Plateau airport oxygen environment adaptive control method and system based on model driving

Abstract

The invention discloses a model-driven plateau airport oxygen environment adaptive control method and system, relates to the technical field of plateau airport environment control, and solves the technical problem that adaptive oxygen supplementation of a plateau airport cannot be realized in the prior art. The method comprises the steps of 1, collecting environment data, operation data and air reference parameters of an airport in real time; 2, correcting oxygen and air density in the environmental data; step 3, establishing a multi-modal fusion adaptive integration model MF-AIEM and performing model training, wherein the MF-AIEM comprises a bottom sensing layer, a middle fusion layer and a top decision layer; 4, the trained MF-AIEM is coupled with multi-source data to predict the oxygen demand quantity, the fresh air volume is regulated and controlled, and the oxygen supplementation quantity is calculated to achieve oxygen supplementation regulation and control; the system effectively adapts to plateau low pressure, dynamic people flow and environment sudden change, and fine guarantee of air quality and improvement of system energy efficiency are achieved.

Description

Technical Field

[0001] This invention relates to the field of environmental control technology for high-altitude airports, specifically to a model-driven adaptive control method and system for oxygen environment in high-altitude airports. Background Technology

[0002] The operational environment of high-altitude airports faces unique challenges, primarily stemming from the combined effects of their special natural environment and dynamic passenger flow. Atmospheric pressure at high altitudes is typically only 60%–70% of that at sea level, resulting in a significantly reduced oxygen partial pressure and insufficient natural oxygen levels within the terminal building, failing to meet human health and comfort needs. Simultaneously, the low temperature and low humidity further exacerbate the problem of dry air. During operation, high-density passenger flow caused by flight takeoffs and landings leads to accelerated oxygen consumption and a rapid increase in CO2 concentration in localized areas, placing additional pressure on environmental quality. Faced with spatiotemporal heterogeneous changes caused by sudden changes in air pressure and flight delays, existing control methods often suffer from delayed oxygen supply response, waste due to excessive oxygen supplementation, or health risks due to insufficient oxygen supply, while also exhibiting high energy consumption. How to accurately establish and update online a dynamic impact model of multiple factors such as passenger density, air pressure, temperature, and humidity on oxygen demand, and how to achieve efficient coordination between the fresh air system and the oxygen supplementation system based on this model, are technical problems that require further solutions. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides a model-driven adaptive oxygen environment control method and system for plateau airports, solving the technical problem that the prior art cannot achieve adaptive oxygen supplementation for plateau airports.

[0004] A model-driven adaptive control method for oxygen environment at high-altitude airports includes: Step 1: Collect environmental data, operational data and air baseline parameters of the airport in real time to provide multi-source input for the subsequent multimodal fusion adaptive model, and support the dynamic control of oxygen supplementation and fresh air system of plateau airport; Step 2: Considering the decrease in air density and the reduction in oxygen transport efficiency caused by low air pressure at high altitudes, correct the oxygen concentration and calculate the air density. Step 3: Establish a multimodal fusion adaptive ensemble model MF-AIEM and train the model. The MF-AIEM includes a bottom-level perception layer, a middle-level fusion layer, and a top-level decision layer. Step 4: Use the trained MF-AIEM to couple the multi-source data corrected in Step 2 to predict oxygen demand, regulate fresh air volume, and calculate supplemental oxygen to achieve supplemental oxygen regulation.

[0005] Furthermore, the environmental data mentioned in step 1 includes: air pressure inside and outside the terminal building. (Pa), temperature (°C), humidity (kg / kg), O2 concentration (%vol), CO2 concentration (ppm); Operational data includes: real-time passenger numbers (People), flight schedules, and regional population density (persons / ㎡); Air reference parameters include: standard atmospheres (Pa), air gas constant (J / kg·K).

[0006] By achieving second-level data synchronization with the airport information system through sensor networks, an input foundation is provided for multimodal fusion models.

[0007] Furthermore, the correction of oxygen concentration in step 2 includes:

[0008] In the formula, This is the corrected oxygen concentration (%vol); The actual oxygen concentration (%vol) measured by the sensor. Standard atmospheric pressure (101325 Pa); This is the actual atmospheric pressure (Pa).

