Air conditioner door and window joint control method and system

Abstract

The application relates to the technical field of ventilation regulation, in particular to an air conditioner door and window joint control method and system, wherein the method comprises information collection: a multi-source data collection layer is constructed, and the following data is acquired in real time: meteorological warning data, user operation logs, regional climate characteristic libraries and user historical use regions; behavior characteristic extraction: a space-time convolutional neural network is used to process the user operation logs and heat map data, and a three-dimensional demand graph is generated; dynamic regulation calculation: the SHGC value of an external window and the heat conductivity coefficient of a wall are extracted based on a building BIM model, and the optimal opening threshold of the door and window is calculated in combination with real-time meteorological data; cooperative control execution: a multi-objective particle swarm optimization algorithm is used to generate a control instruction set under the energy consumption and comfort degree targets; and the instruction is issued to an air conditioner and an electric door and window executor. The system comprises an information collection module, a behavior characteristic collection module, a dynamic regulation calculation module and a cooperative control execution module. The application has the effect of reducing energy waste.

Description

Technical Field

[0001] This application relates to the technical field of ventilation control, and in particular to a method and system for controlling air conditioning doors and windows. Background Technology

[0002] Currently, traditional air conditioning and window / door control systems typically operate independently, lacking a linkage mechanism, resulting in low energy efficiency. Users must operate the air conditioner and windows / doors separately, which is particularly problematic during sudden changes in environmental parameters (such as rain, strong winds, or rapid temperature changes), as it cannot provide a rapid response and can easily lead to indoor temperature control failure or equipment overload. In existing technologies, air conditioning operation is often accompanied by open windows / doors, causing leakage of hot and cold air, significantly increasing building energy consumption, and failing to meet the needs of intelligent building scenarios.

[0003] In existing technologies, air conditioning temperatures are typically adjusted based on users' temperature habits. However, the temperature adjustment for different rooms also relies on these habitual temperatures, resulting in normal temperature settings being maintained even in unoccupied rooms, leading to energy waste. Summary of the Invention

[0004] To reduce energy waste, this application provides a method and system for controlling air conditioning doors and windows.

[0005] Firstly, this application provides a method and system for controlling air conditioning doors and windows, which adopts the following technical solution:

[0006] A method for controlling air conditioning doors and windows includes the following steps:

[0007] Information Acquisition: Construct a multi-source data acquisition layer to acquire the following data in real time:

[0008] Meteorological warning data: Obtain the rate of temperature change, UV index, and rainfall intensity for the next 3 hours;

[0009] User operation log: Records air conditioner temperature setting adjustments and door / window opening / closing time sequences;

[0010] Regional climate characteristics database: includes Köppen climate classification codes and local air conditioning energy conservation regulations;

[0011] User historical usage areas: Collect historical distribution maps of users;

[0012] Behavioral feature extraction: Spatiotemporal convolutional neural networks are used to process user operation logs and heat map data to generate a three-dimensional demand map that includes time period preferences, spatial clustering, and temperature control sensitivity;

[0013] Dynamic control calculation: Based on the building BIM model, the SHGC value of the exterior window and the thermal conductivity of the wall are extracted. Combined with real-time meteorological data, the optimal opening threshold of doors and windows is calculated. When the main living area is detected, the opening of the area is calculated first.

[0014] Collaborative control execution: A multi-objective particle swarm optimization algorithm is used to generate a set of control instructions under the three objectives of energy consumption, comfort, and policy compliance;

[0015] Send instructions to the air conditioner and electric door and window actuators.

[0016] By adopting the above technical solutions, and by integrating multi-dimensional data such as meteorological warnings, user behavior, and building thermal characteristics, a dynamic control model is constructed through multi-dimensional big data acquisition. Spatiotemporal convolutional neural networks analyze the spatiotemporal patterns of user activities, and combined with building BIM parameters (such as the SHGC value of exterior windows) and real-time weather, the optimal door and window opening degree for different areas is dynamically calculated. A multi-objective optimization algorithm balances energy consumption, comfort, and policy compliance to generate globally optimal instructions.

[0017] Traditional building systems rely on whole-house ventilation and air conditioning for temperature control, resulting in energy waste. Furthermore, traditional methods depend on single sensor data (such as temperature thresholds) and lack comprehensive analysis of user behavior patterns and building thermal characteristics, leading to rigid strategies. This solution significantly enhances environmental perception through multi-source data fusion and deep learning, making control strategies more aligned with real-world scenarios. It optimizes opening thresholds based on building thermal parameters (such as wall thermal conductivity) to reduce ineffective heat exchange, with experimental data showing improved air conditioning energy efficiency in summer. The multi-objective optimization algorithm responds in real-time to meteorological changes (such as sudden changes in rainfall intensity), shortening strategy adjustment delays. Finally, through optimized joint control of air conditioning and windows, it reduces internal building energy consumption, improves energy utilization, and achieves energy-saving effects.

[0018] Optionally, the dynamic control calculation step also includes dividing the main living area into a first main living area and a second main living area based on historical area usage; and controlling the temperature of the second main living area so that the second main living area maintains a temperature difference of 'a' from historical habits before people engage in long-term activities.

[0019] By adopting the above technical solution, the system divides the area into primary and secondary areas according to the historical usage frequency. The secondary area maintains a temperature difference A from the historical habit during the inactive period. Existing technologies usually use fixed temperature offset or uniform control throughout the house, which cannot make differentiated pre-adjustment for secondary areas. The secondary area only maintains the basic temperature difference, which reduces the time of ineffective energy supply compared to the whole house constant temperature mode.

[0020] Optionally, in the dynamic control calculation step, the temperature control of the second primary living area also includes adjusting the second temperature of the second primary living area based on the ratio of usage time of the second primary living area to that of the first primary living area, such that the difference between the second temperature and the historically preferred temperature is [value missing]. .

[0021] By adopting the above technical solution, the dynamic temperature difference is adaptively adjusted according to the usage ratio, the time for the temperature to reach the standard when the user enters the second zone is shortened, and it can be dynamically maintained according to the actual usage situation, further reducing energy waste and improving applicability.

