completes the temperature regulation. When the temperature is below -15℃, the geothermal compensation device operates at full power, releasing heat in conjunction with the phase change energy storage unit to quickly increase the soil temperature.

[0116] A closed-loop feedback control mechanism is constructed to dynamically optimize temperature control. After each control operation, a distributed sensor array collects temperature feedback data in real time. The data is transmitted to the control system, which then calculates the control error. ,in This refers to the control variables corresponding to the actual operating parameters of the phase change energy storage unit and the geothermal compensation device. Based on the algorithm interaction formula, the control error is fed back to each algorithm, dynamically adjusting parameters such as the observer gain coefficient, particle swarm inertia weight, and sliding mode control gain to optimize the next round of regulation. Simultaneously, the time of each regulation, input data (real-time temperature, freeze-thaw cycles, UV intensity, wind speed, moisture content, EC value, pH value), algorithm outputs (disturbance estimate, temperature prediction, optimal parameters, control commands), temperature feedback data, and parameter adjustment details are recorded to form a detailed temperature control log, providing data support for subsequent model iterations.

[0117] S7: Humidity-coordinated temperature control auxiliary regulation

[0118] Continuous collection of soil moisture content data provides a basis for humidity control. A distributed sensor array integrates temperature and humidity detection functions, synchronously collecting soil moisture content data at a frequency of 10 minutes per data point. The data collection depth is consistent with temperature data collection, covering a soil layer of 0-50cm. Data and temperature data are synchronously transmitted to the central control system, verified, and then stored for later use (data transmission and backup architecture as follows). Figure 8 (As shown).

[0119] Initiate the moisture content threshold judgment and irrigation system control logic. The control system will monitor the real-time moisture content. Compared to the preset irrigation start threshold of 15%, when the real-time moisture content is higher than or equal to 15%, the current state is maintained. The water-retaining agent, humus, and biochar in the seed germination promoting layer work together to retain water, and there is no need to start the irrigation system. When the real-time moisture content is lower than 15%, the control system automatically triggers the adaptive irrigation system start signal, starts the booster pump, and the atomizing nozzles start at a fixed flow rate of 0.5 L / m³. 2 • Minutes are used for atomized spraying (nozzle spacing is as follows) Figure 6 (As shown).

[0120] Precise calculation of spraying time ensures optimal irrigation results. Spraying time is calculated based on the difference between real-time and target moisture content, soil bulk density, seed germination layer thickness, and spray flow rate. The specific formula is as follows: .in The target moisture content is set at 22%. The soil bulk density was determined to be 1.2 g / cm³ by laboratory testing. 3 , The thickness of the seed germination layer is set to 5 cm (e.g., Figure 5 (The gradient matrix layer upper structure shown) The spray flow rate is set to 0.5 L / m. 2 •min. For example, when the real-time moisture content is 12%, the spraying time is calculated as follows: This ensures that the soil moisture content can accurately return to the target range after irrigation.

[0121] Temperature and humidity are controlled in a coordinated manner to avoid adverse effects of irrigation on temperature control. During the spraying process, a distributed sensor array synchronously monitors soil temperature changes, recording temperature data every 2 minutes and calculating the rate of change in moisture content. ,in The time interval is set to 1 minute. If the temperature deviates from the target range by more than 0.5℃, the rate of change in moisture content is input into the multi-algorithm fusion temperature control decision model. The model automatically adjusts the phase change heat release rate and geothermal compensation power. For example, when the moisture content rises rapidly and causes the soil temperature to drop, the model increases the phase change heat release rate by 10% to ensure that the temperature and humidity are in a suitable range for vegetation growth (temperature 5℃-15℃, moisture content 20%-25%, EC value 1.0mS / cm-2.0mS / cm, pH value 6.5-8.5).

[0122] Record the entire process data of temperature and humidity coordinated regulation and form a dedicated log. Record in detail the start time, real-time moisture content, target moisture content, spraying duration, post-irrigation moisture content, temperature change data (temperature before irrigation, real-time temperature during irrigation, and temperature after irrigation), and temperature control parameter adjustments (adjustment value of phase change heat release rate and geothermal compensation power) for each irrigation. This forms a temperature and humidity coordinated regulation log, which is stored in association with the temperature control log to provide complete temperature and humidity coordinated data for subsequent maintenance and model iteration.