[0009] The calculation of air density includes: The oxygen partial pressure at different altitudes at airports exhibits nonlinear decay; therefore, an altitude-graded compensation factor is introduced into the air density calculation.

[0010] The corrected air density is:

[0011] In the formula, As an altitude compensation factor, it increases non-linearly with increasing altitude to compensate for the impact of the sudden drop in air density at ultra-high altitudes on oxygen transport. The altitude of the airport is (m). Corrected air density (kg / m³) 3 ); The gas constant of air ( ), The temperature is the thermodynamic temperature (K).

[0012] Furthermore, the underlying perception layer described in step 3 uses a spatiotemporal feature extraction module to analyze the spatiotemporal distribution of environmental data, regional heat maps of pedestrian density, and temporal correlation of flight frequencies, respectively. The middle fusion layer constructs a dynamic relationship graph of environment-people flow-flight, and uses graph neural networks to capture cross-modal coupling effects, including the amplification effect of people gathering on local oxygen consumption and changes in CO2 accumulation rate caused by flight delays. The top-level decision-making layer adopts a multi-task deep reinforcement learning framework, with air quality safety and minimum energy consumption as dual objectives, and outputs a joint optimization strategy for oxygen demand, fresh air volume and oxygen supplementation.

[0013] Furthermore, the multi-task deep reinforcement learning framework includes the following modules: Status: Includes corrected environmental parameters, passenger flow data, flight status, current device status, and time information.

[0014] Actions: Outputs include oxygen demand, fresh air volume, oxygen supplementation volume, and equipment control commands (fan speed, oxygen generator power setting, valve opening).

[0015] Reward function: A key signal driving the optimization of multi-task deep reinforcement learning models, including safety reward, comfort reward, and energy consumption reward objectives.

[0016] The key signal is defined as a multi-objective weighted sum:

[0017] In the formula, For multi-objective weighted sum, For the safety reward mechanism, For comfort reward mechanism, As an energy consumption reward mechanism, These are the weights for the corresponding reward mechanisms.

[0018] Furthermore, the security reward mechanism is for the region. and Positive rewards are given when the threshold is reached, and significant negative rewards (penalties) are given when the threshold is approached or exceeded:

[0019] In the formula, This represents the real-time oxygen concentration. To set the oxygen concentration, This represents the real-time carbon dioxide concentration. To set the carbon dioxide concentration; Furthermore, the comfort reward mechanism encourages temperature and humidity to remain within a set comfort range, with greater penalties for larger deviations.

[0020] In the formula, To set a comfortable temperature; To set a comfortable humidity level; Furthermore, the energy consumption reward mechanism is negatively correlated with the total energy consumption of the system (fans, oxygen generators, etc.), encouraging energy consumption reduction; wherein the power is calculated based on the equipment power model and operating status. Furthermore, the expression for predicting oxygen demand in step 4 is as follows:

[0021] In the formula, To predict oxygen demand (m 3 / h); Real-time passenger count (people); Pressure deviation (Pa); Temperature deviation (°C); The actual temperature (°C) is the measured temperature. Set a comfortable temperature (°C); Humidity deviation (kg / kg); The measured humidity is (kg / kg). Set the comfortable humidity (kg / kg); This is the actual measured air pressure; Standard atmospheric pressure; This is due to air pressure deviation; These are the weight coefficients obtained during model training.

[0022] Furthermore, the fresh air volume regulation mentioned in step 4 includes: using the zone air quality safety threshold as a hard constraint, dynamically calculating the fresh air volume in conjunction with energy-saving targets, and reducing system energy consumption by maximizing natural oxygen supply and minimizing mechanical oxygen supplementation. The corresponding fresh air volume calculation formula is as follows:

