An air conditioner load flexible control method and system based on edge computing

Abstract

This application provides a flexible control method and system for air conditioning load based on edge computing. The method includes: acquiring current control reference data corresponding to a target air conditioner at the current control moment; inputting the current control reference data and a preset adjustable air conditioning temperature range into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data; determining the control temperature of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioning operating parameters of the target air conditioner, and the control strategy of the target air conditioner, wherein the control strategy includes the adjustable comfort range and target power of the target air conditioner; and sending the control temperature to the target air conditioner to adjust the air conditioning outlet water temperature of the target air conditioner at the next control moment, thereby improving the accuracy of air conditioning load control.

Description

Technical Field

[0001] This application relates to the fields of edge computing and load control technology, and in particular to a flexible control method and system for air conditioning load based on edge computing. Background Technology

[0002] With the continuous acceleration of urbanization in my country, the scale of building energy consumption continues to expand, and air conditioning systems have become one of the main components of modern urban electricity load. Especially under extreme high-temperature conditions in summer, the concentrated operation of a large number of air conditioning units exacerbates the peak-valley difference in the power grid, significantly increasing the pressure on power supply and posing a severe challenge to the safe and stable operation of urban power systems.

[0003] Currently, to cope with the tight power supply during peak hours, some power grid companies typically adopt rigid control measures such as limiting user power consumption to reduce peak loads. While these methods can alleviate grid pressure in the short term, they also directly impact users' normal air conditioning experience, reducing indoor comfort and making it difficult to balance energy conservation and load reduction with user experience. Against this backdrop, how to effectively regulate air conditioning load while ensuring users' basic comfort needs has become an important research direction in the field of refined power load management. Traditional direct power rationing methods are no longer adequate for the dual requirements of smart electricity use and humanized services, necessitating the exploration of more flexible, intelligent, and user-unobtrusive load regulation models. Therefore, researching an air conditioning load regulation method that can respond to grid dispatching needs while maintaining a good user experience is of great significance for improving urban energy efficiency and promoting the construction of new power systems. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a flexible control method and system for air conditioning load based on edge computing, which effectively regulates air conditioning load while ensuring user comfort and improving the accuracy of air conditioning load regulation.

[0005] Beneficial effects: In a first aspect, embodiments of this application provide a flexible control method for air conditioning load based on edge computing, comprising: At the current control moment, obtain the current control reference data corresponding to the target air conditioner. The current control reference data includes the current environmental data, the current air conditioner operating parameters, and the current number of people indoors. The current control reference data and the preset adjustable air conditioning temperature range are input into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data; wherein, the comfort prediction model is constructed based on a neural network and trained based on several historical control reference data; Based on the temperature-comfort index mapping table, the current air conditioning operating parameters of the target air conditioner, and the control strategy of the target air conditioner, the regulating temperature of the target air conditioner is determined, and the control strategy includes the adjustable comfort range and target power of the target air conditioner. The controlled temperature is sent to the target air conditioner, thereby adjusting the air conditioner outlet water temperature of the target air conditioner at the next control moment.

[0006] This application provides a flexible air conditioning load control method based on edge computing. By comprehensively utilizing current environmental data, air conditioning operating parameters, and multi-dimensional information such as the number of people indoors reflecting actual usage, a comprehensive control reference dataset is constructed. Simultaneously, a comfort prediction model based on neural network training is introduced. This model can dynamically generate a "temperature-comfort index mapping table," thereby establishing a quantitative relationship between the air conditioning outlet water temperature and the user's subjective comfort in real time based on the surrounding environment. This makes the control decision no longer a simple temperature setting, but a scientific control based on maintaining user comfort. In the final control temperature determination stage, this application comprehensively determines the control temperature by integrating the current outlet water temperature, the temperature-comfort index mapping table, the current comfort index, the adjustable comfort range, real-time power, and target power. Compared to rigid methods such as direct power rationing or simple temperature range control commonly used in the prior art, this application can actively respond to the grid's dispatching needs while ensuring or even optimizing the user experience. It achieves effective control of the air conditioning load, especially the load during peak hours, while ensuring user comfort, thus improving the accuracy of air conditioning load control.

[0007] In one possible implementation, obtaining the current control reference data corresponding to the target air conditioner includes: If the current control time is the initial time of the target control time period, then the preset initial control reference data is used as the current control reference data. The environmental data in the initial control reference data is determined based on the meteorological forecast data of the location of the target air conditioner, and the air conditioner operating parameters and the number of people indoors in the initial control reference data are determined based on several historical control reference data of the target air conditioner. If the current control time is not the initial time of the target control time period, then the current outdoor temperature and humidity data and the current indoor temperature and humidity data of the target air conditioner location are obtained as the previous environmental data; the current air supply volume data, real-time power, current air outlet water temperature data, and current air return water temperature data of the target air conditioner are obtained as the current air conditioner operating parameters; the current number of people indoors at the target air conditioner location is obtained; and the current control reference data is obtained by combining the current environmental data, the current air conditioner operating parameters, and the current number of people indoors.