[0123] S8: Maintenance Monitoring and Model Iterative Optimization

[0124] Regular drone inspections are conducted to monitor vegetation growth in real time. A DJI Mavic 3 drone is launched weekly at a fixed time (10:00 AM on a sunny day) for inspection. The drone is set to a flight altitude of 50 meters and a speed of 5 m / s, equipped with a 1080P high-definition camera to collect vegetation growth images of the construction area. The flight path is planned in a grid pattern to ensure coverage of the entire construction area without blind spots. After the image data is transmitted to the control system, a convolutional neural network algorithm (ResNet-50) is used for image recognition to analyze key indicators such as seedling emergence rate, plant height, leaf color, leaf quantity, vegetation cover, plant fresh weight, and plant dry weight. The seedling emergence rate is measured by statistically analyzing 10 randomly selected 1m... 2 The number of seedlings emerging within each quadrat was calculated; plant height was obtained through image pixel ratio conversion, with 10 plants measured per quadrat and the average value taken; leaf color was categorized into four types: dark green, light green, yellowish-green, and yellowish, and the percentage of plants with each color was statistically analyzed; leaf count was calculated based on the average number of leaves per plant; vegetation cover was determined by calculating the proportion of vegetation cover area to the total area; plant fresh weight and dry weight were measured on-site by sampling 10 plants. When the emergence rate was below 70%, the control system automatically initiated an intelligent reseeding program, with the reseeding amount being 50% of the initial seeding amount, and the reseeding area precisely located in the area with insufficient emergence.

[0125] Regular component calibration and maintenance are performed to ensure stable system operation. A comprehensive component maintenance is conducted monthly. First, the distributed sensor array is calibrated by comparing sensor measurements with those of a standard thermometer (PT100), hygrometer (HM1500), EC meter (Cond 3310), and pH meter (PHS-3C). If the temperature measurement error exceeds 0.1℃, the moisture content measurement error exceeds 1%, the EC value measurement error exceeds 0.05mS / cm, or the pH value measurement error exceeds 0.1, the sensor is calibrated and adjusted, and then retested to ensure measurement accuracy remains within ±0.5℃. Next, the adaptive irrigation system is inspected, and nozzle blockages are cleared (sprinkler arrangement as shown). Figure 6 (As shown), test the water pump's operating status and the stability of the spray flow, and replace aging water pipes and nozzles; then check the carbon fiber heating wires of the geothermal compensation device (laying spacing as shown). Figure 6 As shown), use a multimeter (model FLUKE 15B+) to test the continuity of the circuit and replace the damaged insulation layer; finally, check the photothermal conversion unit, phase change energy storage unit, and each layer of the composite structure (such as...). Figure 5 (As shown) clean the surface of the components, remove dust and debris, repair any broken parts on the edges, and ensure that the components are functioning properly.

[0126] Incremental training and iterative optimization of the model were implemented to improve system adaptability. Temperature control logs, temperature and humidity coordinated control logs, and vegetation growth data were collected quarterly to form a model iteration dataset. This dataset includes 300 newly added complete temperature-growth correspondence data sets (covering the correspondence between temperature, water content, EC value, pH value, and vegetation growth indicators). The multi-algorithm fusion temperature control decision model was updated using incremental training. During training, the optimized and stable core parameters (fractional order 0.6, sliding surface basic parameters c=2, d=1) were frozen, and only the algorithm interaction parameters were updated. ) and fitness function weights ( This prevents the model from forgetting existing knowledge. Through incremental training, the model can adapt to environmental changes in the target area (seasonal changes, vegetation growth stages), continuously improving temperature control accuracy and vegetation adaptability.

[0127] Conduct vegetation growth monitoring and regulation at key time points. The first vegetation cover test will be conducted 120 days after construction, using a combination of manual on-site measurement and drone image recognition. Manual measurements will be taken at 20 1m intervals. 2The entire area was identified and covered by drones to ensure accurate test results. When vegetation cover was below 75%, enhanced control measures were implemented, increasing irrigation frequency by 10% (from 10 minutes / time to 9 minutes / time), supplementing with 30% of the initial dosage of mycorrhizal fungi-specific coating agent (Glomus mosseae), and adjusting the temperature control target range to maintain a temperature between 8℃ and 12℃, which is suitable for rapid vegetation growth. Vegetation community stability was tested 360 days after construction, with indicators including species diversity (the proportion of species such as *Leymus chinensis*, *Stipa purpurea*, *Leymus chinensis*, *Agropyron cristatum*, and *Poa annua*), community structure uniformity (the coefficient of variation of vegetation cover in different areas), average plant height consistency (the difference in average plant height in different areas), and leaf health rate (the proportion of plants without yellowing or withered leaves) to ensure a stable community. If a stable state was not achieved, temperature and humidity control parameters were further optimized, and the maintenance period was extended to 450 days.