[0023] In the formula, Fresh air demand (m 3 / h); For safety reasons, Passenger average CO2 productivity (m 3 / h / person); This refers to the upper limit of CO2 concentration (ppm). Outdoor CO2 concentration (ppm); Based on the corrected fresh air oxygen content ( ) and predicted oxygen demand ( Compare: If If the oxygen content of the fresh air is sufficient to meet the demand, then supplemental oxygen is not initiated; otherwise, the oxygen deficiency is calculated, i.e., the predicted oxygen demand exceeds the fresh air supply capacity, and the difference is used to initiate gradient supplemental oxygen supply. The expression is as follows: ; If the real-time O2 concentration in any region is lower than the preset absolute safety threshold Then ignore the current predicted oxygen supplementation amount. And energy consumption optimization goals; directly trigger the "emergency oxygen supplementation mode" for the entire station or the area, maximize the output power of the oxygen generator, adjust the fresh air valve according to safety needs; issue audible and visual alarms; continuously monitor the oxygen concentration in the area, and only release oxygen when it rises back to the safe threshold. Once the above conditions are met and the situation remains stable, the emergency mode will be exited, and normal optimized regulation will resume.

[0024] Furthermore, it also includes step 5: establishing a real-time optimization and security mechanism, constructing a three-layer protection system, and ensuring the robustness and economy of the system in the complex environment of the plateau. The three-layer protection system is as follows: A rolling time-domain optimization layer is used to divide short-term optimization cycles. At the beginning of each cycle, the state input of the MF-AIEM model is updated based on the latest sensor data, flight information and passenger flow status. Prediction and decision-making are rerun, and the set values ​​of actuators such as fans, oxygen generators and valves are dynamically adjusted. This effectively solves the control lag problem caused by dynamic scenarios such as sudden changes in air pressure and pulse fluctuations in passenger flow.

[0025] The safety constraint embedding layer sets a dynamic penalty term in the DRL reward function; when the predicted oxygen concentration approaches the risk threshold or the real-time carbon dioxide concentration exceeds the warning value, the safety reward weight is automatically increased significantly or an additional safety penalty is added, while the energy consumption penalty is temporarily ignored or reduced; this prompts the DRL agent to take safety-protecting actions first, achieving strong safety intervention.

[0026] In the fault-tolerant and energy efficiency management layer, when data from critical sensors (such as O2 and CO2 sensors) is abnormal or lost, the system automatically switches to a backup prediction mode based on historical flight and passenger flow statistics (such as time series prediction and mean filling) and triggers an operation and maintenance warning; ensuring that basic control functions are not interrupted; combining time-of-use electricity price information, utilizing building thermal inertia or energy storage equipment for pre-cooling / pre-heating during off-peak electricity price periods; appropriately adjusting temperature setpoints while meeting comfort requirements; optimizing equipment start-up and shutdown sequences; smoothing out peak electricity loads and reducing overall operating costs.

[0027] Furthermore, the short-term optimization period is dynamically adjusted, with a peak period period of 5 minutes and a stable period period of 15 minutes.

[0028] A model-driven adaptive control system for oxygen environment at high-altitude airports includes: The data acquisition subsystem collects environmental data, operational data, and air baseline parameters from the airport in real time, providing multi-source input for subsequent multimodal fusion adaptive models and supporting the dynamic control of oxygen supplementation and fresh air systems at plateau airports. The data correction module corrects the oxygen concentration and calculates the air density in response to the decrease in air density and the reduction in oxygen transport efficiency caused by the low air pressure at high altitudes. The decision-making oxygen supplementation regulation subsystem adopts a pre-trained multimodal fusion adaptive ensemble model MF-AIEM. Through a hierarchical structure, it achieves deep coupling of environmental, passenger flow, and flight data to predict oxygen demand, regulate fresh air volume, and calculate oxygen supplementation amount to realize oxygen supplementation regulation.

[0029] The beneficial effects of this invention include: This invention uses a data acquisition and preprocessing mechanism, combined with a hierarchical structure, to achieve deep coupling of environmental, passenger flow, and flight data to determine the oxygen supplementation strategy; it then adopts a synergistic approach of maximizing natural oxygen supply and minimizing mechanical oxygen supplementation to reduce system energy consumption; and it uses a combination of intelligent oxygen supplementation regulation and emergency response to ensure that the oxygen concentration in the terminal building is within a safe range. Attached Figure Description

[0030] Figure 1 This is a flowchart of a model-driven adaptive oxygen environment control method for high-altitude airports, as described in an embodiment of this application.

