Energy-saving and carbon-reducing refrigeration control method and system for building

Abstract

The invention provides a building energy-saving and carbon-reducing refrigeration control method and system in the technical field of building environment control and energy conservation. The method comprises the steps that S1, environment video is collected through a camera to recognize personnel information so as to generate activity factors; collecting carbon dioxide concentration through a carbon dioxide sensor to generate metabolic factors; s2, inputting the activity factors and the metabolic factors into the dynamic metabolic heat prediction model to obtain metabolic heat power per unit area; s3, air conditioner operation parameters are collected to correct the metabolic factors so as to dynamically update the unit area metabolic heat power, and the air conditioner operation parameters are adjusted based on the unit area metabolic heat power; s4, predicting a future heat load based on the unit area metabolic heat power; s5, controlling a cold source to execute cold energy storage based on the thermal load; and S6, obtaining carbon intensity emission data of the power grid to control the air conditioning system to work. The method has the advantages that the fineness, the accuracy, the timeliness and the energy efficiency ratio of refrigeration control are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of building environment control and energy-saving technology, and in particular to a building energy-saving and carbon-reducing refrigeration control method and system. Background Art

[0002] With the advancement of global carbon neutrality and building energy conservation policies, air conditioning systems, as a significant source of building energy consumption and carbon emissions, have become a key area of ​​research and engineering application for intelligent energy-saving control. Existing building cooling systems generally use timed and constant temperature control, or simple regional feedback adjustment methods based on temperature and humidity sensors. Some systems incorporate human presence sensors or reservation control functions to improve energy efficiency. However, existing methods have the following problems in actual operation: 1. Ignoring the dynamic contribution of human metabolic heat: As a heat source within a building, the metabolic heat of the human body varies significantly with the number of people and activity intensity. This is especially true in offices, classrooms, shopping malls, and other areas with frequent personnel movement. The impact of human metabolic heat on the cooling load cannot be ignored. Traditional air-conditioning systems are unable to perceive such load fluctuations in real time, resulting in a lagging control phenomenon of "insufficient cooling when many people are present and excessive cooling when few people are present," resulting in energy waste and reduced comfort.

[0003] 2. Carbon dioxide (CO2) concentration has limitations as a crowd indicator: Although CO2 concentration can reflect the density of people in a space, its changes are subject to physical lags and are easily affected by air conditioning wind speed, airflow path, equipment interference, etc. Existing CO2 sensing solutions mostly use single-point static sampling and lack disturbance compensation and multi-point fusion, resulting in significant measurement errors, further amplifying the judgment bias of the control system.

[0004] 3. Rigid air supply response strategies and prominent issues of uneven heating and cooling: Traditional air conditioning systems typically deliver air based on a single temperature setting. They lack a response mechanism to differences in heat load across different areas and are unable to adapt to dynamic changes in occupant distribution. This results in excess heating and cooling in some areas and insufficient comfort in others, impacting user experience and causing uneven energy consumption.

[0005] 4. Cooling load scheduling ignores fluctuations in electricity carbon intensity: Current cooling source control generally uses electricity prices or energy consumption priorities as the basis for start-stop strategies, lacks real-time consideration of the grid's carbon emission intensity, and continues to operate high-load equipment during carbon-intensive periods, which is not conducive to achieving carbon emission peak shaving and environmental goals.

[0006] 5. Delayed response to sudden crowd flow events. For example, the number of people changes suddenly when a meeting room ends or a classroom is replaced. Due to limitations in the sampling cycle or sensing mechanism, traditional air-conditioning systems find it difficult to identify and adjust the load response in a timely manner, resulting in an energy consumption redundancy problem of "wind chasing people".

[0007] Therefore, how to provide a building energy-saving and carbon-reducing refrigeration control method and system to improve the precision, accuracy, timeliness and energy efficiency of refrigeration control has become a technical problem that needs to be solved urgently. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a building energy-saving and carbon-reducing refrigeration control method and system, so as to improve the precision, accuracy, timeliness and energy efficiency of refrigeration control.

[0009] In a first aspect, the present invention provides a building energy-saving and carbon-reducing refrigeration control method, comprising the following steps: Step S1: collecting environmental video using cameras installed in each temperature-controlled area inside the building, identifying personnel information from the environmental video, and generating an activity factor based on the personnel information; collecting carbon dioxide concentration using carbon dioxide sensors installed in each temperature-controlled area inside the building, and generating a metabolic factor based on the carbon dioxide concentration; Step S2: inputting the activity factor and metabolic factor into a pre-trained dynamic metabolic heat prediction model to obtain the metabolic heat power per unit area of ​​each temperature control area; Step S3: collecting in real time the air conditioning operating parameters of the air conditioning system deployed in each temperature-controlled area inside the building, including air supply wind speed, wind direction angle, fan operating frequency, and indoor temperature gradient; correcting the metabolic factor based on each of the air conditioning operating parameters to dynamically update the metabolic heat power per unit area; and dynamically adjusting the air conditioning operating parameters based on each of the metabolic heat power per unit area; Step S4: predicting future heat load based on the metabolic heat power per unit area; Step S5: controlling the cooling source to perform cooling energy storage based on the predicted heat load; Step S6: acquiring carbon intensity emission data of the power grid in real time, and controlling the operation of the air conditioning system based on the carbon intensity emission data; The step S1 is specifically as follows: The edge computing device collects environmental videos in real time through cameras placed in various temperature-controlled areas inside the building. After performing data desensitization operations on the environmental videos, it uses inter-frame difference technology to identify the outlines of people from each environmental image frame in the environmental videos. Based on each of the outlines, it identifies personnel information including at least the number of people, movement trajectory, and activity frequency. Based on the personnel information, it calculates the activity factor. After the calculation is completed, the environmental videos are deleted: ; in, represents the activity factor; Indicates the total number of sampled frames of the environment image frame, with a value range of [10,30]; Indicates the frame number of the environment image frame; Indicates the The number of person silhouettes recognized in the frame; The edge computing device collects carbon dioxide concentrations through carbon dioxide sensors using a pressure differential drainage sampling structure, which are deployed in various temperature-controlled areas within the building. Based on this carbon dioxide concentration, a metabolic factor is generated: ; in, represents metabolic factors; express The carbon dioxide concentration at the time; express The carbon dioxide concentration at the time; Indicates a time interval.

[0010] Furthermore, in step S2, the dynamic metabolic heat prediction model is constructed based on a three-layer feedforward neural network, with two input layer nodes for inputting activity factors and metabolic factors, and one output layer node for outputting metabolic heat power per unit area. The number of hidden layer nodes ranges from [5, 10], and the activation function adopts ReLU. The dynamic metabolic heat prediction model has an online adaptive learning function and dynamically adjusts the model weight parameters based on the air conditioning return air temperature difference and regional energy consumption.

[0011] Furthermore, the step S3 is specifically as follows: The air conditioning system operating parameters of each temperature-controlled area in the building, including air supply speed, wind direction angle, fan operating frequency, and indoor temperature gradient, are collected in real time. The metabolic factor is corrected based on these air conditioning operating parameters: Calculate the perturbation factor: ; in, represents the disturbance factor; 、 、 、 Both represent the experience weight coefficient; Indicates the air supply speed; Indicates wind direction angle; Indicates the fan operating frequency; Indicates the indoor temperature gradient; When the disturbance factor is greater than a preset disturbance threshold, each of the metabolic factors is corrected: ; in, represents the corrected metabolic factor; represents metabolic factors; represents the disturbance correction coefficient, and its value range is [0.6, 0.9]; The metabolic heat powers per unit area are dynamically updated based on the corrected metabolic factors, and the air-conditioning operation parameters are dynamically adjusted based on the metabolic heat powers per unit area.

[0012] Furthermore, the step S4 is specifically as follows: Based on the metabolic heat power per unit area and the area of ​​the temperature control area, the total metabolic heat power of each temperature control area is calculated: ; in, represents the total metabolic heat power; It represents the metabolic heat power per unit area; Indicates the area of ​​the temperature control area; Using a sliding window algorithm combined with an exponentially weighted average method, the heat load for the next 15 to 30 minutes is predicted based on the total metabolic heat power, and the supply air temperature, supply air volume, supply air angle, and return air path of the air conditioning system are controlled based on the predicted heat load; When the heat load reaches the upper limit of the cooling capacity of the air-conditioning system, the temperature control parameters of each temperature control area are adjusted based on the preset air supply priority.

[0013] Furthermore, the step S5 is specifically as follows: A primary cold source, a secondary cold source, and a tertiary cold source are provided. The priority of the primary cold source is higher than that of the secondary cold source, and the priority of the secondary cold source is higher than that of the tertiary cold source. The primary cold source adopts a cold storage unit composed of a water-ice crystal mixed phase change material; the secondary cold source adopts a condensed water heat exchange device; and the tertiary cold source adopts air conditioning host cooling. Setting scheduling rules for the primary cooling source, the secondary cooling source, and the tertiary cooling source, and controlling the primary cooling source, the secondary cooling source, and the tertiary cooling source to perform cold energy storage and cold energy release based on the scheduling rules and the heat load; The step S6 is specifically as follows: The carbon intensity emission data of the power grid is obtained in real time, and the carbon intensity emission data is analyzed. When the carbon intensity is higher than a preset carbon intensity threshold, the cooling load of the compressor of the air-conditioning system is reduced, or the startup of high-energy-consuming equipment is delayed.