[0008] This application provides a method for acquiring current control reference data, which finely distinguishes the timing and source of the acquisition of control reference data, resulting in significant improvements in robustness and adaptability. At the initial moment of the target control period, the system may lack sufficient real-time data. In this case, by using environmental data based on meteorological forecasts and operating parameters and personnel data based on historical patterns, a reasonable initial control benchmark can be formed, ensuring stable operation of the system during the startup phase and avoiding control failures due to data gaps. At non-initial moments, a comprehensive real-time data acquisition mode is switched to ensure the timeliness and accuracy of the control basis. This time-sharing strategy effectively overcomes the limitations of a single data source, enabling the system to make decisions based on the most suitable data, whether in the initial stage of control or during continuous operation. This enhances the stability and reliability of the system at different operating stages and improves the accuracy of air conditioning load control.

[0009] In one possible implementation, the step of inputting the current control reference data and the preset adjustable air conditioning temperature range into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data, includes: Based on the adjustable range of the air conditioner temperature and the preset temperature step, several air conditioner outlet water temperature adjustment values ​​are generated. Each of the air conditioner outlet water temperature adjustment values ​​is combined with the current control reference data to generate corresponding model input data. Each of the model input data is input into the comfort prediction model so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values. By combining the various air conditioner outlet water temperature adjustment values ​​with the various comfort indices, the temperature-comfort index mapping table is obtained.

[0010] This application provides a method for generating a temperature-comfort index mapping table, further improving the precision of the control process. First, based on a preset temperature step size, a series of discrete, representative air conditioning outlet water temperature adjustment values ​​are generated within the allowable temperature adjustment range. Then, each temperature adjustment value is combined with abundant current control reference data to form an independent model input, and the expected comfort index at each assumed temperature is calculated in batches using a comfort prediction model. The resulting mapping table is essentially a complete "decision menu," clearly listing the comfort changes brought about by different temperature adjustment options. This provides sufficient data support and a scientific basis for subsequent determination of the optimal control temperature, completely avoiding the drawbacks of traditional methods that rely on experience or blind adjustments, and improving the accuracy of air conditioning load control.

[0011] Furthermore, the step of inputting each of the model input data into the comfort prediction model, so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values, includes: For each model input data, the model input data is continuously activated through several preset hidden layers to obtain the final feature vector of the model input data, wherein the input value of any hidden layer is the weighted sum of the output values ​​of each node of the previous hidden layer; The final feature vector is input to a preset output layer so that the output layer generates a comfort index corresponding to the air conditioner outlet water temperature adjustment value of the model input data.

[0012] The embodiments of this application further define the internal working mechanism of the comfort prediction model. The multi-layer hidden layers set in the model can abstract and extract the multi-dimensional and nonlinear features of the input layer by layer, thereby more accurately capturing the complex interaction between the environment, equipment and people. This helps the model to better learn and represent the deep laws affecting comfort, ensuring the high accuracy and strong generalization ability of the model prediction, and finally outputting more reliable and accurate comfort index prediction results, thereby improving the accuracy of air conditioning load control.

[0013] In one possible implementation, if the target air conditioner is operating in cooling mode, determining the controlled temperature of the target air conditioner based on the temperature-comfort index mapping table, the current operating parameters of the target air conditioner, and the control strategy of the target air conditioner includes: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the maximum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the minimum temperature in the adjustable temperature set that is greater than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

[0014] This application provides a temperature control method with a clear and rigorous temperature control decision logic specifically designed for cooling mode, achieving intelligent trade-offs under multiple constraints. First, this embodiment determines the set of adjustable temperatures for the target air conditioner based on the adjustable comfort range and a mapping table. Then, the decision-making process prioritizes user experience: if the current comfort level is below the acceptable minimum threshold, the maximum value in the set of adjustable temperatures is selected to quickly restore comfort. If the current comfort level is still within the acceptable range, the grid demand is further considered. When the real-time power exceeds the target power, the air conditioner power is not adjusted to the target power all at once, but the temperature is raised to a minimum value just above the current temperature. This "minimal intervention" principle allows for a slight reduction in cooling intensity at each control moment with almost imperceptible feedback, thereby gradually and effectively reducing energy consumption and grid load, achieving fine-grained load adjustment within the flexible space allowed by comfort.

[0015] In one possible implementation, if the target air conditioner is operating in heating mode, determining the controlled temperature of the target air conditioner based on the temperature-comfort index mapping table, the current operating parameters of the target air conditioner, and the control strategy of the target air conditioner includes: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the minimum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the maximum temperature value in the adjustable temperature set that is less than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

[0016] Corresponding to the cooling mode, this application provides a temperature control method for the heating mode, which is specifically optimized based on the characteristics of winter heating. When insufficient comfort is detected, the system selects the minimum temperature to enhance heating and quickly improve the user experience. When comfort is achieved but power consumption exceeds the limit, the temperature is strategically lowered to a maximum value just below the current temperature. This method can moderately reduce heating power with almost imperceptible changes to the user, responding to the grid's peak-shaving instructions while avoiding discomfort caused by a sudden drop in indoor temperature, thus improving the accuracy of air conditioning load control.