[0031] Figure 2 This is a model structure diagram of the multimodal fusion adaptive ensemble model MF-AIEM involved in the embodiments of this application.

[0032] Figure 3 This is a schematic diagram of intelligent oxygen supplementation regulation and emergency response involved in the embodiments of this application. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0034] Example 1 The following is in conjunction with the appendix Figure 1-3 Specific embodiments of the present invention will be described in detail; Model-driven adaptive control method for oxygen environment at high-altitude airports, such as Figure 1 As shown, it includes: Step 1: By collecting environmental data, operational data and air baseline parameters of the airport in real time, multi-source inputs are provided for the subsequent multimodal fusion adaptive model, supporting the dynamic control of oxygen supplementation and fresh air system in plateau airports; The environmental data includes: air pressure inside and outside the terminal building. (Pa), temperature (°C), humidity (kg / kg), O2 concentration (%vol), CO2 concentration (ppm); Operational data includes: real-time passenger numbers (People), flight schedules, and regional population density (persons / ㎡); Air reference parameters include: standard atmospheres (Pa), air gas constant (J / kg·K).

[0035] By achieving second-level data synchronization with the airport information system through sensor networks, an input foundation is provided for multimodal fusion models.

[0036] Step 2: Considering the decrease in air density and the reduction in oxygen transport efficiency caused by low air pressure at high altitudes, correct the oxygen concentration and calculate the air density. The correction of oxygen concentration includes:

[0037] In the formula, This is the corrected oxygen concentration (%vol); The actual oxygen concentration (%vol) measured by the sensor. Standard atmospheric pressure (101325 Pa); This is the actual atmospheric pressure (Pa).

[0038] The calculation of air density includes: The oxygen partial pressure at different altitudes at airports exhibits nonlinear decay; therefore, an altitude-graded compensation factor is introduced into the air density calculation.

[0039] The corrected air density is:

[0040] In the formula, As an altitude compensation factor, it increases non-linearly with increasing altitude to compensate for the impact of the sudden drop in air density at ultra-high altitudes on oxygen transport. The altitude of the airport is (m). Corrected air density (kg / m³) 3 ); The gas constant of air ( ), The temperature is the thermodynamic temperature (K).

[0041] Step 3: Build and train the multimodal fusion adaptive ensemble model MF-AIEM. Figure 2 As shown, the multimodal fusion adaptive integration model includes a bottom perception layer, a middle fusion layer, and a top decision layer; The underlying perception layer uses a spatiotemporal feature extraction module to analyze the spatiotemporal distribution of environmental data, regional heat maps of pedestrian density, and temporal correlations of flight frequencies. The middle fusion layer constructs a dynamic relationship graph of environment-people flow-flight, and uses graph neural networks to capture cross-modal coupling effects, including the amplification effect of people gathering on local oxygen consumption and changes in CO2 accumulation rate caused by flight delays. The top-level decision-making layer adopts a multi-task deep reinforcement learning framework, with air quality safety and minimum energy consumption as dual objectives, and outputs a joint optimization strategy for oxygen demand, fresh air volume and oxygen supplementation.

[0042] The multi-task deep reinforcement learning framework includes the following modules: Status: Includes corrected environmental parameters, passenger flow data, flight status, current device status, and time information.

[0043] Actions: Outputs include oxygen demand, fresh air volume, oxygen supplementation volume, and equipment control commands (fan speed, oxygen generator power setting, valve opening).

[0044] Reward function: A key signal driving the optimization of multi-task deep reinforcement learning models, including safety reward, comfort reward, and energy consumption reward objectives.

[0045] The key signal is defined as a multi-objective weighted sum:

[0046] In the formula, For multi-objective weighted sum, For the safety reward mechanism, For comfort reward mechanism, As an energy consumption reward mechanism, These are the weights for the corresponding reward mechanisms.

[0047] The security reward mechanism is regional. and Positive rewards are given when the threshold is reached, and significant negative rewards (penalties) are given when the threshold is approached or exceeded:

[0048] In the formula, This represents the real-time oxygen concentration. To set the oxygen concentration, This represents the real-time carbon dioxide concentration. To set the carbon dioxide concentration; The comfort reward mechanism encourages temperature and humidity to remain within a set comfort range, with greater penalties for larger deviations.