[0014] In a second aspect, the present invention provides a building energy-saving and carbon-reducing refrigeration control system, comprising the following modules: An activity metabolic factor generation module is configured to collect environmental video using cameras installed in various temperature-controlled areas within the building, identify personnel information from the environmental video, and generate activity factors based on the personnel information; collect carbon dioxide concentration using carbon dioxide sensors installed in various temperature-controlled areas within the building, and generate metabolic factors based on the carbon dioxide concentration; A metabolic heat power calculation module is used to input the activity factor and metabolic factor into a pre-trained dynamic metabolic heat prediction model to obtain the metabolic heat power per unit area of ​​each temperature control area; The metabolic heat power update module is used to collect in real time the air conditioning operating parameters of the air conditioning system deployed in each temperature-controlled area inside the building, including the air supply wind speed, wind direction angle, fan operating frequency, and indoor temperature gradient, and to correct the metabolic factor based on each of the air conditioning operating parameters to dynamically update the metabolic heat power per unit area, and to dynamically adjust the air conditioning operating parameters based on each of the metabolic heat power per unit area; a heat load prediction module, configured to predict future heat load based on the metabolic heat power per unit area; A cold energy storage module, configured to control a cold source to perform cold energy storage based on the predicted heat load; A cooling control module, configured to obtain carbon intensity emission data from the power grid in real time and control the operation of the air conditioning system based on the carbon intensity emission data; The activity metabolic factor generation module is specifically used for: The edge computing device collects environmental videos in real time through cameras placed in various temperature-controlled areas inside the building. After performing data desensitization operations on the environmental videos, it uses inter-frame difference technology to identify the outlines of people from each environmental image frame in the environmental videos. Based on each of the outlines, it identifies personnel information including at least the number of people, movement trajectory, and activity frequency. Based on the personnel information, it calculates the activity factor. After the calculation is completed, the environmental videos are deleted: ; in, represents the activity factor; Indicates the total number of sampled frames of the environment image frame, with a value range of [10,30]; Indicates the frame number of the environment image frame; Indicates the The number of person silhouettes recognized in the frame; The edge computing device collects carbon dioxide concentrations through carbon dioxide sensors using a pressure differential drainage sampling structure, which are deployed in various temperature-controlled areas within the building. Based on this carbon dioxide concentration, a metabolic factor is generated: ; in, represents metabolic factors; express The carbon dioxide concentration at the time; express The carbon dioxide concentration at the time; Indicates a time interval.

[0015] Furthermore, in the metabolic heat power calculation module, the dynamic metabolic heat prediction model is constructed based on a three-layer feedforward neural network, with two input layer nodes for inputting activity factors and metabolic factors, and one output layer node for outputting metabolic heat power per unit area. The number of hidden layer nodes ranges from [5, 10], and the activation function adopts ReLU. The dynamic metabolic heat prediction model has an online adaptive learning function and dynamically adjusts the model weight parameters based on the air conditioning return air temperature difference and regional energy consumption.

[0016] Furthermore, the metabolic heat power update module is specifically used to: The air conditioning system operating parameters of each temperature-controlled area in the building, including air supply speed, wind direction angle, fan operating frequency, and indoor temperature gradient, are collected in real time. The metabolic factor is corrected based on these air conditioning operating parameters: Calculate the perturbation factor: ; in, represents the disturbance factor; 、 、 、 Both represent the experience weight coefficient; Indicates the air supply speed; Indicates wind direction angle; Indicates the fan operating frequency; Indicates the indoor temperature gradient; When the disturbance factor is greater than a preset disturbance threshold, each of the metabolic factors is corrected: ; in, represents the corrected metabolic factor; represents metabolic factors; represents the disturbance correction coefficient, and its value range is [0.6, 0.9]; The metabolic heat powers per unit area are dynamically updated based on the corrected metabolic factors, and the air-conditioning operation parameters are dynamically adjusted based on the metabolic heat powers per unit area.

[0017] Furthermore, the heat load prediction module is specifically used to: Based on the metabolic heat power per unit area and the area of ​​the temperature control area, the total metabolic heat power of each temperature control area is calculated:

[0018] in, represents the total metabolic heat power; It represents the metabolic heat power per unit area; Indicates the area of ​​the temperature control area; Using a sliding window algorithm combined with an exponentially weighted average method, the heat load for the next 15 to 30 minutes is predicted based on the total metabolic heat power, and the supply air temperature, supply air volume, supply air angle, and return air path of the air conditioning system are controlled based on the predicted heat load; When the heat load reaches the upper limit of the cooling capacity of the air-conditioning system, the temperature control parameters of each temperature control area are adjusted based on the preset air supply priority.

[0019] Furthermore, the cold energy storage module is specifically used for: A primary cold source, a secondary cold source, and a tertiary cold source are provided. The priority of the primary cold source is higher than that of the secondary cold source, and the priority of the secondary cold source is higher than that of the tertiary cold source. The primary cold source adopts a cold storage unit composed of a water-ice crystal mixed phase change material; the secondary cold source adopts a condensed water heat exchange device; and the tertiary cold source adopts air conditioning host cooling. Setting scheduling rules for the primary cooling source, the secondary cooling source, and the tertiary cooling source, and controlling the primary cooling source, the secondary cooling source, and the tertiary cooling source to perform cold energy storage and cold energy release based on the scheduling rules and the heat load; The refrigeration control module is specifically used for: The carbon intensity emission data of the power grid is obtained in real time, and the carbon intensity emission data is analyzed. When the carbon intensity is higher than a preset carbon intensity threshold, the cooling load of the compressor of the air-conditioning system is reduced, or the startup of high-energy-consuming equipment is delayed.

[0020] The advantages of the present invention are: 1. Collect environmental videos through cameras installed in various temperature-controlled areas inside the building, identify personnel information from the environmental videos, and generate activity factors based on the personnel information; collect carbon dioxide concentrations through carbon dioxide sensors installed in various temperature-controlled areas inside the building, generate metabolic factors based on the carbon dioxide concentrations, input the activity factors and metabolic factors into the pre-trained dynamic metabolic heat prediction model to obtain the metabolic heat power per unit area of ​​each temperature-controlled area; then collect the air-conditioning system operating parameters of each temperature-controlled area inside the building, including air supply wind speed, wind direction angle, fan operating frequency, and indoor temperature gradient, in real time, correct the metabolic factors based on the air-conditioning operating parameters to dynamically update the metabolic heat power per unit area, and dynamically adjust the air-conditioning operating parameters based on the metabolic heat power per unit area; then predict the future heat load based on the metabolic heat power per unit area, control the cold source based on the predicted heat load to perform cold energy storage, and obtain the carbon intensity emission data of the power grid in real time. Control the operation of the air-conditioning system; that is, the comprehensive performance of refrigeration control is significantly improved by combining multi-source real-time perception with a dynamic metabolic heat prediction model: using cameras to identify personnel information in each temperature-controlled area in real time to generate activity factors, and synchronously integrating carbon dioxide concentration to generate metabolic factors, and accurately outputting metabolic heat power per unit area through a pre-trained dynamic metabolic heat prediction model, which solves the problem of traditional methods ignoring the dynamic heat contribution of the human body; based on the air-conditioning operating parameters, the metabolic factors are corrected in real time and the metabolic heat power per unit area is dynamically updated, and the air-conditioning operating parameters are adjusted in combination with zone differentiation to eliminate uneven heating and cooling; by predicting future heat loads, the cooling energy storage and carbon intensity emission data are linked to realize the spatiotemporal migration of the cooling load, realize fine control of the zones in the spatial dimension, eliminate response lag in the temporal dimension, and improve parameter accuracy through data fusion and model iteration to achieve the energy efficiency optimization goals of on-demand cooling and peak shaving, ultimately greatly improving the precision, accuracy, timeliness and energy efficiency of refrigeration control.

[0021] 2. By simultaneously utilizing visual information (camera recognition of the number of people, movement trajectories, and activity frequency) and air quality information (CO2 concentration change rate) to quantify human heat production (activity factor) and metabolic intensity (metabolic factor), this method can more accurately reflect the actual and dynamically changing heat load sources in the area than traditional methods that rely solely on fixed occupant density assumptions or single sensor data (such as only the number of people or only CO2). This lays a solid foundation for subsequent precise load forecasting and control, avoids over- or under-cooling, and effectively achieves energy-saving control.

[0022] 3. A dynamic metabolic heat prediction model using a relatively simple but effective three-layer feedforward neural network (5-10 nodes in the hidden layer, ReLU activation) maps activity factors and metabolic factors to metabolic heat power per unit area. It also has an online adaptive learning function and can dynamically adjust the model weight parameters based on the air conditioning return air temperature difference (reflecting the actual cooling effect) and regional energy consumption (reflecting actual demand). This allows the dynamic metabolic heat prediction model to not only predict the current heat load but also continuously learn about environmental changes and system characteristics, continuously self-optimize prediction accuracy, improve system robustness and long-term applicability, and enhance overall energy efficiency.

[0023] 4. By innovatively defining the disturbance factor, the system comprehensively considers the impact of the air conditioning system's own operating conditions, such as air supply speed, wind direction angle, fan operating frequency, and indoor temperature gradient, on the indoor environment (especially CO2 distribution and metabolic factor calculation). When the disturbance is large, the metabolic factor is corrected in real time, solving the problem of metabolic factor measurement distortion caused by changes in the air conditioning system's own operating conditions (such as strong winds blowing away CO2 accumulation areas and wind direction changes affecting airflow organization). This ensures that the prediction of metabolic heat power per unit area and subsequent control remain accurate and reliable under complex operating conditions, preventing control errors and energy waste caused by interference.

[0024] 5. The total metabolic heat power is obtained by combining the predicted metabolic heat power per unit area with the regional area, and the sliding window and exponential weighted average method are further used to predict the heat load for the next 15-30 minutes. Based on this predicted value, the supply air temperature, supply air volume, supply air angle and return air path are dynamically adjusted. A supply air priority adjustment mechanism is specially designed when the cooling capacity is limited to achieve predictive control, respond to load changes in advance, optimize airflow organization, and improve comfort. When the cooling capacity is limited, the priority strategy is used to ensure the comfort of key areas or the efficient operation of the entire system to avoid overload or local discomfort, thereby greatly improving the system efficiency and reliability.

[0025] 6. By setting up a primary cold source (cold storage unit), a secondary cold source (condensate heat exchange device), and a tertiary cold source (air conditioning host cooling), and formulating clear scheduling rules, the high-priority primary cold source takes advantage of its large phase change latent heat and can store cold during low electricity prices / carbon emissions; the secondary cold source recycles waste cold (condensate), which can maximize the utilization of high-efficiency, low-cost, and low-carbon cold sources (especially phase change cold storage and waste heat recovery), reduce dependence on direct cooling of high-energy consumption hosts (tertiary cold source), effectively "save" and "release" cold energy, achieve peak shaving and valley filling, and reduce operating costs and carbon emissions.