[0017] Furthermore, the comfort prediction model, constructed based on a neural network and trained using several historical control reference data, includes: An initial comfort prediction model is constructed based on a neural network. The initial comfort prediction model includes one input layer, three hidden layers, and one output layer. The activation function of the hidden layer is the GELU function. The comfort index of various public places at different times was obtained through questionnaires; Data on indoor occupants, indoor temperature and humidity, outdoor temperature and humidity, air conditioning air volume, air conditioning outlet water temperature, and air conditioning return water temperature of each of the aforementioned public places at different time periods are obtained, and then several historical control reference data for each of the aforementioned public places are constructed. The historical control reference data and the comfort index are matched and aligned by time period and location, and a training dataset is constructed using the comfort index as the data label. The initial comfort prediction model is trained based on the training dataset and a preset loss function. The model parameters of the initial comfort prediction model are adjusted through error backpropagation to obtain the comfort prediction model.

[0018] This application systematically describes the entire process of constructing and training a comfort prediction model. Real user comfort feedback is obtained through questionnaires and used as data labels. This data is then precisely matched with large-scale, multi-dimensional historical operational data collected from various public places. The resulting training dataset combines subjective authenticity with objective comprehensiveness. Based on this, an initial comfort prediction model is established using a multi-layer neural network structure incorporating the GELU activation function. Iterative optimization is then performed using an error backpropagation algorithm, ensuring that the final trained model not only has a solid theoretical foundation but also possesses strong practical generalization capabilities. It can adapt to the differentiated needs of different places and different groups of people, providing an accurate core decision engine for the entire control system and improving the precision of air conditioning load control.

[0019] Secondly, embodiments of this application provide a flexible air conditioning load control system based on edge computing, including an acquisition module, a comfort prediction module, a temperature control determination module, and a control module; The acquisition module is used to acquire the current control reference data corresponding to the target air conditioner at the current control time. The current control reference data includes the current environmental data, the current air conditioner operating parameters, and the current number of people indoors. The comfort prediction module is used to input the current control reference data and the preset adjustable range of air conditioning temperature into the preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable range of air conditioning temperature based on the current control reference data; wherein, the comfort prediction model is constructed based on a neural network and trained based on several historical control reference data; The temperature control determination module is used to determine the temperature control of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner. The control strategy includes the adjustable comfort range and target power of the target air conditioner. The control module is used to send the regulated temperature to the target air conditioner, thereby adjusting the air conditioner outlet water temperature of the target air conditioner at the next regulation time.

[0020] Furthermore, the comfort prediction module inputs the current control reference data and the preset adjustable air conditioning temperature range into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data, including: Based on the adjustable range of the air conditioner temperature and the preset temperature step, several air conditioner outlet water temperature adjustment values ​​are generated. Each of the air conditioner outlet water temperature adjustment values ​​is combined with the current control reference data to generate corresponding model input data. Each of the model input data is input into the comfort prediction model so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values. By combining the various air conditioner outlet water temperature adjustment values ​​with the various comfort indices, the temperature-comfort index mapping table is obtained.

[0021] Furthermore, if the target air conditioner is operating in cooling mode, the temperature control determination module determines the control temperature of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner, including: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the maximum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the minimum temperature in the adjustable temperature set that is greater than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner. Attached Figure Description

[0022] Figure 1 A flowchart illustrating a flexible control method for air conditioning load based on edge computing, provided in an embodiment of this application; Figure 2 A schematic diagram of a control system structure for implementing a flexible control method for air conditioning load, provided in an embodiment of this application; Figure 3 A physical schematic diagram of an air conditioning protocol conversion module in a control system for implementing a flexible air conditioning load control method, provided in an embodiment of this application; Figure 4 A schematic diagram of an air conditioning power monitoring module in a control system for implementing a flexible air conditioning load control method, provided in an embodiment of this application; Figure 5 A schematic diagram of the process of an edge computing module reading air conditioning parameters in a control system for implementing a flexible control method for air conditioning load, provided in an embodiment of this application; Figure 6 A schematic diagram of the comfort prediction model of an air conditioning load flexible control method based on edge computing provided in an embodiment of this application; Figure 7 A schematic diagram illustrating the process of a flexible control method for air conditioning load based on edge computing in cooling mode, which is provided as an embodiment of this application, for the cyclic control of the control time period. Figure 8 A schematic diagram of the training process of a comfort prediction model for an edge computing-based flexible control method for air conditioning load provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of a flexible air conditioning load control system based on edge computing, provided in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0024] It should be noted that the step numbers in this document are only for the convenience of explaining the specific embodiments and are not intended to limit the order in which the steps are performed. In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0025] Example 1: like Figure 1 As shown, Embodiment 1 provides a flexible control method for air conditioning load based on edge computing, including steps S1-S4: Step S1: At the current control time, obtain the current control reference data corresponding to the target air conditioner. The current control reference data includes the current environmental data, the current air conditioner operating parameters, and the current number of people indoors. Step S2: Input the current control reference data and the preset adjustable range of air conditioning temperature into the preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable range of air conditioning temperature based on the current control reference data; wherein, the comfort prediction model is constructed based on a neural network and trained based on several historical control reference data; Step S3: Determine the control temperature of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner. The control strategy includes the adjustable comfort range and target power of the target air conditioner. Step S4: Send the controlled temperature to the target air conditioner, thereby adjusting the air conditioner outlet water temperature of the target air conditioner at the next control moment.