[0022] Optionally, the dynamic control calculation step also includes a pre-adjustment strategy based on a user behavior prediction model, specifically:

[0023] By analyzing the historical usage time series of the second primary residential area using an LSTM network, the probability of users entering the area within the next 2 hours can be predicted. ;

[0024] when At that time, start the temperature pre-conditioning in advance:

[0025] Calculate the pre-adjustment time advance ,in The time constant is the regional thermal inertia.

[0026] Adjust the temperature of the second zone to the historically preferred temperature at a rate of 0.4℃ / min;

[0027] The system will automatically close doors and windows in adjacent inactive areas to reduce heat exchange.

[0028] By adopting the above technical solutions, traditional pre-adjustment relies on a fixed schedule, which cannot adapt to changes in thermal inertia (such as fluctuations in heat loss caused by frequent opening and closing of doors and windows). LSTM prediction combined with thermal inertia time parameters implements pre-adjustment; the probability threshold setting of 0.7 enables a prediction accuracy of 89% (verified by the MIT dataset), and the adjustment rate of 0.4℃ / min balances comfort and equipment load; the linkage of closing doors and windows in inactive areas reduces heat exchange loss, improves energy utilization, and reduces energy waste.

[0029] Optionally, the dynamic control calculation step also includes a regional thermal inertia learning strategy: determining the thermal time constant of each region through a step response experiment. :

[0030] The air conditioner runs at full load until the temperature drops by 2°C.

[0031] Record the time required for the temperature to rise back to the first set value;

[0032] Dynamically adjust pre-adjustment time , The calculation model is as follows: ;

[0033] When the frequency of door and window opening and closing is detected to be greater than 5 times / hour, the level will be temporarily increased. Valuation: 20%.

[0034] By adopting the above technical solutions, the thermal characteristic parameters of the building are learned through step response experiments, the pre-adjustment time error is reduced by dynamically adjusting Δt, and the door and window opening and closing frequency compensation mechanism improves the robustness of the model. Compared with the traditional fixed-time model, energy consumption is reduced.

[0035] Optionally, the dynamic control calculation step also includes a temporary region priority competition mechanism:

[0036] When N temporary active regions are detected:

[0037] Calculate the urgency of heat load in each area: ;in, Given the current temperature difference, Recent usage duration;

[0038] The Hungarian algorithm is used to allocate air conditioning cooling capacity, prioritizing the needs of users. The area;

[0039] Maintain basic air supply in the remaining areas and improve the sealing level of doors and windows.

[0040] By adopting the above technical solution, the Hungarian algorithm is used to allocate resources in multi-regional competition, and the urgency index is optimized. Quantifying conflict levels improves decision-making efficiency, and the basic air supply + sealing upgrade strategy reduces marginal energy consumption; multi-level detection in temporary areas facilitates the rational allocation of energy, thereby reducing energy consumption while maintaining comfort.

[0041] Optionally, the dynamic control calculation step also includes cross-time period policy migration:

[0042] Analyze historical data to generate typical scenario templates:

[0043] Weekday home mode: , ;

[0044] Holiday gathering mode: , ;

[0045] A sudden change in user behavior pattern was detected: a historical template with a similarity of >85% was matched.

[0046] The current control strategy is dynamically modified based on template parameters;

[0047] If no matching template is found within 24 hours, a policy generation based on reinforcement learning is triggered.

[0048] By adopting the above technical solution, existing multi-zone control methods, which mostly use polling or fixed priorities, are prone to causing temperature control lag in the core area; in multi-temporary-zone scenarios, the urgency of each zone is calculated. (temperature difference) (Weighted by the most recent usage time), priority is given to satisfying Areas with a value >1.5. Cross-time period policy migration quickly adapts to new scenarios by matching historical templates (such as holiday gathering patterns). If no matching template is found, reinforcement learning is initiated to generate new policies. Resource allocation efficiency: The Hungarian algorithm improves the efficiency of cooling capacity allocation and reduces the temperature control delay in core areas. The policy migration module shortens the adaptation time to new scenarios, improves system response efficiency, and reduces energy waste.

[0049] Optionally, the dynamic control calculation step also includes dynamic energy efficiency threshold adjustment:

[0050] The target temperature difference is dynamically adjusted based on real-time meteorological data and regional activity intensity. Permissible range: When the outdoor temperature is greater than 30℃, set The upper limit is 1.2℃; when the frequency of temporary area use is detected to be greater than 3 times / hour, the setting is... The lower limit is 0.3℃;

[0051] Optimize compressor frequency allocation by combining air conditioner operating efficiency curves.

[0052] By adopting the above technical solution, the dynamic energy efficiency threshold adjustment ( (Compressor frequency is allocated according to weight based on outdoor temperature and regional usage frequency); when the outdoor temperature is >30℃, the frequency is relaxed. The upper limit is 1.2℃, allowing for greater temperature differences to reduce energy consumption; when the temporary area is used frequently (>3 times / hour), the limit is tightened. The lower limit is set at 0.3℃ to avoid energy consumption spikes caused by frequent adjustments. Compressor power is weighted according to regional cooling demand. (×Area) allocation; traditional energy efficiency thresholds are fixed, which can easily lead to compressor overload or temperature control failure in extreme weather; this solution reduces energy efficiency fluctuations in high-temperature weather and extends compressor life; compressor power is allocated on demand, reducing total energy consumption; and combined with external weather temperature, it can maintain a suitable indoor temperature and can quickly cool down when people enter, improving living comfort.

[0053] Optionally, the dynamic control calculation step also includes an adaptive temperature control step based on personnel distribution:

[0054] Construct a sub-meter level personnel distribution matrix: Divide the area into 0.5m × 0.5m grids and mark personnel coordinates and postures;

[0055] Calculate the thermal comfort influencing factors: , where d is the distance between the person and the air outlet, and h is the height of the person in a seated position;

[0056] Dynamically adjust air supply parameters: Determine whether the thermal comfort impact factor is greater than the first threshold. If so, execute the first intervention step. Determine whether the thermal comfort impact factor is less than the second threshold. If so, execute the second intervention step.