[0128] Complete maintenance data is recorded to create a full-process maintenance archive. This includes detailed records of each drone inspection result (various vegetation growth indicators), component calibration data (sensor calibration errors before and after, component operating parameters), maintenance records (maintenance locations, maintenance measures, maintenance effects), model iteration parameters (updated interaction parameters, weighting coefficients), and vegetation growth indicators (seedling emergence rate, canopy cover, and community stability indicators at each time point). All data is organized chronologically to form a complete maintenance log. This maintenance log is linked to temperature control logs and temperature and humidity coordinated control logs to construct a full-process data archive, providing a reference for long-term system operation evaluation and subsequent engineering applications.

[0129] S9: System Long-Term Performance Verification

[0130] Regular system performance testing will be conducted to comprehensively evaluate the system's operational status. A comprehensive system performance test will be performed every six months after construction, covering wind uplift resistance, freeze-thaw resistance, temperature control accuracy and stability, component connection strength, circuit insulation, and sensor data transmission accuracy. Wind uplift resistance testing will utilize a wind tunnel simulation experiment (model WT-100). System component samples will be placed in the wind tunnel, simulating a level 10 gust (wind speed 24.5 m / s) for 30 minutes. Observation will be conducted to check for loosening or detachment of components, and the displacement and damage will be recorded. Freeze-thaw resistance testing will use a high and low temperature test chamber (model GDW-2005) to simulate the freeze-thaw cycle environment of high-altitude cold regions. One cycle will be from 5℃ to -5℃, and a total of 50 cycles will be performed, each lasting 24 hours. After the test, the structural integrity of the components will be checked (with particular emphasis on verification). Figure 5The structural integrity retention rate is calculated by comparing the connection status of each layer of the composite structure shown. Temperature control accuracy and stability are continuously monitored for 30 days, recording the temperature measurement values ​​at 24 time points per day (once per hour) and the error between the measured temperature and the target temperature, calculating the average and maximum errors. Component connection firmness is tested using a tensile test (model WDW-10) to check if the pull-out force of the anchor nails remains ≥5kN. Line insulation is tested using an insulation resistance meter (model ZC25-3), with an insulation resistance ≥1MΩ. Sensor data transmission accuracy is calculated by comparing the sensor transmission data with standard instrument measurement data, with the error ≤0.5%.

[0131] Targeted optimizations and adjustments will be implemented based on performance test results. If wind uplift resistance fails to meet standards, or if modules become loose or shift, reinforcement measures will be taken, including increasing the number of anchor bolts and densifying the spacing to 1 meter. Additionally, the amount of polyurethane adhesive (model PU-100) will be increased at module joints to enhance connection strength. If the structural integrity retention rate is below 90% after freeze-thaw resistance testing, the paraffin / expanded graphite composite phase change material in the phase change energy storage unit will be replenished, damaged encapsulation films will be replaced, and a permafrost conditioner (polyacrylamide + ferrous sulfate, mass ratio 1:1) will be added to the gradient matrix layer to improve the structure's freeze-thaw resistance. Capabilities: If the average error in the temperature control accuracy and stability test exceeds ±0.5℃, retrain the model, expand the training dataset to 6000 sets, optimize the algorithm interaction parameters, and ensure that the temperature control accuracy is restored to the design standard; if the component connection firmness is not up to standard and the pull-out force is less than 5kN, replace the anchor nail with a higher strength model (basalt fiber nail, 10mm in diameter); if the circuit insulation is not up to standard, rewrap the waterproof insulating tape and replace the damaged wire; if the sensor data transmission accuracy is not up to standard, check the communication protocol and recalibrate the sensor.