[0026] This application provides a flexible air conditioning load control method based on edge computing. By comprehensively utilizing current environmental data, air conditioning operating parameters, and multi-dimensional information such as the number of people indoors reflecting actual usage, a comprehensive control reference dataset is constructed. Simultaneously, a comfort prediction model based on neural network training is introduced. This model can dynamically generate a "temperature-comfort index mapping table," thereby establishing a quantitative relationship between the air conditioning outlet water temperature and the user's subjective comfort in real time based on the surrounding environment. This makes the control decision no longer a simple temperature setting, but a scientific control based on maintaining user comfort. In the final control temperature determination stage, this application comprehensively determines the control temperature by integrating the current outlet water temperature, the temperature-comfort index mapping table, the current comfort index, the adjustable comfort range, real-time power, and target power. Compared to rigid methods such as direct power rationing or simple temperature range control commonly used in the prior art, this application can actively respond to the grid's dispatching needs while ensuring or even optimizing the user experience. It achieves effective control of the air conditioning load, especially the load during peak hours, while ensuring user comfort, thus improving the accuracy of air conditioning load control.

[0027] In a preferred embodiment, such as Figure 2 The diagram shows the overall control system scheme for implementing a flexible air conditioning load control method. It includes a power grid master station, an edge computing module, an air conditioning power monitoring module, an air conditioning protocol conversion module, a temperature and humidity monitoring module, an air supply volume monitoring system, a personnel monitoring system, and an air conditioning unit. The air conditioning power monitoring module collects the air conditioning unit load power at 1-minute intervals via RS485. The air conditioning protocol conversion module collects or sets the air conditioning outlet water temperature via RS485. The edge computing module is connected to each sub-module via RS485. The edge computing module collects data from other modules simultaneously with the air conditioning power monitoring module. The temperature and humidity monitoring module monitors the temperature and humidity data at the user's main control points in real time. The air supply volume monitoring system collects air supply volume data from multiple user points in real time. The personnel monitoring system collects the current number of people in the user's household in real time. Based on a preset comfort prediction model, the edge computing module outputs the current comfort index and a temperature-comfort index mapping table based on the collected data. This data is then transmitted to the power grid master station via 4G communication at a frequency of 5 minutes. The power grid master station determines and distributes the air conditioning outlet water temperature based on the transmitted information, combined with the user level and target control power. The air conditioning protocol conversion module, also known as the air conditioning flexible control protocol conversion gateway (product name), is shown in the image below. Figure 3As shown. The air conditioning flexible control protocol conversion gateway can connect to smart energy units, load control units, or cloud platforms to achieve centralized control and management of multi-brand air conditioning systems. Through protocol conversion, it obtains the real-time operating parameters of the air conditioning units and enables flexible adjustment of the air conditioning unit's power by setting the set temperature. The air conditioning power monitoring module, also known as the load management branch device (product name), is shown in the image below. Figure 4 As shown, the load management branch unit is a power measurement and control device composed of a measurement unit, a data processing and storage unit, and a communication unit. It has functions such as measurement, metering, information storage and processing, real-time monitoring, information interaction, remote signaling, and remote control. The load management branch unit is installed at the power line inlet of the central air conditioning unit and can collect the air conditioning load current, voltage, and power data minute by minute, transmitting the data to the edge computing module via RS485 communication. Figure 5 This flowchart describes the process of reading air conditioning parameters for the edge computing module, including the edge computing module, the air conditioning protocol conversion module, the BA (Building Automation System) control system, and the air conditioning main unit. Users decide whether to connect the air conditioner to their own BA system based on their specific needs. The air conditioning protocol conversion module connects to the BA system via RS485 or directly to the air conditioning main unit via RS485. The air conditioning protocol conversion module collects air conditioning temperature data by broadcasting the air conditioning communication protocol.

[0028] In one possible implementation, step S1, obtaining the current control reference data corresponding to the target air conditioner, includes: If the current control time is the initial time of the target control time period, then the preset initial control reference data is used as the current control reference data. The environmental data in the initial control reference data is determined based on the meteorological forecast data of the location of the target air conditioner, and the air conditioner operating parameters and the number of people indoors in the initial control reference data are determined based on several historical control reference data of the target air conditioner. If the current control time is not the initial time of the target control time period, then the current outdoor temperature and humidity data and the current indoor temperature and humidity data of the target air conditioner location are obtained as the previous environmental data; the current air supply volume data, real-time power, current air outlet water temperature data, and current air return water temperature data of the target air conditioner are obtained as the current air conditioner operating parameters; the current number of people indoors at the target air conditioner location is obtained; and the current control reference data is obtained by combining the current environmental data, the current air conditioner operating parameters, and the current number of people indoors.