[0026] 7. By acquiring the carbon intensity emission data of the power grid in real time and using it as control input, when the carbon intensity is high (i.e., high carbon emissions from power generation), proactive measures are taken to reduce the cooling load of the air-conditioning compressor or delay the start-up of high-energy-consuming equipment. This directly links the operation of the building's air-conditioning system with the carbon emissions of the power grid, proactively reducing energy consumption or shifting loads during "high-carbon" periods of the power grid, and significantly reducing indirect carbon emissions from the building's cooling process.

[0027] 8. Process environmental videos through edge computing devices, perform desensitization operations and identify people's outlines, and then immediately delete the original environmental videos, reducing dependence on cloud bandwidth, improving real-time performance, and effectively protecting user privacy. It avoids the storage and transmission risks of sensitive video data, and improves system compliance and user acceptance.

[0028] 9. By integrating visually recognized personnel activity information with carbon dioxide concentration monitoring data, combining neural networks to dynamically predict metabolic heat power per unit area, and innovatively introducing a disturbance correction mechanism for air-conditioning operating parameters, the accuracy of cooling demand prediction has been significantly improved. On this basis, a multi-level cold source collaborative scheduling strategy (phase change cold storage, waste heat recovery, and host complementarity) and an active control mechanism based on the carbon intensity of the power grid are constructed to achieve real-time optimization of air-conditioning operating parameters and cold energy supply. While ensuring comfort, it effectively reduces building cooling energy consumption and carbon emissions, combining the privacy protection advantages of edge computing data desensitization with the engineering practicality of a clear parameter range. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0030] Figure 1 This is a flow chart of a building energy-saving and carbon-reducing refrigeration control method of the present invention.

[0031] Figure 2 It is a structural schematic diagram of a building energy-saving and carbon-reducing refrigeration control system according to the present invention. DETAILED DESCRIPTION

[0032] The technical solution in the embodiment of the present application has the following overall idea: the comprehensive performance of refrigeration control is significantly improved by combining multi-source real-time perception with a dynamic metabolic heat prediction model: the camera is used to identify the personnel information of each temperature control area in real time to generate an activity factor, and the carbon dioxide concentration is simultaneously integrated to generate a metabolic factor. The pre-trained dynamic metabolic heat prediction model accurately outputs the metabolic heat power per unit area, which solves the problem that the traditional method ignores the dynamic heat contribution of the human body; the metabolic factor is corrected in real time based on the air-conditioning operating parameters and the metabolic heat power per unit area is dynamically updated, and the air-conditioning operating parameters are adjusted in combination with the partition differentiation to eliminate the unevenness of hot and cold; by predicting the future heat load and linking the cold energy storage and carbon intensity emission data, the spatiotemporal migration of the refrigeration load is realized, the partition fine control is achieved in the spatial dimension, and the response lag is eliminated in the time dimension. The parameter accuracy is improved through data fusion and model iteration, and the energy efficiency optimization goals of on-demand cooling and peak shaving are achieved, thereby improving the precision, accuracy, timeliness and energy efficiency ratio of the refrigeration control.

[0033] Please refer to Figures 1 to 2 As shown, a preferred embodiment of a building energy-saving and carbon-reducing refrigeration control method of the present invention includes the following steps: Step S1: collecting environmental video using cameras installed in each temperature-controlled area inside the building, identifying personnel information from the environmental video, and generating an activity factor based on the personnel information; collecting carbon dioxide concentration using carbon dioxide sensors installed in each temperature-controlled area inside the building, and generating a metabolic factor based on the carbon dioxide concentration; Step S2: inputting the activity factor and metabolic factor into a pre-trained dynamic metabolic heat prediction model to obtain the metabolic heat power per unit area of ​​each temperature control area, which represents the heat load level of the population; Step S3: collecting in real time the air conditioning operating parameters of the air conditioning system deployed in each temperature-controlled area inside the building, including air supply wind speed, wind direction angle, fan operating frequency, and indoor temperature gradient; correcting the metabolic factor based on each of the air conditioning operating parameters to dynamically update the metabolic heat power per unit area; and dynamically adjusting the air conditioning operating parameters based on each of the metabolic heat power per unit area; Step S4: predicting future heat load based on the metabolic heat power per unit area; Step S5: controlling the cooling source to perform cooling energy storage based on the predicted heat load; Step S6: Real-time acquisition of carbon intensity emission data from the power grid, and control of the air conditioning system based on the carbon intensity emission data; in specific implementation, different cooling sources may be selected or the main cooling source may be delayed to achieve carbon peak-shifting operation based on the carbon intensity emission data; This invention proposes a unique dual-mode fusion modeling mechanism for human metabolic thermal perception. Combining fuzzy image recognition with CO2 concentration changes, it accurately extracts activity and metabolic factors representing the number of people and activity intensity, enabling real-time prediction of thermal load without violating privacy. Traditional systems generally rely on indoor temperature or manual settings for control, failing to perceive the dynamic thermal impact of a crowd in real time. However, this invention utilizes a low-resolution camera coupled with local edge computing to perform data desensitization and activity extraction. This, supplemented by a differential pressure-drained CO2 sampling structure, effectively avoids wind speed disturbance errors, enabling dynamic perception of human metabolic thermal load.

[0034] In response to the inherent defect of carbon dioxide sensors in responding slowly to instantaneous changes in crowd flow, the "entry and exit event perception trigger mechanism" was introduced for the first time. Through the analysis of short-term image sequences from the camera, sudden changes in crowd flow can be judged. Combined with the temporary sampling frequency increase and rapid load reconstruction strategy, it can achieve early response to typical emergency scenarios such as "meeting ending" and "class switching". This mechanism does not rely on access control or wearable devices. It is based on the coupled judgment of image dynamics and gas response. It has the advantages of simple implementation path, low deployment cost, and fast response speed, which greatly improves the real-time and accuracy of the system's thermal load judgment.

[0035] In terms of air supply response strategy, a dynamic zoned air supply system based on metabolic heat power per unit area was constructed, and "automatic migration of cold and heat" was realized by using adjustable air outlet structure and intelligent return air baffle. Compared with the rigid air supply method of "constant wind speed and constant temperature" in traditional air-conditioning systems, the distribution of air supply resources is adjusted by heat intensity sorting, so that the cold air automatically flows to areas with high metabolic heat load, ensuring cold and hot matching, effectively avoiding the lag phenomenon of "hot when there are many people and cold when people leave", improving comfort while significantly reducing ineffective energy consumption.

[0036] At the level of cooling source utilization and system scheduling, the linkage between the building energy system and the carbon intensity of the power grid is fully considered. This innovative approach incorporates condensed water waste heat exchange and ice crystal phase change material cold storage units. Through a staggered operation mechanism that stores cold at night and releases it during the day, the system dynamically adjusts compressor startup priorities based on real-time carbon intensity data, achieving the dual goals of "peak energy consumption shifting and carbon emission reduction." Traditional systems often ignore the temporal distribution characteristics of the power grid's carbon intensity and schedule cooling sources solely based on economic efficiency. This invention, however, incorporates "carbon priority" into cooling strategies, truly establishing a cooling source control logic that prioritizes "low carbon, economic efficiency, and comfort."

[0037] Combined with real-time recording of key data such as energy consumption, load, carbon emissions, and training optimization models, we continuously improve future operational efficiency and intelligent decision-making capabilities, forming a closed-loop intelligent system from perception-dispatching-execution-optimization. This system has strong synergy and can significantly reduce the system idling ratio, compressor start-up frequency and energy loss. It is suitable for various building scenarios such as office buildings, campuses, and transportation hubs, and has strong engineering feasibility and promotion value.

[0038] The present invention not only makes differentiated innovations in structure, but also makes systematic reinforcements in multiple seemingly small but critical "energy consumption loopholes", demonstrating the reconstruction of the entire link from perception mechanism, data processing, control response to energy scheduling. In particular, without relying on high-cost new materials or external large platforms, it realizes the multiple functions of "dynamic thermal matching", "event-level response" and "carbon priority scheduling" of the building air-conditioning system, and has substantial technical effects and significant energy-saving and carbon-reduction results that are difficult to achieve with traditional systems.

[0039] The step S1 is specifically as follows: The edge computing device collects real-time environmental video through cameras deployed in various temperature-controlled areas within the building. After performing data desensitization operations on the environmental video, it uses inter-frame difference technology to identify the outlines of people from each environmental image frame in the environmental video. Based on each of the human outlines, human information (fuzzy image information) including at least the number of people, movement trajectory, and activity frequency is identified. The activity factor is calculated based on the human information, and the environmental video is deleted after the calculation is completed: ; in, The activity factor per unit time is unitless and is used to reflect the activity level of people in the current area. The value range is: 0 (no activity) to 10 or above (high-frequency flow scene); Indicates the total number of sampled frames of the environment image frame, with a value range of [10,30]; Indicates the frame number of the environment image frame; Indicates the The number of person silhouettes recognized in the frame; In specific implementation, the camera uses a low-resolution camera with a resolution of 160×120 to 320×240 and a frame rate of no less than 5 frames per second, and is installed on the ceiling of the building; The edge computing device collects carbon dioxide concentrations through carbon dioxide sensors using a pressure differential drainage sampling structure, which are deployed in various temperature-controlled areas within the building. Based on this carbon dioxide concentration, a metabolic factor is generated: ; in, The metabolic factor per unit time, that is, the rate of change of carbon dioxide concentration per unit time, is expressed in ppm / min, with a value range of 5-60 ppm / min, and is used to reflect the metabolic intensity of the population; express The carbon dioxide concentration at the time; express The carbon dioxide concentration at the time; Indicates the time interval, in min.

[0040] The CO2 sensor introduces indoor air through a pressure-differential sampling mechanism. The recommended inlet air velocity is 0.1-0.3 m / s, the membrane pore size is 0.3 mm, and the path length is 100-300 mm. CO2 concentration is collected once a minute. To enhance representativeness, multiple CO2 sensors are deployed at different heights (0.5 m, 1.1 m, and 1.7 m) and in crowd hotspots. Their data is fused through a time synchronization mechanism to avoid bias caused by individual CO2 sensors due to turbulence or dilution. In specific implementation, concentration gradient differences, position weighting coefficients, and a temporal trend function are incorporated to generate a more representative average concentration value.