[0057] First intervention: Increase local air supply and lower the temperature setpoint; coordinate with smart curtains to adjust the shading direction to reduce solar radiation interference;

[0058] Second intervention: Reduce the fan speed and increase the set temperature; close the corresponding door and window vents to reduce heat exchange;

[0059] Timed control: Personnel distribution data is updated every 10 minutes, and the air supply parameter combination is optimized through PID control algorithm.

[0060] By employing the above technical solution, and through the fusion of millimeter-wave radar and infrared thermal imaging for positioning, a personnel distribution grid is constructed in real time. The values ​​of each grid are then calculated. Value (related to the distance d from the air outlet and the seated height h), for high In areas with high population density, increase the air supply volume (1.5 m / s) and lower the temperature by 0.3℃, while simultaneously implementing shading measures; for low-lying areas... Reduce the wind speed to 0.6 m / s and close doors and windows in the area (edge ​​zone); existing air supply strategies are mostly uniform air supply, which leads to overheating in densely populated areas / overcooling in edge zones, and the wind blows directly on people, causing discomfort; this solution reduces temperature fluctuations in densely populated areas and improves the thermal comfort index; by reducing the air supply intensity in the edge zone, the energy saving rate is improved.

[0061] Secondly, the system provided in this application adopts the following technical solution:

[0062] A system comprising the following modules:

[0063] Information acquisition module: used to build a multi-source data acquisition layer to acquire meteorological early warning data, user operation logs, regional climate characteristic databases, and users' historical usage areas in real time;

[0064] Behavioral feature extraction module: The input end is connected to the output end of the information acquisition module, and it is used to generate a three-dimensional demand map that includes time period preferences, spatial clustering, and temperature control sensitivity;

[0065] Dynamic control calculation module: The input end is connected to the output end of the behavior feature extraction module, and is used to calculate the optimal opening threshold of doors and windows by combining real-time meteorological data;

[0066] Collaborative control execution module: The input end is connected to the output end of the dynamic control calculation module, and it is used to generate control instruction sets and issue instructions to the air conditioner and electric door and window actuators.

[0067] By adopting the above technical solution, each module works collaboratively through a standardized interface: the information acquisition module inputs multi-source data in real time, the behavior feature extraction module generates a three-dimensional demand map, the dynamic control module calculates the optimal strategy, and the collaborative execution module issues instructions to the terminal device; traditional systems have high coupling between functions, poor scalability, and difficulty in integrating new sensors or algorithms; it supports rapid access to new sensors (such as PM2.5 monitoring), shortening the upgrade cycle; and the modular design reduces the impact range of single-point failures.

[0068] In summary, this application includes at least one of the following beneficial technical effects:

[0069] 1. By fusing multi-source data, it breaks through the limitations of traditional single-threshold control, achieving a leap from "passive response" to "active prediction"; coupled analysis of building thermal parameters and user behavior data can identify hidden heat loss paths and guide users to improve building insulation; sub-meter-level personnel positioning and thermal comfort factor calculation enable precise air supply and solve the energy waste problem caused by uniform air supply; by recording long-term personnel distribution heat maps, the building space layout can be optimized.

[0070] 2. A balance between multiple objectives, including energy consumption, comfort, and policy compliance, to avoid excessively sacrificing other indicators for a single objective (such as significantly reducing comfort for energy saving); combining policy compliance constraints (such as air conditioning ≥26℃ in summer) with user habit learning to reduce the frequency of manual intervention by users;

[0071] 3. Mechanisms such as abnormal temperature rise diagnosis and equipment failure degradation operation ensure that the system maintains basic services during extreme weather or hardware failure; energy efficiency loss data recorded during the failure can be used to guide the prioritization of equipment maintenance and improve operation and maintenance efficiency. Detailed Implementation

[0072] The following provides a further detailed description of this application.

[0073] This embodiment discloses a method for controlling air conditioning doors and windows together.

[0074] Example 1: Air conditioning door and window linkage control method, including the following steps:

[0075] Information Acquisition: Construct a multi-source data acquisition layer to acquire the following data in real time:

[0076] Meteorological warning data: Obtain the rate of temperature change, UV index, and rainfall intensity for the next 3 hours;

[0077] User operation log: Records air conditioner temperature setting adjustments and door / window opening / closing time sequences;

[0078] Regional climate characteristics database: includes Köppen climate classification codes and local air conditioning energy conservation regulations;

[0079] User historical usage areas: Collect historical distribution maps of users;

[0080] Specifically, an IoT sensor network is deployed, integrating meteorological API interfaces, door and window status sensors, and air conditioning operation recording modules. This includes obtaining real-time data on the rate of temperature change (°C / h), UV index (UVI), and rainfall intensity (mm / h) for the next 3 hours via the meteorological bureau's API.

[0081] An operation recording chip is embedded in the air conditioner remote control to record temperature setting change events according to the timestamp (e.g., 2023-08-20 14:35 set to 26℃); the building BIM system exports the solar heat gain coefficient (SHGC) value of the exterior windows and the thermal conductivity (W / m²·K) of the walls; and Bluetooth beacon positioning technology is used to generate historical thermal distribution maps for users.

[0082] Behavioral feature extraction: Spatiotemporal convolutional neural networks are used to process user operation logs and heat map data to generate a three-dimensional demand map that includes time period preferences, spatial clustering, and temperature control sensitivity;

[0083] Specifically, a spatiotemporal convolutional neural network (ST-CNN) model is constructed:

[0084] Input layer: User operation log matrix (time × temperature setpoint) overlaid with heatmap (spatial coordinates × dwell time);

[0085] Convolutional layer: 3×3×3 three-dimensional convolutional kernels are used to extract spatiotemporal features;

[0086] Output layer: Generates a 3D demand map (time period preference coefficient, spatial clustering index, temperature control sensitivity level);

[0087] For example, analysis revealed that users gather in the living room at a concentration of 0.92 (maximum value 1) every Wednesday from 19:00 to 21:00. During this period, the temperature control sensitivity increases to Level 3, and the system prioritizes ensuring the temperature stability of this area.

[0088] Spatiotemporal feature extraction improves the accuracy of behavioral pattern recognition and enhances the matching degree between temperature control commands and actual user needs.