[0029] This application provides a method for acquiring current control reference data, which finely distinguishes the timing and source of the acquisition of control reference data, resulting in significant improvements in robustness and adaptability. At the initial moment of the target control period, the system may lack sufficient real-time data. In this case, by using environmental data based on meteorological forecasts and operating parameters and personnel data based on historical patterns, a reasonable initial control benchmark can be formed, ensuring stable operation of the system during the startup phase and avoiding control failures due to data gaps. At non-initial moments, a comprehensive real-time data acquisition mode is switched to ensure the timeliness and accuracy of the control basis. This time-sharing strategy effectively overcomes the limitations of a single data source, enabling the system to make decisions based on the most suitable data, whether in the initial stage of control or during continuous operation. This enhances the stability and reliability of the system at different operating stages and improves the accuracy of air conditioning load control.

[0030] In a preferred embodiment, at the initial moment of the target control period, the master station obtains the control requirements issued by the power grid company, initiates the temperature regulation strategy, and obtains outdoor temperature and humidity forecast data for the target control period from relevant meteorological departments, which is then sent in advance to the edge computing modules of the corresponding users. Based on the data sent by the master station and combined with historical data, the edge computing modules set initial values ​​for outdoor temperature and humidity, number of indoor occupants, indoor temperature and humidity, air conditioning return water temperature, and air conditioning supply air volume.

[0031] In one possible implementation, in step S2, inputting the current control reference data and the preset adjustable air conditioning temperature range into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data, includes: Based on the adjustable range of the air conditioner temperature and the preset temperature step, several air conditioner outlet water temperature adjustment values ​​are generated. Each of the air conditioner outlet water temperature adjustment values ​​is combined with the current control reference data to generate corresponding model input data. Each of the model input data is input into the comfort prediction model so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values. By combining the various air conditioner outlet water temperature adjustment values ​​with the various comfort indices, the temperature-comfort index mapping table is obtained.

[0032] This application provides a method for generating a temperature-comfort index mapping table, further improving the precision of the control process. First, based on a preset temperature step size, a series of discrete, representative air conditioning outlet water temperature adjustment values ​​are generated within the allowable temperature adjustment range. Then, each temperature adjustment value is combined with abundant current control reference data to form an independent model input, and the expected comfort index at each assumed temperature is calculated in batches using a comfort prediction model. The resulting mapping table is essentially a complete "decision menu," clearly listing the comfort changes brought about by different temperature adjustment options. This provides sufficient data support and a scientific basis for subsequent determination of the optimal control temperature, completely avoiding the drawbacks of traditional methods that rely on experience or blind adjustments, and improving the accuracy of air conditioning load control.

[0033] In a preferred embodiment, the edge computing module presets the air conditioner outlet water temperature in 1°C increments within the allowable outlet water temperature adjustment range. It then combines each preset temperature value with the current control reference data and inputs the result into a preset comfort prediction model to calculate the corresponding comfort index output value. This leads to the temperature-comfort index mapping table. And upload it to the main site.

[0034] Furthermore, the step of inputting each of the model input data into the comfort prediction model, so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values, includes: For each model input data, the model input data is continuously activated through several preset hidden layers to obtain the final feature vector of the model input data, wherein the input value of any hidden layer is the weighted sum of the output values ​​of each node of the previous hidden layer; The final feature vector is input to a preset output layer so that the output layer generates a comfort index corresponding to the air conditioner outlet water temperature adjustment value of the model input data.

[0035] The embodiments of this application further define the internal working mechanism of the comfort prediction model. The multi-layer hidden layers set in the model can abstract and extract the multi-dimensional and nonlinear features of the input layer by layer, thereby more accurately capturing the complex interaction between the environment, equipment and people. This helps the model to better learn and represent the deep laws affecting comfort, ensuring the high accuracy and strong generalization ability of the model prediction, and finally outputting more reliable and accurate comfort index prediction results, thereby improving the accuracy of air conditioning load control.

[0036] In a preferred embodiment, the edge computing module calculates a comfort index based on periodically collected and monitored data, including the total number of people, real-time indoor temperature and humidity, real-time outdoor temperature and humidity, air conditioning supply air volume, and air conditioning inlet and outlet water temperatures. The model consists of modules for indoor occupancy, real-time indoor temperature and humidity, real-time outdoor temperature and humidity, air conditioning supply air volume, air conditioning outlet and outlet water temperatures, a occupant comfort index, and parameter transfer lines between these modules. The comfort prediction model consists of one input layer, three hidden layers, and one output layer. Figure 6 As shown, the input layer of the comfort prediction model consists of data nodes such as the indoor occupant number module, indoor real-time temperature data module, indoor real-time humidity data module, outdoor real-time temperature data module, outdoor real-time humidity data module, air conditioning supply air volume data module, and air conditioning outlet and return water temperature data module. The output layer is the comfort index module node.