[0041] The pressure differential drainage sampling structure includes an external shell, a built-in microporous pressure-stabilizing air inlet membrane and a dual-channel air guide path. One end of the air guide path is connected to the indoor activity area, and the other end is equipped with a pressure equalization device and an electric micro air valve, which automatically adjusts the sampling speed to maintain a stable sampling environment and prevent wind speed fluctuations from affecting carbon dioxide concentration readings.

[0042] In specific implementation, in order to improve stability, a 3-5 point sliding average method can be used to smooth the .

[0043] In step S2, the dynamic metabolic heat prediction model is constructed based on a three-layer feedforward neural network, with two input layer nodes for inputting activity factors and metabolic factors, and one output layer node for outputting metabolic heat power per unit area. The number of hidden layer nodes ranges from [5 to 10], and the activation function adopts ReLU. The dynamic metabolic heat prediction model has an online adaptive learning function, which dynamically adjusts the model weight parameters based on the air conditioning return air temperature difference and regional energy consumption to adapt to the crowd behavior characteristics of a specific building space.

[0044] The unit of metabolic heat power per unit area is W / m 2 , represents the heat power generated by people per unit area. The output range is: ordinary office area: 20~35W / m 2 , areas with strong activities (such as fitness areas and large conference rooms): 35~70W / m 2; This result serves as the core indicator for determining the priority of regional cooling. In the next step, the system will dynamically adjust the supply air temperature, air volume and direction according to the metabolic heat power per unit area of ​​different regions, that is, guide the air supply strategy of each temperature-controlled area and realize the adaptive control strategy of "cooling follows heat".

[0045] The step S3 is specifically as follows: The air conditioning system operating parameters of each temperature-controlled area in the building, including air supply speed, wind direction angle, fan operating frequency, and indoor temperature gradient, are collected in real time. The metabolic factor is corrected based on these air conditioning operating parameters: Calculate the perturbation factor: ; in, Indicates the disturbance factor, that is, the disturbance intensity coefficient, in ppm / min; 、 、 、 Both represent the empirical weight coefficient, ranging from 0.1 to 1.0; Indicates the air supply speed in m / s, ranging from 0.5 to 3.0 m / s; Indicates the wind direction angle, in degrees, which refers to the angle of the air supply relative to the sensor axis; Indicates the fan operating frequency in Hz, reflecting the air volume; Indicates the indoor temperature gradient in °C / m, used to determine the local convection intensity; When the disturbance factor is greater than a preset disturbance threshold (e.g., 10 ppm / min), it indicates that the carbon dioxide signal is strongly interfered with and needs to be corrected. The metabolic factors are corrected: ; in, represents the corrected metabolic factor; represents metabolic factors; The disturbance correction coefficient is in the range of [0.6, 0.9]. This means that the metabolic factor is corrected through a real-time interpolation algorithm to reduce the measurement deviation caused by high wind dilution and turbulence. The metabolic heat output per unit area is dynamically updated based on the corrected metabolic factors, and air conditioning operating parameters are dynamically adjusted based on the metabolic heat output per unit area. The air conditioning system features at least one electrically operated air vent structure with adjustable air outlet angle, wind speed, and flow rate. This system prioritizes air supply and redistributes cooling and heating capacity based on differences in the metabolic heat output per unit area across various zones. Furthermore, a "cooling follows hot zone" response mode is implemented through a series-controlled intelligent baffle structure for the return air vents.

[0046] The step S4 is specifically as follows: Based on the metabolic heat power per unit area and the area of ​​the temperature control area, the total metabolic heat power of each temperature control area is calculated: ; in, Indicates total metabolic heat power, unit W; It represents metabolic heat power per unit area in W / m²; Indicates the area of ​​the temperature control area, unit: m 2 ; Using a sliding window algorithm combined with an exponentially weighted mean average method (EVVMA), the heat load for the next 15 to 30 minutes is predicted based on the total metabolic heat power, and the supply air temperature, supply air volume, supply air angle, and return air path of the air conditioning system are controlled based on the predicted heat load; When the heat load reaches the cooling capacity limit of the air conditioning system, the temperature control parameters of each temperature-controlled zone are adjusted based on the preset air supply priority. In specific implementation, an optional LSTM lightweight network model can be used to improve accuracy, which is particularly suitable for highly dynamic scenarios such as conference venues and offices.

[0047] Air temperature control: The higher the heat load of the target temperature control area, the lower the supply air temperature, and the recommended control range is 17~22℃; Supply air volume control: Air is distributed according to the heat load ratio, and the maximum air volume is controlled at 80~100% of the rated air volume; Supply air angle adjustment: Automatically adjust the horizontal / vertical air outlet angle according to the distribution of personnel and obstruction conditions to ensure that the cold air reaches the heat source concentration area directly; Return air path control: By adjusting the variable baffle structure, the hot return air is guided to pass through the high-load area first, thereby improving the system's heat exchange efficiency.

[0048] When the heat load reaches the cooling capacity limit of the air conditioning system, air supply is prioritized based on the metabolic heat level corresponding to the metabolic heat power per unit area of ​​each area. Cooling is first supplied to densely populated areas with high metabolic heat, reducing cooling energy waste. At the same time, some air outlets in low-load areas are switched to minimum maintenance mode (only to maintain temperature stability) to minimize cooling energy input to non-target areas.

[0049] The step S5 is specifically as follows: A primary cold source, a secondary cold source, and a tertiary cold source are provided. The priority of the primary cold source is higher than that of the secondary cold source, and the priority of the secondary cold source is higher than that of the tertiary cold source. The primary cold source adopts a cold storage unit composed of a water-ice crystal mixed phase change material. The secondary cold source adopts a condensed water heat exchange device. The tertiary cold source adopts air conditioning host cooling. The cold storage unit accumulates cold energy during the night when electricity prices are low and carbon intensity is low, and releases cold energy during the day or during periods of higher carbon intensity to replace part of the host cooling, thereby reducing the total carbon emissions of the system. Set the scheduling rules for the primary cold source, secondary cold source, and tertiary cold source, and control the primary cold source, secondary cold source, and tertiary cold source to perform cold energy storage and cold energy release based on the scheduling rules and the heat load. The cold storage temperature range of the primary cold source is -1°C to 5°C; it can store energy during low electricity price periods or low carbon intensity periods at night; no additional power consumption is required when releasing cold energy, and the carbon footprint is extremely low; the secondary cold source recovers the condensed water discharged from the air conditioning system for indirect heat exchange, can provide refrigerant-side cooling at 5 - 8°C, and assists in maintaining the supply air temperature, which is suitable for the low-load stability stage and has good economy and environmental protection; the tertiary cold source is a compression refrigeration system, with high power consumption but rapid response; it is only called when the cold storage is insufficient or during peak load periods, and is a backup plan with a relatively high carbon intensity.

[0050] The scheduling rule adopts the strategy of "energy storage at night + release during the day", specifically as follows: Cold energy storage stage (at night or when the carbon intensity < 400 gCO2 / kWh): Start the chiller, transfer the cold energy to the cold storage unit, and control the outlet temperature of the cold storage unit to be 3 - 5°C; at the same time, introduce the condensed water into the heat exchange pipeline of the condensed water heat exchange device to pre-cool the condensed water heat exchange device. Cold energy release stage (during the day or when the carbon intensity > 600 gCO2 / kWh): Shut down the chiller, and switch the cold energy source to the cold storage unit or the condensed water heat exchange device; the cold storage flow is automatically adjusted according to the heat load change, and the cooling capacity per unit area is controlled at 80 - 120 W / m2. The scheduling priority follows the principle of "low carbon first + use cold storage first + economic and controllable": 1. Read the current carbon intensity Cg, the available capacity of cold storage Qs, and the currently predicted heat load ; 2. If Cg < 400 and Qs < Qmax: Start the chiller for refrigeration to store cold. 3. If Cg > 600: Pause the chiller refrigeration, and preferentially enable the cold storage unit and the condensed water heat exchange device. 4. If > Qs + Qc: Start the chiller for refrigeration to operate in a way to make up the difference. 5. Record the operating duration and energy consumption of each type of cold source throughout the process for optimizing subsequent scheduling strategies.

[0051] Among them, Cg represents the real-time grid carbon intensity (unit: gCO2 / kWh); Qs represents the current remaining cold storage capacity (unit: kWh); Qc represents the maximum cold load capacity of the condensed water heat exchange device (unit: kW); Qmax represents the maximum capacity of the cold storage unit (unit: kWh); The specific content of step S6 is as follows: Real-time carbon intensity emission data from the power grid is acquired and analyzed. When the carbon intensity exceeds a preset threshold, the air conditioning system's compressor cooling load is reduced or the startup of high-energy-consuming equipment is delayed. In specific implementation, this data, combined with weather forecasts, electricity load curves, and cold storage status, is used to establish a cooling source priority decision map. When the carbon intensity exceeds the set threshold, the system automatically reduces the compressor cooling load and transfers the primary cooling task to low-carbon cooling sources, forming a multi-objective control strategy combining energy conservation, comfort, and low carbon.

[0052] During specific implementation, the current electricity carbon intensity is obtained in real time through the data interface with the power operation platform or regional energy consumption management center. (Carbon intensity emission data), the unit is gCO2 / kWh, and the refresh frequency is no less than once every 5 minutes. To enhance the forecasting ability, it also connects to the meteorological platform and the historical carbon emission database to build a short-term trend forecast model and output the carbon intensity estimate within the next 1 hour. .

[0053] The carbon intensity is divided into three operating levels and the control strategy is preset: Green zone (low carbon): The host cooling and cold storage operate synchronously, and can actively store cold in this range; Yellow zone (medium carbon): The priority of the cooling source is increased to the cold storage and condensing water systems, and the main engine cooling is limited; Red zone (high carbon): Reduce or shut down the host cooling, reduce the load by ≥30%, and temporarily suspend the operation of high-energy-consuming equipment.