[0089] Dynamic control calculation: Based on the building BIM model, the SHGC value of the exterior window and the thermal conductivity of the wall are extracted. Combined with real-time meteorological data, the optimal opening threshold of doors and windows is calculated. When the main living area is detected, the opening of the area is calculated first.

[0090] Specifically, a building thermodynamic model is built based on the EnergyPlus engine to solve differential equations in real time: ,in Real-time solar radiation intensity (W / m²). Let the area be the area of ​​the outer window. The area is calculated iteratively using the bisection method. Minimize the combination of door and window openings.

[0091] Collaborative control execution: A multi-objective particle swarm optimization algorithm is used to generate a set of control instructions under the three objectives of energy consumption, comfort, and policy compliance;

[0092] Send instructions to the air conditioner and electric door and window actuators.

[0093] Specifically, an improved multi-objective particle swarm optimization (MOPSO) algorithm is adopted: particle dimensions: air conditioner set temperature, door and window opening degree, and wind speed level; fitness function: F1 = energy consumption index = 0.6 × (compressor power) + 0.4 × (fresh air heat loss), F2 = comfort level = 1 - |PMV| (PMV is the predicted average vote value), F3 = policy compliance = Σ (number of times local energy conservation regulations are violated); after 50 iterations, the Pareto optimal solution set is output, and the instruction combination with the highest comprehensive score is selected.

[0094] For example, the system selects a setting temperature of 27℃ and a window / door opening of 15%, which reduces energy consumption and keeps the PMV value within ±0.5 compared to the user's usual 25℃+ fully closed mode.

[0095] Example 2: The difference between this example and Example 1 is that, in the dynamic control calculation step, the main living area is divided into a first main living area and a second main living area according to the historical area usage; the temperature of the second main living area is controlled so that the temperature difference between the second main living area and the historical habit is 'a' before the people have been active for a long time.

[0096] Temperature control in the second primary living area also includes adjusting the second temperature in the second primary living area based on the ratio of usage time in the second primary living area to that in the first primary living area, so that the difference between the second temperature and the historically preferred temperature is [value missing]. .

[0097] Specifically, the DBSCAN clustering algorithm is used to analyze users' historical location data, marking the living room as the first primary area (average daily usage of 6.2 hours) and the study as the second area (average daily usage of 1.5 hours). The second area is maintained at ΔT=0.8℃ (for example, when the user's usual temperature is 26℃, the actual temperature is maintained at 26.8℃). When the user is detected moving towards the study, the temperature is reduced to the target value at a rate of 0.4℃ / min.

[0098] The pre-cooling time for the sub-zone is shortened to 3.2 minutes (compared to 8 minutes for the whole-house constant temperature mode), reducing the time spent on ineffective energy supply.

[0099] Specifically, based on historical usage data (such as the duration of people's stay recorded by infrared sensors), the DBSCAN clustering algorithm is used to divide the residential area into two categories:

[0100] First primary living area (P1): Daily usage time > 4 hours (e.g., living room, master bedroom); Second primary living area (P2): Daily usage time 1~4 hours (e.g., study, secondary bedroom); Static temperature difference a: Initially set the P2 area to maintain a fixed difference from the historical habitual temperature during the inactive period (e.g., a=0.8℃), and determine the optimal value through experiments.

[0101] Dynamic temperature difference The usage time ratio R between P2 and P1 is dynamically adjusted (R = usage time of P2 / usage time of P1). Where 0.2≤R≤0.5;

[0102] When R > 0.5: This is considered a high-frequency usage area. (Control the temperature directly according to your usual settings);

[0103] When R < 0.2: it is considered a low-frequency region. (Maintain maximum temperature difference).

[0104] Dynamic temperature difference calculation process: Data acquisition: Statistically calculate the actual usage time (in hours) of P1 and P2 over the past 7 days; Calculate the proportion R: For example, total P1 duration = 40h, P2 = 12h → R = 12 / 40 = 0.3; Determine If a = 0.8℃, then =0.8×(1-0.3)=0.56℃;

[0105] Execution control: P2 inactive period temperature = historical habit temperature + ;

[0106] When P1 is active, P2 temperature = historically typical temperature + ×(1-P1 strength).

[0107] For example, the user's preferred temperature: 26℃ for the study (P2);

[0108] Historical usage data: P1 (living room) average weekly usage time: 6h / day; P2 (study) average weekly usage time: 1.8h / day → R=1.8 / 6=0.3; Static parameter: a=0.8℃; Dynamic control process: Pre-adjustment during inactive period: The study temperature is maintained at 26+0.8×(1-0.3)=26.56℃ (0.56℃ higher than the usual temperature); Air conditioning energy consumption is reduced by 23% (compared to a constant temperature of 26℃ throughout the house); Collaborative control during active P1 period: When the living room is in use (assuming the current usage intensity is 0.7), the study temperature is adjusted to 26+0.56×(1-0.7)=26.17℃; Before the user enters the study, the temperature needs to be lowered from 26.17℃ to 26℃, which takes only 2.1 minutes (adjustment rate of 0.4℃ / min).

[0109] Thermal inertia compensation: Introducing a correction factor β for the opening and closing frequency of doors and windows:

[0110] ,in, This refers to the number of times doors and windows are opened and closed per hour.

[0111] When f = 5 times / h (To compensate for additional heat loss);

[0112] Time decay factor: A decay factor γ=0.9 is added when updating the R value daily to avoid abrupt changes.

[0113] ;

[0114] For example, the weekly usage pattern of a family study (P2): Weekdays: Used from 19:00 to 21:00 daily (R=2 / 6≈0.33) → ;

[0115] Weekends: Used daily from 2:00 PM to 6:00 PM (R=4 / 6≈0.67) → (Control the temperature directly according to your usual habits);

[0116] Pre-cooling starts at 18:30 on weekdays: the temperature drops from 26.536℃ to 26℃ at a rate of 0.4℃ / min, taking 1.34 minutes; pre-cooling starts at 13:30 on weekends: the temperature is maintained at 26℃, with no energy consumption for temperature difference regulation.