[0132] Long-term operational monitoring and evaluation will be conducted, and an annual operational report will be generated. The system's operational status will be continuously monitored for up to its designed 5-year lifespan. Each year, data on temperature control effectiveness (average temperature control accuracy, temperature fluctuation range), vegetation restoration (vegetation cover, community stability, biomass), structural stability (wind uplift resistance, freeze-thaw resistance, component integrity rate), energy consumption (power consumption of geothermal compensation devices, water consumption of irrigation systems), and component operation (sensor failure rate, irrigation system failure rate, geothermal compensation device failure rate) will be compiled into an annual operational report. The report will analyze the system's strengths and weaknesses and propose targeted improvement suggestions (such as optimizing algorithm parameters, replacing with higher-performance components, and adjusting maintenance cycles), providing detailed technical references for the design, construction, and operation of subsequent high-altitude cold-region vegetation soil temperature control projects.

[0133] The core technology of this invention lies in constructing an algorithm fusion model, the core functions of which are explained below:

[0134] 1. Fractional-order Cooperative Disturbance Observer: Its core function is to overcome the limitations of traditional disturbance observation methods and accurately capture dynamic disturbances from multiple sources coupled in high-altitude and cold regions. Utilizing the memory characteristics of fractional-order differential operators, this observer can not only identify the current disturbance intensity but also trace the historical trends of disturbance changes, achieving a comprehensive characterization of complex disturbances such as freeze-thaw cycles, ultraviolet radiation, and gusts. Its output disturbance estimate provides reliable input for subsequent optimization and control algorithms, enabling optimization parameters to specifically avoid the impact of disturbances and control algorithms to counteract disturbance effects in real time, laying the foundation for the precise operation of the temperature control system.

[0135] 2. Hyperbolic Tangent-Cauchy Mutation Particle Swarm Optimization Algorithm: Its core function is to achieve the globally optimal solution for temperature control parameters, balancing temperature control accuracy and energy consumption. By dynamically adjusting the inertia weight through the hyperbolic tangent function, the algorithm maintains strong global search capability in the early stages of iteration, extensively exploring the parameter space; in the later stages of iteration, it enhances local search capability, accurately converging to the optimal solution; the introduction of Cauchy mutation effectively avoids the algorithm getting trapped in local optima, ensuring that the phase change heat release rate and geothermal compensation power parameters found are globally optimal solutions. This algorithm achieves weighted optimization of temperature control accuracy and energy consumption through a fitness function, minimizing system energy consumption and improving the economic efficiency of system operation while ensuring that the soil temperature remains stable within the target range.

[0136] 3. Improved Lyapunov Adaptive Sliding Mode Control Algorithm: Its core function is to enhance the robustness and response speed of the temperature control system, ensuring stable operation under complex disturbances. The design of the nonlinear sliding surface makes the system respond faster to temperature deviations and has smaller steady-state errors; the adaptive law can dynamically adjust the control gain to offset the effects of disturbance observation errors and model uncertainties in real time; the integration of the feedforward compensation term uses temperature prediction values ​​to adjust the control quantity in advance, reducing temperature fluctuations; the use of the saturation function effectively suppresses the chattering problem of traditional sliding mode control, making the temperature control process smoother. This algorithm ensures that the temperature control system can still quickly and accurately control the soil temperature within the target range under extreme environments and complex disturbances in high-altitude and cold regions, providing a stable temperature environment for vegetation growth.

[0137] 4. Gaussian Process Regression-Fuzzy Cognitive Map: Its core function is to achieve accurate short-term prediction of soil temperature, driving the transformation of temperature control from passive response to active regulation. The probabilistic reasoning capability of Gaussian process regression ensures the accuracy and reliability of temperature prediction, while the causal reasoning capability of the fuzzy cognitive map enables the model to dynamically adjust prediction results based on multi-source environmental parameters. The fusion of these two technologies achieves accurate prediction of soil temperature 30 minutes in the future. This prediction provides a preliminary reference for the optimization algorithm, guiding parameter optimization towards future temperature requirements; it also provides feedforward compensation for the control algorithm, adjusting control variables in advance, effectively reducing temperature deviation and fluctuations, enabling the temperature control system to respond to environmental changes in advance, and significantly improving the foresight and accuracy of temperature control.

[0138] The four algorithms form an organic whole through a deep interactive fusion mechanism. The output of each algorithm serves as the input or optimization basis for other algorithms, realizing the full-process collaboration of disturbance observation, temperature prediction, parameter optimization, and precise control. This breaks through the limitations of traditional single-algorithm temperature control, significantly improving the accuracy, stability, adaptability, and economy of temperature control for vegetation soil in high-altitude and cold regions, and providing key temperature protection for vegetation restoration in these regions.