[0054] The carbon intensity threshold can be flexibly configured according to regional policies or carbon trading platform requirements, and carbon intensity control can be incorporated into the cooling load forecasting and scheduling model to form a multi-objective optimization logic: Goal 1: Keep the indoor temperature in the comfort zone (such as 24±1°C); Goal 2: Minimize compressor refrigeration carbon emissions; Goal 3: Prioritize the consumption of cold storage resources to avoid carbon emissions peaking; Goal 4: Limit non-essential regional cooling during high-carbon periods.

[0055] The optimization model uses piecewise linear programming to ensure a dynamic balance between carbon constraints and comfort. In actual operation, scheduling switches are automatically performed according to the following logic: 1. Read the current time , carbon intensity , total load forecast value ; 2. If High threshold; ; 3. If the carbon intensity peak is predicted to occur within 15 to 30 minutes, start the cold storage and cooling process in advance; 4. Switch the air conditioner in non-core areas to the maintenance mode (such as air volume , temperature control is relaxed); 5. Record carbon emission curves throughout the entire cycle and gradually train the control model based on the user's energy consumption targets; This mechanism can be combined with the carbon credit feedback mechanism to incentivize carbon emission reduction behaviors and help meet green building assessment standards.

[0056] If carbon intensity data is interrupted or there's an unexpected load surge, the system automatically reverts to the default control logic to ensure cooling continuity and comfort. The host load limit, temperature and humidity tolerance range, and cold storage discharge rate all serve as system protection parameters in the scheduling optimization process.

[0057] In addition, the edge computing device also integrates the following core subroutines: Trajectory lag buffer control subroutine: When a person stays in a certain area for more than 180 seconds and then leaves quickly (<30 seconds), it is judged based on historical path information that the person's feeling of cold has not yet fully appeared. The air supply parameters in this area are maintained for 90 seconds to delay the attenuation of cooling capacity, thereby avoiding the "people getting cold when the wind stops" caused by the lag in cooling sensation.

[0058] Contour change detection subroutine: Calculates the rate of change of person contours in consecutive image frames in real time and compares it with a set threshold ΔM; ΔM is a system calibration value set to 8% of the total number of image contour pixels; if the contour variation rate ΔM exceeds the threshold within a unit time (the default is 3 seconds), it indicates that multiple people may have entered or exited the current area quickly (such as when a meeting starts or ends).

[0059] When ΔM>threshold, the fast sampling window Ws is triggered: The sampling period of the CO2 detection unit is shortened from the normal 60 seconds to 10 seconds; Over the next 30 seconds, take 3 to 5 consecutive CO2 concentration readings; The metabolic factors were corrected for a short time using a local second-order curve fitting method; This method effectively captures sudden load changes caused by large influxes of people, improving the system's adaptability to dynamic human traffic.

[0060] In addition, the relative pixel ratio R=W_px / H_px of a person's shoulder width and height is extracted by blurring the camera image. This ratio does not involve face or identity recognition and still complies with the privacy protection design.

[0061] The preset body segmentation and cooling load correction rules are as follows: R ≥ 0.5, it is judged as "large body size", and its unit metabolic heat estimation is increased by 10%; 0.35≤R<0.5, judged as "medium body type" and no adjustment is made; R<0.35, judged as "small size", and its valuation was reduced by 8%.

[0062] After counting the average body size classification in each air supply area, the metabolic heat power per unit area of ​​the area is weighted and corrected to generate Qh_corr, which is used as the basis for cooling capacity scheduling and air outlet control.

[0063] First, by deploying low-resolution blurry cameras in conjunction with edge computing devices, real-time extraction of the number of people, activity intensity, and movement trajectory is achieved to form a basic activity factor. Unlike traditional solutions that rely solely on CO2 concentration changes and sampling lag, which prevents the timely detection of instantaneous crowd entry and exit, this innovative method embeds a contour change detection subroutine within the image processing process to determine the degree of rapid movement of person boundaries between image frames. When the image contour change ΔM per unit time exceeds the set threshold, it indicates that a large number of people may have entered or exited the current area. The rapid sampling window Ws is triggered, automatically shortening the CO2 sampling period from the conventional 60 seconds to 10 seconds. The metabolic factor is then corrected using a short-term second-order fitting algorithm to rapidly update the metabolic heat power per unit area. This mechanism effectively addresses the problem of traditional systems' delayed response to short-term fluctuations in crowd flow, ensuring that cooling scheduling is synchronized with the actual heat load of the crowd, significantly improving energy efficiency and response sensitivity in dynamic scenarios.

[0064] Considering that the human body has a physiological response lag of 2 to 5 minutes to temperature changes, the traditional system's "regulate as soon as data changes" strategy often results in misaligned regulation when "the person has left the hot zone but the wind has just started," causing the person to overshoot the cold sensation and reduce comfort. This invention proposes a mechanism to compensate for the delayed response of the human body to cold sensation. Based on the individual path trajectory and historical heat exposure time recorded by the camera, a "path thermal memory map" is constructed in each area. A "delayed release strategy" is introduced during regulation. That is, even if a person has just left the high-heat zone, the air supply stability is still maintained for a certain period of time, forming a dynamic "slow cooling" regulation mode. This ensures a high degree of rhythmic matching between the human body's physiological cold sensation and the cold source regulation, avoiding energy waste while enhancing physical comfort.

[0065] To overcome the drawbacks of uniformly modeling everyone and ignoring individual metabolic differences, a fuzzy image-based body shape estimation and fine-tuning control mechanism is proposed. Specifically, edge computing devices extract a person's "height-to-shoulder width ratio" when capturing image outlines, and based on this, classify individuals into different body types (such as thin, average, and overweight). A "body shape correction factor" Qh_corr is then introduced and weightedly fused with the basal metabolic heat Qh to fine-tune the air supply intensity. This method, without collecting any sensitive image data or introducing complex recognition algorithms, enables precise adjustment of cooling capacity in scenarios where people of different body shapes gather, significantly improving cooling accuracy and energy-saving potential.

[0066] A preferred embodiment of a building energy-saving and carbon-reducing refrigeration control system of the present invention includes the following modules: An activity metabolic factor generation module is configured to collect environmental video using cameras installed in various temperature-controlled areas within the building, identify personnel information from the environmental video, and generate activity factors based on the personnel information; collect carbon dioxide concentration using carbon dioxide sensors installed in various temperature-controlled areas within the building, and generate metabolic factors based on the carbon dioxide concentration; A metabolic heat power calculation module is used to input the activity factor and metabolic factor into a pre-trained dynamic metabolic heat prediction model to obtain the metabolic heat power per unit area of ​​each temperature control area, indicating the heat load level of the population; The metabolic heat power update module is used to collect in real time the air conditioning operating parameters of the air conditioning system deployed in each temperature-controlled area inside the building, including the air supply wind speed, wind direction angle, fan operating frequency, and indoor temperature gradient, and to correct the metabolic factor based on each of the air conditioning operating parameters to dynamically update the metabolic heat power per unit area, and to dynamically adjust the air conditioning operating parameters based on each of the metabolic heat power per unit area; a heat load prediction module, configured to predict future heat load based on the metabolic heat power per unit area; A cold energy storage module, configured to control a cold source to perform cold energy storage based on the predicted heat load; A cooling control module is used to obtain real-time carbon intensity emission data from the power grid and control the operation of the air conditioning system based on the carbon intensity emission data. In specific implementation, different cooling sources can be selected or the main cooling source can be delayed to achieve carbon peak operation based on the carbon intensity emission data; This invention proposes a unique dual-mode fusion modeling mechanism for human metabolic thermal perception. Combining fuzzy image recognition with CO2 concentration changes, it accurately extracts activity and metabolic factors representing the number of people and activity intensity, enabling real-time prediction of thermal load without violating privacy. Traditional systems generally rely on indoor temperature or manual settings for control, failing to perceive the dynamic thermal impact of a crowd in real time. However, this invention utilizes a low-resolution camera coupled with local edge computing to perform data desensitization and activity extraction. This, supplemented by a differential pressure-drained CO2 sampling structure, effectively avoids wind speed disturbance errors, enabling dynamic perception of human metabolic thermal load.

[0067] In response to the inherent defect of carbon dioxide sensors in responding slowly to instantaneous changes in crowd flow, the "entry and exit event perception trigger mechanism" was introduced for the first time. Through the analysis of short-term image sequences from the camera, sudden changes in crowd flow can be judged. Combined with the temporary sampling frequency increase and rapid load reconstruction strategy, it can achieve early response to typical emergency scenarios such as "meeting ending" and "class switching". This mechanism does not rely on access control or wearable devices. It is based on the coupled judgment of image dynamics and gas response. It has the advantages of simple implementation path, low deployment cost, and fast response speed, which greatly improves the real-time and accuracy of the system's thermal load judgment.

[0068] In terms of air supply response strategy, a dynamic zoned air supply system based on metabolic heat power per unit area was constructed, and "automatic migration of cold and heat" was realized by using adjustable air outlet structure and intelligent return air baffle. Compared with the rigid air supply method of "constant wind speed and constant temperature" in traditional air-conditioning systems, the distribution of air supply resources is adjusted by heat intensity sorting, so that the cold air automatically flows to areas with high metabolic heat load, ensuring cold and hot matching, effectively avoiding the lag phenomenon of "hot when there are many people and cold when people leave", improving comfort while significantly reducing ineffective energy consumption.

[0069] At the level of cooling source utilization and system scheduling, the linkage between the building energy system and the carbon intensity of the power grid is fully considered. This innovative approach incorporates condensed water waste heat exchange and ice crystal phase change material cold storage units. Through a staggered operation mechanism that stores cold at night and releases it during the day, the system dynamically adjusts compressor startup priorities based on real-time carbon intensity data, achieving the dual goals of "peak energy consumption shifting and carbon emission reduction." Traditional systems often ignore the temporal distribution characteristics of the power grid's carbon intensity and schedule cooling sources solely based on economic efficiency. This invention, however, incorporates "carbon priority" into cooling strategies, truly establishing a cooling source control logic that prioritizes "low carbon, economic efficiency, and comfort."