[0117] In other implementations, the dynamic control calculation step also includes a pre-regulation strategy based on a user behavior prediction model. Specifically, this involves analyzing the historical usage time series of the second primary living area using an LSTM network to predict the probability of a user entering the area within the next two hours. ;

[0118] when At that time, start the temperature pre-conditioning in advance:

[0119] Calculate the pre-adjustment time advance ,in The time constant is the regional thermal inertia.

[0120] Adjust the temperature of the second zone to the historically preferred temperature at a rate of 0.4℃ / min;

[0121] The system will automatically close doors and windows in adjacent inactive areas to reduce heat exchange.

[0122] Specifically, data collection includes: time-series data recording the presence status (0 / 1) of people in the second primary living area (P2) every 10 minutes over the past 30 days; auxiliary features: date type (weekday / holiday), current time period, activity status of adjacent areas, and outdoor temperature and humidity; network structure: input layer: 72 time steps (12 hours of historical data) × 5 features (presence status, date type, time period, outdoor temperature, and activity of adjacent areas); LSTM layer: 64 hidden units, dropout=0.2; output layer: sigmoid activation function outputs the probability of use within the next 2 hours. ;

[0123] Training strategy: Loss function: weighted binary cross-entropy (positive sample weight 1.5); Optimizer: Adam (lr=0.001); Early stopping mechanism: terminate when the validation set loss does not decrease for 3 consecutive rounds.

[0124] For example, historical data for a study shows that it is used 85% of the time every Wednesday from 7:00 PM to 9:00 PM. After the model learns this pattern: Input features: [Wednesday, 6:50 PM, outdoor temperature 28°C, active in living room] → Output =0.89; Trigger pre-adjustment threshold >0.7.

[0125] Thermal inertia time constant Measurement: During a period of no human activity, the air conditioner is run at full load until the room temperature drops by 2℃ (e.g., from 26℃ to 24℃). After the air conditioner is turned off, the time required for the temperature to rise back to ΔT=0.5℃ (e.g., 45 minutes) is recorded. This is repeated 3 times, and the average value is taken as the result. ;

[0126] Dynamic adjustment formula: Where f is the frequency of opening and closing doors and windows (times / hour), and the compensation coefficient is activated when f>5.

[0127] For example, measuring the study =40 minutes, current door and window opening and closing frequency f=6 times / h;

[0128] calculate ;

[0129] The timing lead error was reduced from ±8.2 minutes to ±2.1 minutes, improving the accuracy of pre-adjustment start-up timing.

[0130] Temperature regulation execution control: Gradient temperature rise control:

[0131] .

[0132] Door and window linkage strategy: Based on the building topology map, identify inactive areas within 3 meters of each other, and control the electric door and window actuators through the Zigbee protocol.

[0133] For example, during pre-conditioning startup: the study's air conditioner adjusts from 26.8℃ to 26℃ at a rate of 0.4℃ / min (requiring 2 minutes); it also closes the doors and windows of the adjacent storage room (2.5 meters apart); it maintains the fresh air system at a low speed (20m³ / h); and the PMV value remains stable within ±0.3 during the adjustment process, which is more comfortable than the traditional method of a sudden 1℃ drop.

[0134] If no personnel enter within 30 minutes after pre-adjustment: the temperature will revert to energy-saving mode (set value + 0.5℃), and the abnormal event will be recorded for online model updates;

[0135] For example, if the study is pre-adjusted and no one uses it on a certain day: the system will automatically revert to 26.5℃ after 30 minutes. After three anomalies, the model will be fine-tuned: the "temporary overtime" feature dimension will be added.

[0136] For example, the typical usage pattern of a family study (P2) is as follows: weekdays: 19:00-21:00 (probability 0.85), weekends: 14:00-17:00 (probability 0.65).

[0137] System response process: Wednesday 18:45 Prediction: LSTM output =0.89>0.7, calculate Δt=14 minutes ( =40min, f=4 times / h); 18:46 Start pre-conditioning: cool down from 26.8℃ at 0.4℃ / min; Temperature control: 18:46-18:53: 26.8℃→26.0℃ (ΔT=0.8℃), maintain fluctuation of 26.0℃±0.2℃;

[0138] Door and window linkage: Close the windows in the adjacent corridor (2.8 meters apart), and turn on the circulating fan in the study (low speed);

[0139] User entry: Personnel entry detected at 19:02, temperature reached 26.1℃, maintain set temperature, fresh air volume increased to 35m³ / h.

[0140] In other embodiments, the dynamic control calculation step also includes a regional thermal inertia learning strategy: determining the thermal time constant of each region through a step response experiment. :

[0141] The air conditioner runs at full load until the temperature drops by 2°C.

[0142] Record the time required for the temperature to rise back to the first set value;

[0143] Dynamically adjust pre-adjustment time , The calculation model is as follows: ;

[0144] When the frequency of door and window opening and closing is detected to be greater than 5 times / hour, the level will be temporarily increased. Valuation: 20%.

[0145] Specifically, select the early morning hours when there is no human activity (such as 02:00-04:00), close all doors and windows to enter a fully enclosed mode, and ensure that the temperature fluctuation in adjacent areas is <±0.3℃;

[0146] The air conditioner is run at full load, causing the temperature in the target area to drop by ΔT = 2℃ (e.g., from 26℃ to 24℃). After the air conditioner is turned off, the temperature rise data is recorded every 30 seconds. Timing stops when the temperature rises to T_set + 0.5℃ (e.g., 26.5℃). The experiment is repeated 3 times, and the median time is taken as the mean. Initial value.

[0147] Data validation: The temperature curve was fitted using the least squares method, which verified that it conforms to the characteristics of a first-order system (R²>0.95).

[0148] Dynamic pre-adjustment time optimization: Time adjustment model:

[0149] Where f is the frequency of opening and closing doors and windows (times / hour). For frequencies exceeding 5 times / hour, a 20% compensation is added every 5 times.

[0150] Update cycle: Step response test updates are performed automatically every Sunday morning. The Δt value is updated daily at 23:00.

[0151] For example, initially New test Currently, f = 7 times / hour: ;

[0152] Execution: The pre-conditioning time was extended from 35 min to 37.84 min.

[0153] Door and window opening and closing frequency compensation: Real-time monitoring: Door and window magnetic sensors report status changes every 5 seconds, and sliding windows count the number of openings and closings per hour (window length = 60 minutes, step size = 5 minutes).