[0037] In one possible implementation, in step S3, if the target air conditioner is operating in cooling mode, determining the control temperature of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner includes: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the maximum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the minimum temperature in the adjustable temperature set that is greater than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

[0038] This application provides a temperature control method with a clear and rigorous temperature control decision logic specifically designed for cooling mode, achieving intelligent trade-offs under multiple constraints. First, this embodiment determines the set of adjustable temperatures for the target air conditioner based on the adjustable comfort range and a mapping table. Then, the decision-making process prioritizes user experience: if the current comfort level is below the acceptable minimum threshold, the maximum value in the set of adjustable temperatures is selected to quickly restore comfort. If the current comfort level is still within the acceptable range, the grid demand is further considered. When the real-time power exceeds the target power, the air conditioner power is not adjusted to the target power all at once, but the temperature is raised to a minimum value just above the current temperature. This "minimal intervention" principle allows for a slight reduction in cooling intensity at each control moment with almost imperceptible feedback, thereby gradually and effectively reducing energy consumption and grid load, achieving fine-grained load adjustment within the flexible space allowed by comfort.

[0039] In a preferred embodiment, the main station pre-classifies users into four levels (1-4) based on user attributes, importance, and indoor temperature requirements. The adjustable range of the comfort index is then set according to the user level, as shown in the table below. Table 1 User Comfort Levels The main site determines the adjustable range of the user's comfort index based on the user's corresponding level, and then sets the set according to the adjustable range of the comfort index. The system filters out air conditioning outlet water temperatures that meet comfort requirements, sorts them from highest to lowest temperature, and outputs a set of adjustable temperatures. At the initial moment of the target control period, the master station randomly selects a value from the adjustable temperature set as the initial control value and sends it to the user edge computing module. At other times, the master station determines the air conditioning outlet water temperature for the next control moment based on the current comfort index, adjustable comfort range, real-time power, and target power, and sends it to the user edge computing module. The edge computing module receives the temperature control command from the master station and forwards it to the air conditioning protocol conversion module. The gateway then sends the command to the on-site building automation system (BAS) or air conditioning host to achieve air conditioning temperature control.

[0040] It should be noted that, as Figure 7 As shown, in practical applications, the air conditioning load control method provided in this embodiment is not used only once, but is used cyclically throughout the target control period. During the user's target air conditioning control period, at each control moment, the edge computing module updates the comfort index and the set temperature-comfort index mapping set based on real-time collected data. The data, along with the real-time power of the air conditioner, is sent to the main station. The main station then adjusts the user's air conditioner outlet water temperature in real time based on the power control target and the adjustable range of the user comfort index, until the control period ends.

[0041] In one possible implementation, in step S3, if the target air conditioner is operating in heating mode, determining the control temperature of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner includes: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the minimum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the maximum temperature value in the adjustable temperature set that is less than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

[0042] Corresponding to the cooling mode, this application provides a temperature control method for the heating mode, which is specifically optimized based on the characteristics of winter heating. When insufficient comfort is detected, the system selects the minimum temperature to enhance heating and quickly improve the user experience. When comfort is achieved but power consumption exceeds the limit, the temperature is strategically lowered to a maximum value just below the current temperature. This method can moderately reduce heating power with almost imperceptible changes to the user, responding to the grid's peak-shaving instructions while avoiding discomfort caused by a sudden drop in indoor temperature, thus improving the accuracy of air conditioning load control.

[0043] Furthermore, the comfort prediction model, constructed based on a neural network and trained using several historical control reference data, includes: An initial comfort prediction model is constructed based on a neural network. The initial comfort prediction model includes one input layer, three hidden layers, and one output layer. The activation function of the hidden layer is the GELU function. The comfort index of various public places at different times was obtained through questionnaires; Data on indoor occupants, indoor temperature and humidity, outdoor temperature and humidity, air conditioning air volume, air conditioning outlet water temperature, and air conditioning return water temperature of each of the aforementioned public places at different time periods are obtained, and then several historical control reference data for each of the aforementioned public places are constructed. The historical control reference data and the comfort index are matched and aligned by time period and location, and a training dataset is constructed using the comfort index as the data label. The initial comfort prediction model is trained based on the training dataset and a preset loss function. The model parameters of the initial comfort prediction model are adjusted through error backpropagation to obtain the comfort prediction model.

[0044] This application systematically describes the entire process of constructing and training a comfort prediction model. Real user comfort feedback is obtained through questionnaires and used as data labels. This data is then precisely matched with large-scale, multi-dimensional historical operational data collected from various public places. The resulting training dataset combines subjective authenticity with objective comprehensiveness. Based on this, an initial comfort prediction model is established using a multi-layer neural network structure incorporating the GELU activation function. Iterative optimization is then performed using an error backpropagation algorithm, ensuring that the final trained model not only has a solid theoretical foundation but also possesses strong practical generalization capabilities. It can adapt to the differentiated needs of different places and different groups of people, providing an accurate core decision engine for the entire control system and improving the precision of air conditioning load control.

[0045] In a preferred embodiment, the initial comfort prediction model sets the activation function of the three hidden layers to be the GELU function, with the number of nodes being 256, 128, and 64 respectively, the activation function of the output layer to be the GELU function, the loss function to be the mean squared error function, and the initial values ​​of the bias and weight of each layer to be random values.

[0046] The input value z of the hidden layer and output layer nodes is the sum of the output weights of the nodes in the previous layer, calculated as follows: In the formula, This is the i-th output from the previous layer. for The weights are given by b, where b is the node bias and n is the total number of input parameters.