[0070] Combined with real-time recording of key data such as energy consumption, load, carbon emissions, and training optimization models, we continuously improve future operational efficiency and intelligent decision-making capabilities, forming a closed-loop intelligent system from perception-dispatching-execution-optimization. This system has strong synergy and can significantly reduce the system idling ratio, compressor start-up frequency and energy loss. It is suitable for various building scenarios such as office buildings, campuses, and transportation hubs, and has strong engineering feasibility and promotion value.

[0071] The present invention not only makes differentiated innovations in structure, but also makes systematic reinforcements in multiple seemingly small but critical "energy consumption loopholes", demonstrating the reconstruction of the entire link from perception mechanism, data processing, control response to energy scheduling. In particular, without relying on high-cost new materials or external large platforms, it realizes the multiple functions of "dynamic thermal matching", "event-level response" and "carbon priority scheduling" of the building air-conditioning system, and has substantial technical effects and significant energy-saving and carbon-reduction results that are difficult to achieve with traditional systems.

[0072] The activity metabolic factor generation module is specifically used for: The edge computing device collects real-time environmental video through cameras deployed in various temperature-controlled areas within the building. After performing data desensitization operations on the environmental video, it uses inter-frame difference technology to identify the outlines of people from each environmental image frame in the environmental video. Based on each of the human outlines, human information (fuzzy image information) including at least the number of people, movement trajectory, and activity frequency is identified. The activity factor is calculated based on the human information, and the environmental video is deleted after the calculation is completed: ; in, The activity factor per unit time is unitless and is used to reflect the activity level of people in the current area. The value range is: 0 (no activity) to 10 or above (high-frequency flow scene); Indicates the total number of sampled frames of the environment image frame, with a value range of [10,30]; Indicates the frame number of the environment image frame; Indicates the The number of person silhouettes recognized in the frame; In specific implementation, the camera uses a low-resolution camera with a resolution of 160×120 to 320×240 and a frame rate of no less than 5 frames per second, and is installed on the ceiling of the building; The edge computing device collects carbon dioxide concentrations through carbon dioxide sensors using a pressure differential drainage sampling structure, which are deployed in various temperature-controlled areas within the building. Based on this carbon dioxide concentration, a metabolic factor is generated: ; in, The metabolic factor per unit time, that is, the rate of change of carbon dioxide concentration per unit time, is expressed in ppm / min, with a value range of 5-60 ppm / min, and is used to reflect the metabolic intensity of the population; express The carbon dioxide concentration at the time; express The carbon dioxide concentration at the time; Indicates the time interval, in min.

[0073] The CO2 sensor introduces indoor air through a pressure-differential sampling mechanism. The recommended inlet air velocity is 0.1-0.3 m / s, the membrane pore size is 0.3 mm, and the path length is 100-300 mm. CO2 concentration is collected once a minute. To enhance representativeness, multiple CO2 sensors are deployed at different heights (0.5 m, 1.1 m, and 1.7 m) and in crowd hotspots. Their data is fused through a time synchronization mechanism to avoid bias caused by individual CO2 sensors due to turbulence or dilution. In specific implementation, concentration gradient differences, position weighting coefficients, and a temporal trend function are incorporated to generate a more representative average concentration value.

[0074] The pressure differential drainage sampling structure includes an external shell, a built-in microporous pressure-stabilizing air inlet membrane and a dual-channel air guide path. One end of the air guide path is connected to the indoor activity area, and the other end is equipped with a pressure equalization device and an electric micro air valve, which automatically adjusts the sampling speed to maintain a stable sampling environment and prevent wind speed fluctuations from affecting carbon dioxide concentration readings.

[0075] In specific implementation, in order to improve stability, a 3-5 point sliding average method can be used to smooth the .

[0076] In the metabolic heat power calculation module, the dynamic metabolic heat prediction model is constructed based on a three-layer feedforward neural network, with two input layer nodes for inputting activity factors and metabolic factors, and one output layer node for outputting metabolic heat power per unit area. The number of hidden layer nodes ranges from [5 to 10], and the activation function adopts ReLU. The dynamic metabolic heat prediction model has an online adaptive learning function, which dynamically adjusts the model weight parameters based on the air conditioning return air temperature difference and regional energy consumption to adapt to the crowd behavior characteristics of a specific building space.

[0077] The unit of metabolic heat power per unit area is W / m 2 , represents the heat power generated by people per unit area. The output range is: ordinary office area: 20~35W / m 2 , areas with strong activities (such as fitness areas and large conference rooms): 35~70W / m 2; This result serves as the core indicator for determining the priority of regional cooling. In the next step, the system will dynamically adjust the supply air temperature, air volume and direction according to the metabolic heat power per unit area of ​​different regions, that is, guide the air supply strategy of each temperature-controlled area and realize the adaptive control strategy of "cooling follows heat".

[0078] The metabolic heat power update module is specifically used for: The air conditioning system operating parameters of each temperature-controlled area in the building, including air supply speed, wind direction angle, fan operating frequency, and indoor temperature gradient, are collected in real time. The metabolic factor is corrected based on these air conditioning operating parameters: Calculate the perturbation factor: ; in, Indicates the disturbance factor, that is, the disturbance intensity coefficient, in ppm / min; 、 、 、 Both represent the empirical weight coefficient, ranging from 0.1 to 1.0; Indicates the air supply speed in m / s, ranging from 0.5 to 3.0 m / s; Indicates the wind direction angle, in degrees, which refers to the angle of the air supply relative to the sensor axis; Indicates the fan operating frequency in Hz, reflecting the air volume; Indicates the indoor temperature gradient in °C / m, used to determine the local convection intensity; When the disturbance factor is greater than a preset disturbance threshold (e.g., 10 ppm / min), it indicates that the carbon dioxide signal is strongly interfered with and needs to be corrected. The metabolic factors are corrected: ; in, represents the corrected metabolic factor; represents metabolic factors; The disturbance correction coefficient is in the range of [0.6, 0.9]. This means that the metabolic factor is corrected through a real-time interpolation algorithm to reduce the measurement deviation caused by high wind dilution and turbulence. The metabolic heat output per unit area is dynamically updated based on the corrected metabolic factors, and air conditioning operating parameters are dynamically adjusted based on the metabolic heat output per unit area. The air conditioning system features at least one electrically operated air vent structure with adjustable air outlet angle, wind speed, and flow rate. This system prioritizes air supply and redistributes cooling and heating capacity based on differences in the metabolic heat output per unit area across various zones. Furthermore, a "cooling follows hot zone" response mode is achieved through a series-controlled intelligent baffle structure for return air vents.

[0079] The heat load prediction module is specifically used for: Based on the metabolic heat power per unit area and the area of ​​the temperature control area, the total metabolic heat power of each temperature control area is calculated: ; in, Indicates total metabolic heat power, unit W; It represents metabolic heat power per unit area in W / m²; Indicates the area of ​​the temperature control area, unit: m 2 ; Using a sliding window algorithm combined with an exponentially weighted mean average method (EVVMA), the heat load for the next 15 to 30 minutes is predicted based on the total metabolic heat power, and the supply air temperature, supply air volume, supply air angle, and return air path of the air conditioning system are controlled based on the predicted heat load; When the heat load reaches the cooling capacity limit of the air conditioning system, the temperature control parameters of each temperature-controlled zone are adjusted based on the preset air supply priority. In specific implementation, an optional LSTM lightweight network model can be used to improve accuracy, which is particularly suitable for highly dynamic scenarios such as conference venues and offices.

[0080] Air temperature control: The higher the heat load of the target temperature control area, the lower the supply air temperature, and the recommended control range is 17~22℃; Supply air volume control: Air is distributed according to the heat load ratio, and the maximum air volume is controlled at 80~100% of the rated air volume; Supply air angle adjustment: Automatically adjust the horizontal / vertical air outlet angle according to the distribution of personnel and obstruction conditions to ensure that the cold air reaches the heat source concentration area directly; Return air path control: By adjusting the variable baffle structure, the hot return air is guided to pass through the high-load area first, thereby improving the system's heat exchange efficiency.

[0081] When the heat load reaches the cooling capacity limit of the air conditioning system, air supply is prioritized based on the metabolic heat level corresponding to the metabolic heat power per unit area of ​​each area. Cooling is first supplied to densely populated areas with high metabolic heat, reducing cooling energy waste. At the same time, some air outlets in low-load areas are switched to minimum maintenance mode (only to maintain temperature stability) to minimize cooling energy input to non-target areas.

[0082] The cold energy storage module is specifically used for: A primary cold source, a secondary cold source, and a tertiary cold source are provided. The priority of the primary cold source is higher than that of the secondary cold source, and the priority of the secondary cold source is higher than that of the tertiary cold source. The primary cold source adopts a cold storage unit composed of a water-ice crystal mixed phase change material. The secondary cold source adopts a condensed water heat exchange device. The tertiary cold source adopts air conditioning host cooling. The cold storage unit accumulates cold energy during the night when electricity prices are low and carbon intensity is low, and releases cold energy during the day or during periods of higher carbon intensity to replace part of the host cooling, thereby reducing the total carbon emissions of the system. Set the scheduling rules for the primary cold source, secondary cold source, and tertiary cold source, and control the primary cold source, secondary cold source, and tertiary cold source to perform cold energy storage and cold energy release based on the scheduling rules and heat load; The cold storage temperature range of the primary cold source is -1°C to 5°C; it can store energy during low electricity price periods or low carbon intensity periods at night; no additional power consumption is required when releasing cold energy, and the carbon footprint is extremely low; the secondary cold source recovers the condensate water discharged from the air conditioning system for indirect heat exchange, can provide refrigerant-side cooling at 5 - 8°C, and helps maintain the supply air temperature, which is suitable for the low-load stability maintenance stage and has good economy and environmental protection; the tertiary cold source is a compression refrigeration system, which has high power consumption but rapid response; it is only called when the cold storage is insufficient or during peak load periods, and is a backup plan with a relatively high carbon intensity.