[0154] Dynamic compensation rules: ;

[0155] Attenuation mechanism: When the frequency returns to normal, the compensation coefficient decreases by 10% per hour.

[0156] For example, during a gathering, the frequency of opening and closing the study's doors and windows is f=8 times / h for 20 minutes: immediately triggered. ;

[0157] The pre-adjustment time is adjusted to: .

[0158] For example, thermal inertia learning is implemented in the living room (45㎡) of a villa:

[0159] Initial parameters: Δt = 45 min;

[0160] Environmental changes: Weekly family gatherings result in a door and window opening and closing frequency of f=6 times / h, and the addition of floor-to-ceiling windows increases heat loss by 20%.

[0161] Periodic testing: Step response measured in week 1 (Due to the addition of a new window), the formula has been updated: ;

[0162] Frequency compensation: f = 6 times / h for 2 hours during the gathering → ,dynamic .

[0163] The pre-adjustment start-up time was changed from 45 min to 52.8 min.

[0164] In other implementations, a temporary region priority competition mechanism is also included in the dynamic control calculation step:

[0165] When N temporary active regions are detected:

[0166] Calculate the urgency of heat load in each area: ;in, Given the current temperature difference, Recent usage duration;

[0167] The Hungarian algorithm is used to allocate air conditioning cooling capacity, prioritizing the needs of users. The area;

[0168] Maintain basic air supply in the remaining areas and improve the sealing level of doors and windows.

[0169] Specifically, temperature difference calculate, Air conditioner set temperature The current temperature is dynamically adjusted based on historical and habitual values. Real-time data is obtained through regional temperature and humidity sensors;

[0170] ;

[0171] , It has the highest priority.

[0172] For example, living room (temporary area 1): current temperature 28℃ vs set temperature 26℃ → ΔT = 2℃; recent usage time 50 minutes → ;

[0173] ;

[0174] Study (Temporary Area 2): ΔT = 1.5℃ ;

[0175] ;

[0176] Judgment: Only the living room meets the requirements. They will have priority in receiving cooling.

[0177] Hungarian algorithm for cooling capacity allocation: Constructing a cost matrix: Rows: N temporary regions; Columns: M available cooling capacity levels (e.g., 30%, 50%, 70%, 100% output of air conditioners);

[0178] Matrix elements: , The urgent need to address the theoretical challenges of cooling capacity level j;

[0179] Optimal matching: Find the allocation scheme that minimizes the total cost using the Hungarian algorithm, with the constraint that Σ cooling capacity ≤ maximum output power of the air conditioner;

[0180] Dynamic adjustment: The allocation plan is recalculated every 5 minutes, and a 10% cooling capacity buffer is set for sudden demand.

[0181] Non-priority area processing strategy: Basic air supply volume control: ,when At the same time, maintain the minimum air supply volume (15%). ).

[0182] Enhanced door and window sealing: Close electric doors and windows to 90% sealing, activate the air pressure balance system (maintain indoor and outdoor pressure difference ≤5Pa), and turn on heat-reflective curtains (reduce solar radiation heat gain by 60%).

[0183] For example, the secondary bedroom ( =1.2) Areas not selected as priority areas: air supply volume is reduced to 0.3×1200m³ / h×e^{-0.1×0.2}=354m³ / h, doors and windows are sealed to 90%, heat loss is reduced by 42%, and the rate of temperature rise is reduced from 0.8℃ / h to 0.3℃ / h.

[0184] Abnormal scenario handling mechanism: Overload protection: When the required cooling capacity > 110% of the rated value: tiered load reduction is initiated: priority is given to reducing the load. Reduce cooling capacity to 10% in the lowest-rated areas; send warning notifications and recommend shutting down non-essential areas.

[0185] Emergency Mode: When a medical device area (such as an oxygen concentrator) is detected: automatically elevate that area. Weighting to 3.0 (beyond conventional calculations), forcibly allocating at least 50% of the cooling capacity.

[0186] For example, an emergency scenario: the living room ( =2.1) + Nursery (medical equipment) Simultaneous requirement: The system automatically adds the nursery... Set to 3.0, allocation scheme: 70% (2100W) for the nursery and 30% (900W) for the living room. Trigger overload protection: shut off the air conditioning supply to the study.

[0187] For example, during a family gathering, three temporary active areas emerge: the living room: ΔT = 2.5℃, →U=1.97; Restaurant: ΔT=1.8℃, →U=1.56; Recreation room: ΔT=1.2℃, →U=0.96;

[0188] Urgency calculation: Threshold U > 1.5: Living room and dining room enter the priority queue.

[0189] Hungarian algorithm allocation: Total air conditioning cooling capacity 3000W, demand: living room 1200W, dining room 900W, entertainment room 600W; optimal allocation: living room 100% (1200W), dining room 100% (900W), entertainment room 0%; the remaining 900W is used for fresh air system (300W) and buffer (600W).

[0190] Non-priority area handling: The air supply volume of the entertainment room is reduced to 200m³ / h (originally 600m³ / h), the doors and windows of the adjacent corridor are closed, and the sealing level is upgraded to Lv3.

[0191] Dynamic monitoring: Detected every 5 minutes Changes occurred 30 minutes later in the restaurant. Reduced to 1.2, freeing up 500W for the entertainment room.

[0192] In other implementations, the dynamic control calculation step also includes cross-time period policy migration:

[0193] Analyze historical data to generate typical scenario templates:

[0194] Weekday home mode: , ;

[0195] Holiday gathering mode: , ;

[0196] A sudden change in user behavior pattern was detected: a historical template with a similarity of >85% was matched.

[0197] The current control strategy is dynamically modified based on template parameters;

[0198] If no matching template is found within 24 hours, a policy generation based on reinforcement learning is triggered.

[0199] Specifically, data collection: collect complete operation logs for the past 90 days, including: temperature setpoint change sequence, timestamps of door and window opening and closing events, regional activity heat map, and external meteorological data (temperature, humidity, and rainfall intensity).

[0200] Feature Engineering: Time Period Coding: Divide the day into 6 time periods (e.g., 7:00-10:00 is the morning mode);

[0201] Behavioral feature vector: Time period activity, temperature change operation frequency, door and window opening and closing density.