[0049] In the formula, To set a comfortable temperature; To set a comfortable humidity level; The energy consumption reward mechanism is negatively correlated with the total energy consumption of the system (fans, oxygen generators, etc.) to encourage energy reduction; the power is calculated based on the equipment power model and operating status. Step 4: Using the trained MF-AIEM coupled with the multi-source data corrected in Step 2, predict oxygen demand, regulate fresh air volume, and calculate supplemental oxygen volume to achieve supplemental oxygen regulation. Specifically, as follows... Figure 3 As shown.

[0050] The expression for predicting oxygen demand is as follows:

[0051] In the formula, To predict oxygen demand (m 3 / h); Real-time passenger count (people); Pressure deviation (Pa); Temperature deviation (°C); The actual temperature (°C) is the measured temperature. Set a comfortable temperature (°C); Humidity deviation (kg / kg); The measured humidity is (kg / kg). Set the comfortable humidity (kg / kg); The measured air pressure (Pa); Standard atmospheric pressure (Pa); Pressure deviation (Pa); These are the weight coefficients obtained during model training.

[0052] The regulation of fresh air volume includes: using the zoned air quality safety threshold as a hard constraint, dynamically calculating the fresh air volume in conjunction with energy-saving targets, and reducing system energy consumption by maximizing natural oxygen supply and minimizing mechanical oxygen supplementation. The corresponding fresh air volume calculation formula is as follows:

[0053] In the formula, Fresh air demand (m 3 / h); The safety margin is determined based on experience. Passenger average CO2 productivity (m3 / h / person); This refers to the upper limit of CO2 concentration (ppm). Outdoor CO2 concentration (ppm); Based on the corrected fresh air oxygen content ( ) and predicted oxygen demand ( Compare: If If the oxygen content of the fresh air is sufficient to meet the demand, then supplemental oxygen is not initiated; otherwise, the oxygen deficiency is calculated, i.e., the predicted oxygen demand exceeds the fresh air supply capacity, and the difference is used to initiate gradient supplemental oxygen supply. The expression is as follows: ; If the real-time O2 concentration in any region is lower than the preset absolute safety threshold Then ignore the current predicted oxygen supplementation amount. And energy consumption optimization goals; directly trigger the "emergency oxygen supplementation mode" for the entire station or the area, maximize the output power of the oxygen generator, adjust the fresh air valve according to safety needs; issue audible and visual alarms; continuously monitor the oxygen concentration in the area, and only release oxygen when it rises back to the safe threshold. Once the above conditions are met and the situation remains stable, the emergency mode will be exited, and normal optimized regulation will resume.

[0054] In another embodiment, step 5 is also included: establishing a real-time optimization and security mechanism, constructing a three-layer protection system to ensure the robustness and economy of the system in the complex environment of the plateau. The three-layer protection system is as follows: A rolling time-domain optimization layer is used to divide short-term optimization cycles. At the beginning of each cycle, based on the latest sensor data, flight information, and passenger flow status, the state input of the MF-AIEM model is updated, and the prediction and decision-making process is rerun, dynamically adjusting the setpoints of actuators such as fans, oxygen generators, and valves. This effectively solves the control lag problem caused by dynamic scenarios such as sudden changes in air pressure and pulse fluctuations in passenger flow. The time cycle is dynamically adjusted, with a peak period cycle of 5 minutes and a stable period cycle of 15 minutes.

[0055] In another embodiment, a model-driven adaptive control system for oxygen environment at a high-altitude airport is involved, comprising: The data acquisition subsystem collects environmental data, operational data, and air baseline parameters from the airport in real time, providing multi-source input for subsequent multimodal fusion adaptive models and supporting the dynamic control of oxygen supplementation and fresh air systems at plateau airports. The data correction module corrects the oxygen concentration and calculates the air density in response to the decrease in air density and the reduction in oxygen transport efficiency caused by the low air pressure at high altitudes. The decision-making oxygen supplementation regulation subsystem adopts a pre-trained multimodal fusion adaptive ensemble model MF-AIEM. Through a hierarchical structure, it achieves deep coupling of environmental, passenger flow, and flight data to predict oxygen demand, regulate fresh air volume, and calculate oxygen supplementation amount to realize oxygen supplementation regulation.