[0083] The scheduling rule adopts the strategy of "energy storage at night + release during the day", which is specifically as follows: Cold energy storage stage (at night or when the carbon intensity < 400 gCO2 / kWh): Start the chiller, transfer the cold energy to the cold storage unit, and control the outlet temperature of the cold storage unit to be 3 - 5°C; at the same time, introduce the condensate water into the heat exchange pipeline of the condensate water heat exchange device to pre-cool the condensate water heat exchange device; Cold energy release stage (during the day or when the carbon intensity > 600 gCO2 / kWh): Shut down the chiller, and switch the cold energy source to the cold storage unit or the condensate water heat exchange device; the cold storage flow is automatically adjusted according to the heat load change, and the cooling capacity per unit area is controlled at 80 - 120 W / m2; The scheduling priority follows the principle of "low carbon first + use cold storage first + economic and controllable": 1. Read the current carbon intensity Cg, the available capacity of cold storage Qs, and the currently predicted heat load ; 2. If Cg < 400 and Qs < Qmax: Start the chiller to refrigerate for cold storage; 3. If Cg > 600: Pause the chiller refrigeration, and preferentially enable the cold storage unit and the condensate water heat exchange device; 4. If > Qs + Qc: Start the chiller to refrigerate and operate in a way to make up the difference; 5. Record the operation duration and energy consumption of each type of cold source throughout the process for optimizing subsequent scheduling strategies.

[0084] Among them, Cg represents the real-time grid carbon intensity (unit: gCO2 / kWh); Qs represents the current remaining cold storage capacity (unit: kWh); Qc represents the maximum cold load capacity of the condensate water heat exchange device (unit: kW); Qmax represents the maximum capacity of the cold storage unit (unit: kWh); The refrigeration control module is specifically used for: Real-time carbon intensity emission data from the power grid is acquired and analyzed. When the carbon intensity exceeds a preset threshold, the air conditioning system's compressor cooling load is reduced or the startup of high-energy-consuming equipment is delayed. In specific implementation, this data, combined with weather forecasts, electricity load curves, and cold storage status, is used to establish a cooling source priority decision map. When the carbon intensity exceeds the set threshold, the system automatically reduces the compressor cooling load and transfers the primary cooling task to low-carbon cooling sources, forming a multi-objective control strategy combining energy conservation, comfort, and low carbon.

[0085] During specific implementation, the current electricity carbon intensity is obtained in real time through the data interface with the power operation platform or regional energy consumption management center. (Carbon intensity emission data), the unit is gCO2 / kWh, and the refresh frequency is no less than once every 5 minutes. To enhance the forecasting ability, it also connects to the meteorological platform and the historical carbon emission database to build a short-term trend forecast model and output the carbon intensity estimate within the next 1 hour. .

[0086] The carbon intensity is divided into three operating levels and the control strategy is preset: Green zone (low carbon): The host cooling and cold storage operate synchronously, and can actively store cold in this range; Yellow zone (medium carbon): The priority of the cooling source is increased to the cold storage and condensing water systems, and the main engine cooling is limited; Red zone (high carbon): Reduce or shut down the host cooling, reduce the load by ≥30%, and temporarily suspend the operation of high-energy-consuming equipment.

[0087] The carbon intensity threshold can be flexibly configured according to regional policies or carbon trading platform requirements, and carbon intensity control can be incorporated into the cooling load forecasting and scheduling model to form a multi-objective optimization logic: Goal 1: Keep the indoor temperature in the comfort zone (such as 24±1°C); Goal 2: Minimize compressor refrigeration carbon emissions; Goal 3: Prioritize the consumption of cold storage resources to avoid carbon emissions peaking; Goal 4: Limit non-essential regional cooling during high-carbon periods.

[0088] The optimization model uses piecewise linear programming to ensure a dynamic balance between carbon constraints and comfort. In actual operation, scheduling switches are automatically performed according to the following logic: 1. Read the current time , carbon intensity , total load forecast value ; 2. If High threshold, ; 3. If the carbon intensity peak is predicted to occur within 15 to 30 minutes, start the cold storage and cooling process in advance; 4. Switch the air conditioner in non-core areas to the maintenance mode (such as air volume , temperature control is relaxed); 5. Record carbon emission curves throughout the entire cycle and gradually train the control model based on the user's energy consumption targets; This mechanism can be combined with the carbon credit feedback mechanism to incentivize carbon emission reduction behaviors and help meet green building assessment standards.

[0089] This mechanism can be combined with the carbon credit feedback mechanism to incentivize carbon emission reduction behaviors and help meet green building assessment standards.

[0090] If carbon intensity data is interrupted or there's an unexpected load surge, the system automatically reverts to the default control logic to ensure cooling continuity and comfort. The host load limit, temperature and humidity tolerance range, and cold storage discharge rate all serve as system protection parameters in the scheduling optimization process.

[0091] In addition, the edge computing device also integrates the following core subroutines: Trajectory lag buffer control subroutine: When a person stays in a certain area for more than 180 seconds and then leaves quickly (<30 seconds), it is judged based on historical path information that the person's feeling of cold has not yet fully appeared. The air supply parameters in this area are maintained for 90 seconds to delay the attenuation of cooling capacity, thereby avoiding the "people getting cold when the wind stops" caused by the lag in cooling sensation.

[0092] Contour change detection subroutine: Calculates the rate of change of person contours in consecutive image frames in real time and compares it with a set threshold ΔM; ΔM is a system calibration value set to 8% of the total number of image contour pixels; if the contour variation rate ΔM exceeds the threshold within a unit time (the default is 3 seconds), it indicates that multiple people may have entered or exited the current area quickly (such as when a meeting starts or ends).

[0093] When ΔM>threshold, the fast sampling window Ws is triggered: The sampling period of the CO2 detection unit is shortened from the normal 60 seconds to 10 seconds; Over the next 30 seconds, take 3 to 5 consecutive CO2 concentration readings; The metabolic factors were corrected for a short time using a local second-order curve fitting method; This method effectively captures sudden load changes caused by large influxes of people, improving the system's adaptability to dynamic human traffic.

[0094] In addition, the relative pixel ratio R=W_px / H_px of a person's shoulder width and height is extracted by blurring the camera image. This ratio does not involve face or identity recognition and still complies with the privacy protection design.

[0095] The preset body segmentation and cooling load correction rules are as follows: R ≥ 0.5, it is judged as "large body size", and its unit metabolic heat estimation is increased by 10%; 0.35≤R<0.5, judged as "medium body type" and no adjustment is made; R<0.35, judged as "small size", and its valuation was reduced by 8%.

[0096] After counting the average body size classification in each air supply area, the metabolic heat power per unit area of ​​the area is weighted and corrected to generate Qh_corr, which is used as the basis for cooling capacity scheduling and air outlet control.

[0097] First, by deploying low-resolution blurry cameras in conjunction with edge computing devices, real-time extraction of the number of people, activity intensity, and movement trajectory is achieved to form a basic activity factor. Unlike traditional solutions that rely solely on CO2 concentration changes and sampling lag, which prevents the timely detection of instantaneous crowd entry and exit, this innovative method embeds a contour change detection subroutine within the image processing process to determine the degree of rapid movement of person boundaries between image frames. When the image contour change ΔM per unit time exceeds the set threshold, it indicates that a large number of people may have entered or exited the current area. The rapid sampling window Ws is triggered, automatically shortening the CO2 sampling period from the conventional 60 seconds to 10 seconds. The metabolic factor is then corrected using a short-term second-order fitting algorithm to rapidly update the metabolic heat power per unit area. This mechanism effectively addresses the problem of traditional systems' delayed response to short-term fluctuations in crowd flow, ensuring that cooling scheduling is synchronized with the actual heat load of the crowd, significantly improving energy efficiency and response sensitivity in dynamic scenarios.

[0098] Considering that the human body has a physiological response lag of 2 to 5 minutes to temperature changes, the traditional system's "regulate as soon as data changes" strategy often results in misaligned regulation when "the person has left the hot zone but the wind has just started," causing the person to overshoot the cold sensation and reduce comfort. This invention proposes a mechanism to compensate for the delayed response of the human body to cold sensation. Based on the individual path trajectory and historical heat exposure time recorded by the camera, a "path thermal memory map" is constructed in each area. A "delayed release strategy" is introduced during regulation. That is, even if a person has just left the high-heat zone, the air supply stability is still maintained for a certain period of time, forming a dynamic "slow cooling" regulation mode. This ensures a high degree of rhythmic matching between the human body's physiological cold sensation and the cold source regulation, avoiding energy waste while enhancing physical comfort.

[0099] To overcome the drawbacks of uniformly modeling everyone and ignoring individual metabolic differences, a fuzzy image-based body shape estimation and fine-tuning control mechanism is proposed. Specifically, edge computing devices extract a person's "height-to-shoulder width ratio" when capturing image outlines, and based on this, classify individuals into different body types (such as thin, average, and overweight). A "body shape correction factor" Qh_corr is then introduced and weightedly fused with the basal metabolic heat Qh to fine-tune the air supply intensity. This method, without collecting any sensitive image data or introducing complex recognition algorithms, enables precise adjustment of cooling capacity in scenarios where people of different body shapes gather, significantly improving cooling accuracy and energy-saving potential.

[0100] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A building energy-saving and carbon-reducing refrigeration control method, characterized by: The steps include: Step S1: collecting environmental video using cameras installed in each temperature-controlled area inside the building, identifying personnel information from the environmental video, and generating an activity factor based on the personnel information; collecting carbon dioxide concentration using carbon dioxide sensors installed in each temperature-controlled area inside the building, and generating a metabolic factor based on the carbon dioxide concentration; Step S2: inputting the activity factor and metabolic factor into a pre-trained dynamic metabolic heat prediction model to obtain the metabolic heat power per unit area of ​​each temperature control area; Step S3: collecting in real time the air conditioning operating parameters of the air conditioning system deployed in each temperature-controlled area inside the building, including air supply wind speed, wind direction angle, fan operating frequency, and indoor temperature gradient; correcting the metabolic factor based on each of the air conditioning operating parameters to dynamically update the metabolic heat power per unit area; and dynamically adjusting the air conditioning operating parameters based on each of the metabolic heat power per unit area; Step S4: predicting future heat load based on the metabolic heat power per unit area; Step S5: controlling the cooling source to perform cooling energy storage based on the predicted heat load; Step S6: acquiring carbon intensity emission data of the power grid in real time, and controlling the operation of the air conditioning system based on the carbon intensity emission data; The step S1 is specifically as follows: The edge computing device collects environmental videos in real time through cameras placed in various temperature-controlled areas inside the building. After performing data desensitization operations on the environmental videos, it uses inter-frame difference technology to identify the outlines of people from each environmental image frame in the environmental videos. Based on each of the outlines, it identifies personnel information including at least the number of people, movement trajectory, and activity frequency. Based on the personnel information, it calculates the activity factor. After the calculation is completed, the environmental videos are deleted: ; in, represents the activity factor; Indicates the total number of sampled frames of the environment image frame, with a value range of [10,30]; Indicates the frame number of the environment image frame; Indicates the The number of person silhouettes recognized in the frame; The edge computing device collects carbon dioxide concentrations through carbon dioxide sensors using a pressure differential drainage sampling structure, which are deployed in various temperature-controlled areas within the building. Based on this carbon dioxide concentration, a metabolic factor is generated: ; in, represents metabolic factors; express The carbon dioxide concentration at the time; express The carbon dioxide concentration at the time; Indicates a time interval.