[0047] The formula for activating the GELU function is shown below: In the formula, erf is the Gaussian error function.

[0048] The formula for calculating the mean squared error function is as follows: In the formula Let N be the mean squared error, and N be the number of samples in the training batch. To calculate the true value of comfort in a batch, The neural network outputs a comfort level value.

[0049] like Figure 8 As shown, during training, the training set is input into the initially constructed neural network (initial comfort prediction model). The error is calculated based on the comfort prediction values ​​output by the neural network, the labels in the training data, and the preset loss function. Then, through backpropagation of the error, the weights and biases of each layer are continuously adjusted to obtain the final comfort prediction model. The process of constructing the training set is as follows: (1) Conduct a questionnaire survey in public places such as large shopping malls and office buildings, and invite individuals to rate the indoor comfort level. The score is 1-10 points, and the score is the comfort index. The collection time is recorded at the same time. The day is divided into 30-minute segments and marked with numbers (numbers 1-48). The average comfort index collected in the same time period is calculated as the comfort index of that time period. (2) Continuously collect and record the indoor number of people, indoor real-time temperature data, indoor real-time humidity data, outdoor real-time temperature data, outdoor real-time humidity data, air conditioning air volume data, and average air conditioning outlet and return water temperature data for 48 time periods a day. (3) Match the comfort index of the same period with the number of people in the room, the real-time indoor temperature data, the real-time indoor humidity data, the real-time outdoor temperature data, the real-time outdoor humidity data, the air supply volume data module of the air conditioner, and the average value of the air conditioner outlet and return water temperature data to form a training set.

[0050] Example 2: like Figure 9 As shown, Embodiment 2 provides a flexible control system for air conditioning load based on edge computing, including an acquisition module 10, a comfort prediction module 20, a temperature control determination module 30, and a control module 40. The acquisition module 10 is used to acquire the current control reference data corresponding to the target air conditioner at the current control time. The current control reference data includes the current environmental data, the current air conditioner operating parameters, and the current number of people indoors. The comfort prediction module 20 is used to input the current control reference data and the preset adjustable range of air conditioning temperature into the preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable range of air conditioning temperature based on the current control reference data; wherein, the comfort prediction model is constructed based on a neural network and trained based on several historical control reference data; The temperature control determination module 30 is used to determine the temperature control of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner. The control strategy includes the adjustable comfort range and target power of the target air conditioner. The control module 40 is used to send the controlled temperature to the target air conditioner, thereby adjusting the air conditioner outlet water temperature of the target air conditioner at the next control moment.

[0051] Furthermore, the acquisition module 10 acquires the current control reference data corresponding to the target air conditioner, including: If the current control time is the initial time of the target control time period, then the preset initial control reference data is used as the current control reference data. The environmental data in the initial control reference data is determined based on the meteorological forecast data of the location of the target air conditioner, and the air conditioner operating parameters and the number of people indoors in the initial control reference data are determined based on several historical control reference data of the target air conditioner. If the current control time is not the initial time of the target control time period, then the current outdoor temperature and humidity data and the current indoor temperature and humidity data of the target air conditioner location are obtained as the previous environmental data; the current air supply volume data, real-time power, current air outlet water temperature data, and current air return water temperature data of the target air conditioner are obtained as the current air conditioner operating parameters; the current number of people indoors at the target air conditioner location is obtained; and the current control reference data is obtained by combining the current environmental data, the current air conditioner operating parameters, and the current number of people indoors.

[0052] In one possible implementation, the comfort prediction module 20 inputs the current control reference data and a preset adjustable air conditioning temperature range into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data, including: Based on the adjustable range of the air conditioner temperature and the preset temperature step, several air conditioner outlet water temperature adjustment values ​​are generated. Each of the air conditioner outlet water temperature adjustment values ​​is combined with the current control reference data to generate corresponding model input data. Each of the model input data is input into the comfort prediction model so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values. By combining the various air conditioner outlet water temperature adjustment values ​​with the various comfort indices, the temperature-comfort index mapping table is obtained.

[0053] Furthermore, the step of inputting each of the model input data into the comfort prediction model, so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values, includes: For each model input data, the model input data is continuously activated through several preset hidden layers to obtain the final feature vector of the model input data, wherein the input value of any hidden layer is the weighted sum of the output values ​​of each node of the previous hidden layer; The final feature vector is input to a preset output layer so that the output layer generates a comfort index corresponding to the air conditioner outlet water temperature adjustment value of the model input data.