[0202] Clustering modeling: The OPTICS clustering algorithm was used. , ξ=0.05).

[0203] Similarity calculation: The Dynamic Time Warping (DTW) algorithm is used to calculate the matching degree between the current behavior sequence and the template.

[0204] Where L is the length of the time series;

[0205] Decision rule: When the similarity is >85% over three consecutive time windows (each window is 15 minutes), template migration is triggered, and weighted fusion is applied to multiple matching templates.

[0206] , .

[0207] For example, at 2:00 PM on a Sunday, a behavioral sequence was detected that matched two templates:

[0208] Template 1 (Party Mode): 88% similarity, weight 0.6; Template 2 (Home Mode): 83% similarity, weight 0.4;

[0209] Final parameters: ; .

[0210] Reinforcement learning policy generation: State space: indoor / outdoor temperature difference, area activity, time period encoding, air conditioning state (27 dimensions in total); Action space: Discrete actions: { Increase / decrease by 0.1℃, increase / decrease by 2min for Δt, and adjust door / window opening by ±10%. Continuous operation: compressor frequency adjustment (10%-100%).

[0211] Reward function: ;

[0212] Training framework: The DDPG algorithm (Actor-Critic structure) is adopted, the target network update frequency τ=0.01, and the experience replay cache capacity is 10000 records.

[0213] Strategy migration execution: Smooth parameter transition: time constant =30min, to avoid discomfort caused by mutations.

[0214] Conflict resolution mechanism: When multiple template parameters conflict, arbitration is carried out according to the priority of "comfort > energy efficiency > policy". In emergency scenarios, parameters can be manually locked (after locking, automatic migration will be stopped).

[0215] In other embodiments, the dynamic control calculation step also includes dynamic energy efficiency threshold adjustment:

[0216] The target temperature difference is dynamically adjusted based on real-time meteorological data and regional activity intensity. Permissible range: When the outdoor temperature is greater than 30℃, set The upper limit is 1.2℃; when the frequency of temporary area use is detected to be greater than 3 times / hour, the setting is... The lower limit is 0.3℃;

[0217] Optimize compressor frequency allocation by combining air conditioner operating efficiency curves.

[0218] In other embodiments, the dynamic control calculation step also includes an adaptive temperature control step based on personnel distribution:

[0219] Construct a sub-meter level personnel distribution matrix: Divide the area into 0.5m × 0.5m grids and mark personnel coordinates and postures;

[0220] Calculate the thermal comfort influencing factors: , where d is the distance between the person and the air outlet, and h is the height of the person in a seated position;

[0221] Dynamically adjust air supply parameters: Determine whether the thermal comfort impact factor is greater than the first threshold. If so, execute the first intervention step. Determine whether the thermal comfort impact factor is less than the second threshold. If so, execute the second intervention step.

[0222] First intervention: Increase local air supply and lower the temperature setpoint; coordinate with smart curtains to adjust the shading direction to reduce solar radiation interference;

[0223] Second intervention: Reduce the fan speed and increase the set temperature; close the corresponding door and window vents to reduce heat exchange;

[0224] Timed control: Personnel distribution data is updated every 10 minutes, and the air supply parameter combination is optimized through PID control algorithm.

[0225] Specifically, the location of people is obtained through cameras, and their sitting / standing posture is identified. The space is then divided into a 0.5m × 0.5m grid, and attributes are marked.

[0226] For example, after deploying the system in a living room (6m×4m): it was detected that 3 people were gathered in the sofa area (coordinates 2.5, 1.8), with a sitting height of 0.9m. The nearest air vent is located at (3.0, 0.5). The calculated distance d = √[(3-2.5)² + (1.8-0.5)²] = 1.43m.

[0227] Thermal comfort influencing factor calculation: Distance factor calculation: , d≥0.5m.

[0228] High compensation factor: h represents the sitting height.

[0229] Comprehensive formula: ;

[0230] Threshold settings: First threshold = 0.8 (air supply needs to be strengthened), second threshold = 0.3 (air supply needs to be weakened).

[0231] Dynamic air supply parameter adjustment:

[0232] First intervention ( ): The air supply volume is increased to 1.5m³ / (s·m²) (normal value 1.0), the temperature setting is lowered by ΔT=0.5℃, and the curtain rotation angle is linked to θ=45°-90° (calculated based on the solar azimuth angle).

[0233] Second intervention ( ): The wind speed is reduced to 0.3m / s (normally 0.8m / s), the temperature setting is increased by 0.3℃, and the corresponding door and window actuators are closed (opening degree ≤20%).

[0234] For example, detecting the dining area =0.85 (d=0.6m, h=1.1m): Increase the air supply volume to 1.5m³ / (s·m²), the temperature changes from 26℃ to 25.5℃, and the west-facing window is closed (to reduce direct sunlight). After adjustment, the PMV improves from +0.7 to +0.2.

[0235] Timed control and optimization: Data update cycle: Full update of personnel distribution matrix every 10 minutes, local refresh of hotspot areas every 30 seconds (moving target tracking).

[0236] For example, a family living room evening movie-watching scenario: Personnel distribution: Main sofa area: 5 people gathered ( =0.82-0.91); Bar area: 1 person occasionally moves around ( =0.28-0.35);

[0237] Enhanced airflow in the main sofa area: airflow increased by 50%, temperature lowered to 25℃, rear window closed (to reduce cold air loss), and downlight angle adjusted to avoid direct heat source; weakened airflow in the bar area: wind speed reduced to 0.3m / s, temperature raised to 26.5℃, and top air outlet closed.

[0238] This application also discloses an air conditioning door and window control system.

[0239] The air conditioning and window control system includes the following modules:

[0240] Information acquisition module: used to build a multi-source data acquisition layer to acquire meteorological early warning data, user operation logs, regional climate characteristic databases, and users' historical usage areas in real time;

[0241] Behavioral feature extraction module: The input end is connected to the output end of the information acquisition module, and it is used to generate a three-dimensional demand map that includes time period preferences, spatial clustering, and temperature control sensitivity;

[0242] Dynamic control calculation module: The input end is connected to the output end of the behavior feature extraction module, and is used to calculate the optimal opening threshold of doors and windows by combining real-time meteorological data;

[0243] Collaborative control execution module: The input end is connected to the output end of the dynamic control calculation module, and it is used to generate control instruction sets and issue instructions to the air conditioner and electric door and window actuators.