2. A building energy-saving and carbon-reducing refrigeration control method according to claim 1, characterized in that: In step S2, the dynamic metabolic heat prediction model is constructed based on a three-layer feedforward neural network, with two input layer nodes for inputting activity factors and metabolic factors, and one output layer node for outputting metabolic heat power per unit area. The number of hidden layer nodes ranges from [5 to 10], and the activation function adopts ReLU. The dynamic metabolic heat prediction model has an online adaptive learning function and dynamically adjusts the model weight parameters based on the air conditioning return air temperature difference and regional energy consumption.

3. The building energy-saving and carbon-reducing refrigeration control method according to claim 1, characterized in that: The step S3 is specifically as follows: The air conditioning system operating parameters of each temperature-controlled area in the building, including air supply speed, wind direction angle, fan operating frequency, and indoor temperature gradient, are collected in real time. The metabolic factor is corrected based on these air conditioning operating parameters: Calculate the perturbation factor: ; in, represents the disturbance factor; 、 、 、 Both represent the experience weight coefficient; Indicates the air supply speed; Indicates wind direction angle; Indicates the fan operating frequency; Indicates the indoor temperature gradient; When the disturbance factor is greater than a preset disturbance threshold, each of the metabolic factors is corrected: ; in, represents the corrected metabolic factor; represents metabolic factors; represents the disturbance correction coefficient, and its value range is [0.6, 0.9]; The metabolic heat powers per unit area are dynamically updated based on the corrected metabolic factors, and the air-conditioning operation parameters are dynamically adjusted based on the metabolic heat powers per unit area.

4. The building energy-saving and carbon-reducing refrigeration control method according to claim 1, characterized in that: The step S4 is specifically as follows: Based on the metabolic heat power per unit area and the area of ​​the temperature control area, the total metabolic heat power of each temperature control area is calculated: ; in, represents the total metabolic heat power; It represents the metabolic heat power per unit area; Indicates the area of ​​the temperature control area; Using a sliding window algorithm combined with an exponentially weighted average method, the heat load for the next 15 to 30 minutes is predicted based on the total metabolic heat power, and the supply air temperature, supply air volume, supply air angle, and return air path of the air conditioning system are controlled based on the predicted heat load; When the heat load reaches the upper limit of the cooling capacity of the air-conditioning system, the temperature control parameters of each temperature control area are adjusted based on the preset air supply priority.

5. The building energy-saving and carbon-reducing refrigeration control method according to claim 1, characterized in that: The step S5 is specifically as follows: A primary cold source, a secondary cold source, and a tertiary cold source are provided. The priority of the primary cold source is higher than that of the secondary cold source, and the priority of the secondary cold source is higher than that of the tertiary cold source. The primary cold source adopts a cold storage unit composed of a water-ice crystal mixed phase change material; the secondary cold source adopts a condensed water heat exchange device; and the tertiary cold source adopts air conditioning host cooling. Setting scheduling rules for the primary cooling source, the secondary cooling source, and the tertiary cooling source, and controlling the primary cooling source, the secondary cooling source, and the tertiary cooling source to perform cold energy storage and cold energy release based on the scheduling rules and the heat load; The step S6 is specifically as follows: The carbon intensity emission data of the power grid is obtained in real time, and the carbon intensity emission data is analyzed. When the carbon intensity is higher than a preset carbon intensity threshold, the cooling load of the compressor of the air-conditioning system is reduced, or the startup of high-energy-consuming equipment is delayed.

6. A building energy-saving and carbon-reducing refrigeration control system, characterized by: Includes the following modules: An activity metabolic factor generation module is configured to collect environmental video using cameras installed in various temperature-controlled areas within the building, identify personnel information from the environmental video, and generate activity factors based on the personnel information; collect carbon dioxide concentration using carbon dioxide sensors installed in various temperature-controlled areas within the building, and generate metabolic factors based on the carbon dioxide concentration; A metabolic heat power calculation module is used to input the activity factor and metabolic factor into a pre-trained dynamic metabolic heat prediction model to obtain the metabolic heat power per unit area of ​​each temperature control area; The metabolic heat power update module is used to collect in real time the air conditioning operating parameters of the air conditioning system deployed in each temperature-controlled area inside the building, including the air supply wind speed, wind direction angle, fan operating frequency, and indoor temperature gradient, and to correct the metabolic factor based on each of the air conditioning operating parameters to dynamically update the metabolic heat power per unit area, and to dynamically adjust the air conditioning operating parameters based on each of the metabolic heat power per unit area; a heat load prediction module, configured to predict future heat load based on the metabolic heat power per unit area; A cold energy storage module, configured to control a cold source to perform cold energy storage based on the predicted heat load; A cooling control module, configured to obtain carbon intensity emission data from the power grid in real time and control the operation of the air conditioning system based on the carbon intensity emission data; The activity metabolic factor generation module is specifically used for: The edge computing device collects environmental videos in real time through cameras placed in various temperature-controlled areas inside the building. After performing data desensitization operations on the environmental videos, it uses inter-frame difference technology to identify the outlines of people from each environmental image frame in the environmental videos. Based on each of the outlines, it identifies personnel information including at least the number of people, movement trajectory, and activity frequency. Based on the personnel information, it calculates the activity factor. After the calculation is completed, the environmental videos are deleted: ; in, represents the activity factor; Indicates the total number of sampled frames of the environment image frame, with a value range of [10,30]; Indicates the frame number of the environment image frame; Indicates the The number of person silhouettes recognized in the frame; The edge computing device collects carbon dioxide concentrations through carbon dioxide sensors using a pressure differential drainage sampling structure, which are deployed in various temperature-controlled areas within the building. Based on this carbon dioxide concentration, a metabolic factor is generated: ; in, represents metabolic factors; express The carbon dioxide concentration at the time; express The carbon dioxide concentration at the time; Indicates a time interval.

7. A building energy-saving and carbon-reducing refrigeration control system according to claim 6, characterized in that: In the metabolic heat power calculation module, the dynamic metabolic heat prediction model is constructed based on a three-layer feedforward neural network, with two input layer nodes for inputting activity factors and metabolic factors, and one output layer node for outputting metabolic heat power per unit area. The number of hidden layer nodes ranges from [5 to 10], and the activation function adopts ReLU. The dynamic metabolic heat prediction model has an online adaptive learning function and dynamically adjusts the model weight parameters based on the air conditioning return air temperature difference and regional energy consumption.

8. The building energy-saving and carbon-reducing refrigeration control system according to claim 6, characterized in that: The metabolic heat power update module is specifically used for: The air conditioning system operating parameters of each temperature-controlled area in the building, including air supply speed, wind direction angle, fan operating frequency, and indoor temperature gradient, are collected in real time. The metabolic factor is corrected based on these air conditioning operating parameters: Calculate the perturbation factor: ; in, represents the disturbance factor; 、 、 、 Both represent the experience weight coefficient; Indicates the air supply speed; Indicates wind direction angle; Indicates the fan operating frequency; Indicates the indoor temperature gradient; When the disturbance factor is greater than a preset disturbance threshold, each of the metabolic factors is corrected: ; in, represents the corrected metabolic factor; represents metabolic factors; represents the disturbance correction coefficient, and its value range is [0.6, 0.9]; The metabolic heat powers per unit area are dynamically updated based on the corrected metabolic factors, and the air-conditioning operation parameters are dynamically adjusted based on the metabolic heat powers per unit area.

9. The building energy-saving and carbon-reducing refrigeration control system according to claim 6, characterized in that: The heat load prediction module is specifically used for: Based on the metabolic heat power per unit area and the area of ​​the temperature control area, the total metabolic heat power of each temperature control area is calculated: ; in, represents the total metabolic heat power; It represents the metabolic heat power per unit area; Indicates the area of ​​the temperature control area; Using a sliding window algorithm combined with an exponentially weighted average method, the heat load for the next 15 to 30 minutes is predicted based on the total metabolic heat power, and the supply air temperature, supply air volume, supply air angle, and return air path of the air conditioning system are controlled based on the predicted heat load; When the heat load reaches the upper limit of the cooling capacity of the air-conditioning system, the temperature control parameters of each temperature control area are adjusted based on the preset air supply priority.

10. The building energy-saving and carbon-reducing refrigeration control system according to claim 6, characterized in that: The cold energy storage module is specifically used for: A primary cold source, a secondary cold source, and a tertiary cold source are provided. The priority of the primary cold source is higher than that of the secondary cold source, and the priority of the secondary cold source is higher than that of the tertiary cold source. The primary cold source adopts a cold storage unit composed of a water-ice crystal mixed phase change material; the secondary cold source adopts a condensed water heat exchange device; and the tertiary cold source adopts air conditioning host cooling. Setting scheduling rules for the primary cooling source, the secondary cooling source, and the tertiary cooling source, and controlling the primary cooling source, the secondary cooling source, and the tertiary cooling source to perform cold energy storage and cold energy release based on the scheduling rules and the heat load; The refrigeration control module is specifically used for: The carbon intensity emission data of the power grid is obtained in real time, and the carbon intensity emission data is analyzed. When the carbon intensity is higher than a preset carbon intensity threshold, the cooling load of the compressor of the air-conditioning system is reduced, or the startup of high-energy-consuming equipment is delayed.

Images