[0056] The embodiments described above are merely illustrative of specific implementations of this application, and while the descriptions are detailed, they should not be construed as limiting the scope of protection of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the technical solution of this application, and these modifications and improvements all fall within the scope of protection of this application.

Claims

1. A model-driven adaptive control method for oxygen environment at high-altitude airports, characterized in that, Includes the following steps: Step 1: Collect multi-source data from the airport in real time, including environmental data, operational data, and air baseline parameters; Step 2: Considering the decrease in air density and the reduction in oxygen transport efficiency caused by low air pressure at high altitudes, the oxygen and air density in the environmental data are corrected to obtain the corrected multi-source data. Step 3: Establish a multimodal fusion adaptive ensemble model MF-AIEM and train the model. The MF-AIEM includes a bottom-level perception layer, a middle-level fusion layer, and a top-level decision layer. Step 4: Use the trained MF-AIEM to couple the multi-source data corrected in Step 2 to predict oxygen demand, regulate fresh air volume, and calculate supplemental oxygen to achieve supplemental oxygen regulation.

2. The model-driven adaptive control method for oxygen environment at high-altitude airports according to claim 1, characterized in that, Step 2, which involves correcting the oxygen and air density data in the environmental data, includes: The corrected oxygen is: ; In the formula, This is the corrected oxygen concentration; The actual oxygen concentration measured by the sensor; Standard atmospheric pressure; This is the actual atmospheric pressure; The oxygen partial pressure at different altitudes at airports exhibits nonlinear decay; therefore, an altitude-graded compensation factor is introduced into the air density calculation. ; The corrected air density is: ; In the formula, As an altitude compensation factor, it increases non-linearly with increasing altitude to compensate for the impact of the sudden drop in air density at ultra-high altitudes on oxygen transport. This refers to the airport's altitude. This is the corrected air density; The gas constant of air, It is the thermodynamic temperature.

3. The model-driven adaptive control method for oxygen environment at high-altitude airports according to claim 1, characterized in that, In step 3, the underlying perception layer uses a spatiotemporal feature extraction module to analyze the spatiotemporal distribution of environmental data, regional heat maps of pedestrian density, and temporal correlation of flight frequencies. The middle fusion layer constructs a dynamic relationship graph of environment-people flow-flight, and uses graph neural networks to capture cross-modal coupling effects, including the amplification effect of people gathering on local oxygen consumption and the change in CO2 accumulation rate caused by flight delays; The top-level decision-making layer adopts a multi-task deep reinforcement learning framework, with air quality safety and minimum energy consumption as dual objectives, and outputs a joint optimization strategy for oxygen demand, fresh air volume and oxygen supplementation.

4. The model-driven adaptive control method for oxygen environment at high-altitude airports according to claim 3, characterized in that, The multi-task deep reinforcement learning framework includes the following modules: Status: Includes corrected environmental parameters, passenger flow data, flight status, current device status, and time information; Actions: Outputs include oxygen demand, fresh air volume, oxygen replenishment volume, and equipment control commands; Reward function: a key signal driving the optimization of multi-task deep reinforcement learning models, including safety reward, comfort reward, and energy consumption reward objectives; The key signal is defined as a multi-objective weighted sum: ; In the formula, For multi-objective weighted sum, For the safety reward mechanism, For comfort reward mechanism, As an energy consumption reward mechanism, These are the weights for the corresponding safety reward mechanism, comfort reward mechanism, and energy consumption reward mechanism, respectively.

5. The model-driven adaptive control method for oxygen environment at high-altitude airports according to claim 4, characterized in that, The security reward mechanism is based on the area. Receive a positive reward when the opportunity arises, otherwise receive a negative reward: ; In the formula, This represents the real-time oxygen concentration. To set the oxygen concentration, This represents the real-time carbon dioxide concentration. To set the carbon dioxide concentration; The comfort reward mechanism is used to maintain temperature and humidity within a set comfort range. ; In the formula, To set a comfortable temperature; To set a comfortable humidity level; The energy consumption reward mechanism is negatively correlated with the total energy consumption of the system in order to reduce energy consumption; the power is calculated based on the device power model and the operating state.