Claims

1. A multi-algorithm fusion method for vegetated soil temperature control based on complex disturbance scenarios in high-altitude and cold regions, characterized in that, include: S1: Collect multi-source environmental data of the target area in the high-altitude cold region through a distributed sensor array, while collecting historical meteorological data and conducting soil sample testing to obtain soil suitability parameters; S2: After pre-treating the slope of the target area, deploy intelligent temperature control core components, which include a photothermal conversion unit, a phase change energy storage unit, a distributed sensor array, an adaptive irrigation system, and a geothermal compensation device. S3: Based on multi-source environmental data and soil suitability parameters, determine the benchmark values ​​and dynamic adjustment ranges of core temperature control indicators; S4: Construct a fusion algorithm model that includes a fractional-order cooperative perturbation observer, Gaussian process regression-fuzzy cognitive graph, hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, and improved Lyapunov adaptive sliding mode control algorithm, and design a deep interactive fusion mechanism for the algorithm; S5: Train the fusion algorithm model, optimize the algorithm parameters, and then adjust the algorithm interaction parameters through overall testing to complete the model optimization; S6: The distributed sensor array is activated to continuously collect real-time data. Based on the core temperature control index benchmark value and dynamic adjustment range, the current target temperature is determined. Based on the optimized fusion algorithm model, multi-source dynamic disturbances are captured by a fractional-order collaborative disturbance observer. Short-term temperature prediction is performed using Gaussian process regression-fuzzy cognitive map. The current target temperature, temperature prediction value, and disturbance estimate are taken as inputs. The optimal heat release rate of the phase change energy storage unit and the optimal output power of the geothermal compensation device are solved by the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm. Control commands are generated based on the improved Lyapunov adaptive sliding mode control algorithm to drive the intelligent temperature control core components to perform temperature control operations: the photothermal conversion unit absorbs ultraviolet rays and converts them into infrared thermal radiation to provide energy for the phase change energy storage unit to store heat; the phase change energy storage unit adjusts the heat release intensity through temperature control valves according to the optimal heat release rate to achieve heat storage during the day and heat release at night. The geothermal compensation device controls the power supply status of the carbon fiber heating wire through pulse width modulation signal according to the optimal output power, starts or stops heating, and adjusts the heating power at the same time. The fractional-order cooperative disturbance observer in S4 is constructed using fractional-order calculus theory. It utilizes the memory characteristics of fractional-order differential operators to trace the historical trend of disturbance changes, thereby achieving comprehensive capture of multi-source dynamic coupling disturbances such as freeze-thaw cycles, ultraviolet radiation, and gusts. The output disturbance estimate provides data support for the search range of the constraint parameters of the hyperbolic tangent-Cauchy mutant particle swarm optimization algorithm and the adjustment of the control gain by the improved Lyapunov adaptive sliding mode control algorithm. The S4 Gaussian process regression-fuzzy cognitive graph integrates the probabilistic prediction characteristics of Gaussian process regression with the causal reasoning ability of fuzzy cognitive graphs to construct a node system that includes real-time temperature, freeze-thaw cycle count, ultraviolet intensity, and disturbance estimates. By dynamically adjusting the strength of causal relationships between nodes, it achieves accurate short-term prediction of soil temperature, providing a forward-looking adaptation basis for the current target temperature determined based on the core temperature control index benchmark value and dynamic adjustment range, and assisting in optimizing the control timing of phase change energy storage units and geothermal compensation devices. The hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm in S4 dynamically adjusts the inertia weight through the hyperbolic tangent function to balance the global search and local search capabilities of the algorithm. At the same time, it introduces the Cauchy mutation mechanism to perturb the particle position to avoid the algorithm getting trapped in local optima. Taking the deviation between the current target temperature and the real-time temperature and the system energy consumption as optimization objectives, it solves the optimal heat release rate of the phase change energy storage unit and the optimal output power of the geothermal compensation device to meet the core temperature control index requirements. The improved Lyapunov adaptive sliding mode control algorithm in S4 designs a nonlinear sliding mode surface, introduces an adaptive law to dynamically adjust the control gain, offsets the effects of disturbance observation errors and model uncertainties, incorporates temperature prediction values ​​as feedforward compensation terms, and uses a saturation function to suppress chattering, ensuring that the generated control commands can accurately drive the temperature control valve of the phase change energy storage unit and the carbon fiber heating wire of the geothermal compensation device, stabilizing the soil temperature within the target range specified by the core temperature control index; The deep interaction and fusion mechanism of the S4 algorithm is as follows: the perturbation estimate of the fractional-order cooperative perturbation observer is input into the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm and the improved Lyapunov adaptive sliding mode control algorithm respectively, constraining the parameter search range and adjusting the control gain; the temperature prediction value of Gaussian process regression-fuzzy cognitive map guides the direction of parameter optimization and serves as feedforward compensation; the optimal heat release rate and optimal output power output by the optimization algorithm are fed back to correct the node relationship strength of the prediction model, and at the same time serve as the initial parameters of the control algorithm to shorten the convergence time; the control error feedback of the control algorithm adjusts the observer gain coefficient, forming a closed-loop cooperative system that adapts to the core temperature control index requirements.