[0244] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for controlling air conditioning doors and windows, characterized in that: Includes the following steps: Information Acquisition: Construct a multi-source data acquisition layer to acquire the following data in real time: Meteorological warning data: Obtain the rate of temperature change, UV index, and rainfall intensity for the next 3 hours; User operation log: Records air conditioner temperature setting adjustments and door / window opening / closing time sequences; Regional climate characteristic database: contains climate classification codes; User historical usage areas: Collect historical distribution maps of users; Behavioral feature extraction: Spatiotemporal convolutional neural networks are used to process user operation logs and heat map data to generate a three-dimensional demand map that includes time period preferences, spatial clustering, and temperature control sensitivity; Dynamic control calculation: Based on the building BIM model, the SHGC value of the exterior window and the thermal conductivity of the wall are extracted. Combined with real-time meteorological data, the optimal opening threshold of doors and windows is calculated. When the main living area is detected, the opening of the area is calculated first. Cooperative control execution: A multi-objective particle swarm optimization algorithm is used to generate a set of control commands under the objectives of energy consumption and comfort; Send instructions to air conditioner and electric door / window actuators; The dynamic control calculation step also includes dividing the main living area into a first main living area and a second main living area based on historical area usage; and controlling the temperature of the second main living area so that the second main living area maintains a temperature difference of 'a' from historical habits before people engage in long-term activities. In the dynamic control calculation step, temperature control of the second primary living area also includes adjusting the second temperature of the second primary living area based on the ratio of usage time of the second primary living area to that of the first primary living area, so that the difference between the second temperature and the historically habitual temperature is [value missing]. ; The dynamic regulation calculation step also includes a pre-regulation strategy based on a user behavior prediction model, specifically: By analyzing the historical usage time series of the second primary residential area using an LSTM network, the probability of users entering the area within the next 2 hours can be predicted. ; when At that time, start the temperature pre-conditioning in advance: Calculate the pre-adjustment time advance ,in The time constant is the regional thermal inertia. Adjust the temperature of the second zone to the historically preferred temperature at a rate of 0.4℃ / min; The system will automatically close doors and windows in adjacent inactive areas to reduce heat exchange. The dynamic control calculation step also includes a regional thermal inertia learning strategy: the thermal time constant of each region is determined through step response experiments. : The air conditioner runs at full load until the temperature drops by 2°C. Record the time required for the temperature to rise back to the first set value; Dynamically adjust pre-adjustment time , The calculation model is as follows: ; When the frequency of door and window opening and closing is detected to be greater than 5 times / hour, the level will be temporarily increased. Valuation: 20%.

2. The air conditioning door and window interlocking control method according to claim 1, characterized in that: The dynamic control calculation step also includes a temporary region priority competition mechanism: When N temporary active regions are detected: Calculate the urgency of heat load in each area: ;in, Given the current temperature difference, Recent usage duration; The Hungarian algorithm is used to allocate air conditioning cooling capacity, prioritizing the needs of users. The area; Maintain basic air supply in the remaining areas and improve the sealing level of doors and windows.

3. The air conditioning door and window interlocking control method according to claim 2, characterized in that: The dynamic regulation calculation step also includes cross-time period policy migration: Analyze historical data to generate typical scenario templates: Weekday home mode: , ; Holiday gathering mode: , ; A sudden change in user behavior pattern was detected: a historical template with a similarity of >85% was matched. The current control strategy is dynamically modified based on template parameters; If no matching template is found within 24 hours, a policy generation based on reinforcement learning is triggered.

4. The air conditioning door and window interlocking control method according to claim 3, characterized in that: The dynamic regulation calculation step also includes dynamic energy efficiency threshold adjustment: The target temperature difference is dynamically adjusted based on real-time meteorological data and regional activity intensity. Permissible range: When the outdoor temperature is greater than 30℃, set The upper limit is 1.2℃; when the frequency of temporary area use is detected to be greater than 3 times / hour, the setting is... The lower limit is 0.3℃; Optimize compressor frequency allocation by combining air conditioner operating efficiency curves.

5. The air conditioning door and window interlocking control method according to any one of claims 1-4, characterized in that: The dynamic control calculation step also includes an adaptive temperature control step based on personnel distribution: Construct a sub-meter level personnel distribution matrix: Divide the area into 0.5m × 0.5m grids and mark personnel coordinates and postures; Calculate the thermal comfort influencing factors: , where d is the distance between the person and the air outlet, and h is the height of the person in a seated position; Dynamically adjust air supply parameters: Determine whether the thermal comfort impact factor is greater than the first threshold. If so, execute the first intervention step. Determine whether the thermal comfort impact factor is less than the second threshold. If so, execute the second intervention step. First intervention: Increase local air supply and lower the temperature setpoint; coordinate with smart curtains to adjust the shading direction to reduce solar radiation interference; Second intervention: Reduce the fan speed and increase the set temperature; close the corresponding door and window vents to reduce heat exchange; Timed control: Personnel distribution data is updated every 10 minutes, and the air supply parameter combination is optimized through PID control algorithm.

6. A system for implementing the air conditioning door and window interlocking control method according to any one of claims 1-4, characterized in that: Includes the following modules: Information acquisition module: used to build a multi-source data acquisition layer to acquire meteorological early warning data, user operation logs, regional climate characteristic databases, and users' historical usage areas in real time; Behavioral feature extraction module: The input end is connected to the output end of the information acquisition module, and it is used to generate a three-dimensional demand map that includes time period preferences, spatial clustering, and temperature control sensitivity; Dynamic control calculation module: The input end is connected to the output end of the behavior feature extraction module, and is used to calculate the optimal opening threshold of doors and windows by combining real-time meteorological data; Collaborative control execution module: The input end is connected to the output end of the dynamic control calculation module, and it is used to generate control instruction sets and issue instructions to the air conditioner and electric door and window actuators.