6. The model-driven adaptive control method for oxygen environment at high-altitude airports according to claim 1, characterized in that, The expression for predicting oxygen demand in step 4 is as follows: ； In the formula, To predict oxygen demand; Real-time passenger count; This is due to air pressure deviation; Temperature deviation; This is the measured temperature; To set a comfortable temperature; Humidity deviation; This is the actual measured humidity. To set a comfortable humidity level; This is the actual measured air pressure; Standard atmospheric pressure; This is due to air pressure deviation; These are the weight coefficients obtained during model training.

7. The model-driven adaptive control method for oxygen environment at high-altitude airports according to claim 1, characterized in that, Step 4, which involves regulating the fresh air volume, includes: using the zoned air quality safety threshold as a hard constraint, dynamically calculating the fresh air volume in conjunction with energy-saving targets, and reducing system energy consumption by maximizing natural oxygen supply and minimizing mechanical oxygen supplementation. The corresponding fresh air volume calculation formula is as follows: ； In the formula, For the demand for fresh air, To predict oxygen demand, For the corrected oxygen concentration, For safety reasons, Average CO2 productivity per passenger; This represents the upper limit of CO2 concentration. This represents the outdoor CO2 concentration.

8. The model-driven adaptive control method for oxygen environment at high-altitude airports according to claim 7, characterized in that, Correction of fresh air oxygen content With predicted oxygen demand Comparison: If If the oxygen content of the fresh air is sufficient to meet the demand, then supplemental oxygen is not initiated; otherwise, the oxygen deficiency is calculated, i.e., the predicted oxygen demand exceeds the fresh air supply capacity, and the difference is used to initiate gradient supplemental oxygen supply. The expression is as follows: ; If the real-time O2 concentration in any region is lower than the preset absolute safety threshold Then ignore the current predicted oxygen supplementation amount. And energy consumption optimization goals; directly trigger the emergency oxygen supplementation mode for the entire station or the area, maximize the output power of the oxygen generator, adjust the fresh air valve according to safety needs, and issue audible and visual alarms, then maintain the oxygen concentration in the area until it rises back to the safe threshold. Once the above conditions are met and the situation remains stable, the emergency mode will be exited, and normal optimized regulation will resume.

9. The model-driven adaptive control method for oxygen environment at high-altitude airports according to claim 1, characterized in that, It also includes step 5: establishing a real-time optimization and security mechanism and constructing a three-layer protection system, including a rolling time-domain optimization layer, a security constraint embedding layer, and a fault tolerance and energy efficiency management layer; The rolling time-domain optimization layer divides short-term optimization cycles. At the beginning of each cycle, prediction and decision-making are rerun based on the latest data, and the set values ​​of the actuators are dynamically adjusted. The safety constraint embedding layer sets a dynamic penalty term in the DRL reward function to achieve strong safety intervention when the predicted oxygen concentration approaches the risk threshold or the real-time carbon dioxide concentration exceeds the warning value. The fault tolerance and energy efficiency management system automatically switches to a backup prediction mode based on historical flight and passenger flow statistics when critical sensor data is abnormal or lost, and triggers an operation and maintenance warning.

10. A model-driven adaptive control method for oxygen environment at high-altitude airports, characterized in that, A model-driven adaptive control system for oxygen environment at high-altitude airports includes: The data acquisition subsystem collects environmental data, operational data, and air baseline parameters from the airport in real time, providing multi-source input for subsequent multimodal fusion adaptive models and supporting the dynamic control of oxygen supplementation and fresh air systems at plateau airports. The data correction module corrects the oxygen concentration and calculates the air density in response to the decrease in air density and the reduction in oxygen transport efficiency caused by the low air pressure at high altitudes. The decision-making oxygen supplementation regulation subsystem adopts a pre-trained multimodal fusion adaptive ensemble model MF-AIEM. Through a hierarchical structure, it achieves deep coupling of environmental, passenger flow, and flight data to predict oxygen demand, regulate fresh air volume, and calculate oxygen supplementation amount to realize oxygen supplementation regulation.

Images