2. The multi-algorithm fusion method for vegetated soil temperature control based on complex disturbance scenarios in high-altitude and cold regions as described in claim 1, is characterized in that... Also includes: S7: The system continuously collects soil moisture data through a distributed sensor array, compares the soil moisture content with the irrigation start-up moisture content benchmark value and the dynamic adjustment range. If the soil moisture content deviates from the suitable range, the system will start the adaptive irrigation system to replenish water. During the water replenishment process, the soil temperature changes are monitored simultaneously. The moisture content change information is input into the fusion algorithm model. The model combines the core temperature control index benchmark value and the dynamic adjustment range to adjust the optimal temperature control parameters, driving the phase change energy storage unit and geothermal compensation device to adapt and adjust the heat release rate and heating power, so as to achieve temperature and humidity coordinated adaptation.

3. The multi-algorithm fusion method for vegetated soil temperature control based on complex disturbance scenarios in high-altitude cold regions as described in claim 1, is characterized in that... The training and optimization of the fusion algorithm model in S5 includes: specifically optimizing the gain coefficient of the fractional-order cooperative perturbation observer, the kernel function parameters and node weights of the Gaussian process regression-fuzzy cognitive graph, the inertial weights and mutation parameters of the hyperbolic tangent-Cauchy mutation particle swarm optimization algorithm, and the sliding surface parameters and adaptive gain of the improved Lyapunov adaptive sliding mode control algorithm. The optimal parameters of each algorithm are determined through stability analysis and error verification. Then, based on the core temperature control index requirements of temperature control accuracy, response speed, robustness, energy economy, and adaptive capability, the algorithm interaction parameters are adjusted to ensure that the fusion model can accurately drive the core components of intelligent temperature control to adapt to the complex environment of high-altitude and cold regions.

4. The multi-algorithm fusion method for vegetated soil temperature control based on complex disturbance scenarios in high-altitude cold regions as described in claim 1, characterized in that, In S6, closed-loop feedback control is implemented. Specifically, the temperature feedback data after temperature control is executed is compared with the current target temperature specified by the core temperature control index to obtain the control error. Based on this control error, the gain coefficient of the fractional-order cooperative disturbance observer, the inertial weight of the hyperbolic tangent-Cauchy mutant particle swarm optimization algorithm, and the control gain of the improved Lyapunov adaptive sliding mode control algorithm are dynamically adjusted. This optimizes the heat release rate of the next phase change energy storage unit and the output power of the geothermal compensation device, ensuring that the temperature remains stable within the dynamic adjustment range of the core temperature control index.

5. The multi-algorithm fusion method for vegetated soil temperature control based on complex disturbance scenarios in high-altitude and cold regions according to claim 2, characterized in that, The temperature and humidity coordinated adaptation described in S7 is as follows: During the water replenishment process of the adaptive irrigation system, the distributed sensor array collects soil temperature data at fixed intervals, calculates the temperature change rate and inputs it into the fusion algorithm model. The model combines the target temperature range of the core temperature control index and the moisture content change trend to adjust the opening of the temperature control valve of the phase change energy storage unit and the power supply duty cycle of the carbon fiber heating wire of the geothermal compensation device in real time, and adjusts the heat release rate and heating power accordingly to avoid the temperature deviating from the suitable range specified by the core temperature control index due to water replenishment.

Images