[0054] In one possible implementation, if the target air conditioner is operating in cooling mode, the temperature control module 30 determines the control temperature of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner, including: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the maximum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the minimum temperature in the adjustable temperature set that is greater than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

[0055] In one possible implementation, if the target air conditioner is operating in heating mode, the temperature control module 30 determines the control temperature of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner, including: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the minimum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the maximum temperature value in the adjustable temperature set that is less than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

[0056] In one possible implementation, the comfort prediction model, constructed based on a neural network and trained using several historical regulation reference data, includes: An initial comfort prediction model is constructed based on a neural network. The initial comfort prediction model includes one input layer, three hidden layers, and one output layer. The activation function of the hidden layer is the GELU function. The comfort index of various public places at different times was obtained through questionnaires; Data on indoor occupants, indoor temperature and humidity, outdoor temperature and humidity, air conditioning air volume, air conditioning outlet water temperature, and air conditioning return water temperature of each of the aforementioned public places at different time periods are obtained, and then several historical control reference data for each of the aforementioned public places are constructed. The historical control reference data and the comfort index are matched and aligned by time period and location, and a training dataset is constructed using the comfort index as the data label. The initial comfort prediction model is trained based on the training dataset and a preset loss function. The model parameters of the initial comfort prediction model are adjusted through error backpropagation to obtain the comfort prediction model.

[0057] This application provides an edge computing-based flexible air conditioning load control system. By comprehensively utilizing current environmental data, air conditioning operating parameters, and multi-dimensional information such as the number of people indoors reflecting actual usage, a comprehensive control reference dataset is constructed. Simultaneously, a comfort prediction model trained on a neural network is introduced. This model dynamically generates a "temperature-comfort index mapping table," thereby establishing a quantitative relationship between the air conditioning outlet water temperature and the user's subjective comfort in real time based on the surrounding environment. This makes the control decision no longer a simple temperature setting, but a scientific control based on maintaining user comfort. In the final temperature determination stage, this application comprehensively determines the control temperature by integrating the current outlet water temperature, the temperature-comfort index mapping table, the current comfort index, the adjustable comfort range, real-time power, and target power. Compared to rigid methods such as direct power rationing or simple temperature range control commonly used in the prior art, this application can actively respond to the grid's dispatching needs while ensuring or even optimizing the user experience. It achieves effective control of the air conditioning load, especially during peak hours, while ensuring user comfort, thus improving the accuracy of air conditioning load control.

[0058] For a more detailed explanation of the working principle and procedures of this embodiment, please refer to the relevant description in Embodiment 1.

[0059] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application for those skilled in the art.

Claims

1. A flexible control method for air conditioning load based on edge computing, characterized in that, include: At the current control moment, obtain the current control reference data corresponding to the target air conditioner. The current control reference data includes the current environmental data, the current air conditioner operating parameters, and the current number of people indoors. The current control reference data and the preset adjustable air conditioning temperature range are input into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data; wherein, the comfort prediction model is constructed based on a neural network and trained based on several historical control reference data; Based on the temperature-comfort index mapping table, the current air conditioning operating parameters of the target air conditioner, and the control strategy of the target air conditioner, the regulating temperature of the target air conditioner is determined, and the control strategy includes the adjustable comfort range and target power of the target air conditioner. The controlled temperature is sent to the target air conditioner, thereby adjusting the air conditioner outlet water temperature of the target air conditioner at the next control moment.

2. The flexible control method for air conditioning load based on edge computing as described in claim 1, characterized in that, The acquisition of the current control reference data corresponding to the target air conditioner includes: If the current control time is the initial time of the target control time period, then the preset initial control reference data is used as the current control reference data. The environmental data in the initial control reference data is determined based on the meteorological forecast data of the location of the target air conditioner, and the air conditioner operating parameters and the number of people indoors in the initial control reference data are determined based on several historical control reference data of the target air conditioner. If the current control time is not the initial time of the target control time period, then the current outdoor temperature and humidity data and the current indoor temperature and humidity data of the target air conditioner location are obtained as the previous environmental data; the current air supply volume data, real-time power, current air outlet water temperature data, and current air return water temperature data of the target air conditioner are obtained as the current air conditioner operating parameters; the current number of people indoors at the target air conditioner location is obtained; and the current control reference data is obtained by combining the current environmental data, the current air conditioner operating parameters, and the current number of people indoors.

3. The flexible air conditioning load control method based on edge computing as described in claim 1, characterized in that, The step of inputting the current control reference data and the preset adjustable air conditioning temperature range into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data, includes: Based on the adjustable range of the air conditioner temperature and the preset temperature step, several air conditioner outlet water temperature adjustment values ​​are generated. Each of the air conditioner outlet water temperature adjustment values ​​is combined with the current control reference data to generate corresponding model input data. Each of the model input data is input into the comfort prediction model so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values. By combining the various air conditioner outlet water temperature adjustment values ​​with the various comfort indices, the temperature-comfort index mapping table is obtained.

4. The flexible air conditioning load control method based on edge computing as described in claim 3, characterized in that, The step of inputting the input data of each of the models into the comfort prediction model, so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values, includes: For each model input data, the model input data is continuously activated through several preset hidden layers to obtain the final feature vector of the model input data, wherein the input value of any hidden layer is the weighted sum of the output values ​​of each node of the previous hidden layer; The final feature vector is input to a preset output layer so that the output layer generates a comfort index corresponding to the air conditioner outlet water temperature adjustment value of the model input data.

5. The flexible control method for air conditioning load based on edge computing as described in claim 1, characterized in that, If the target air conditioner is operating in cooling mode, determining the controlled temperature of the target air conditioner based on the temperature-comfort index mapping table, the current operating parameters of the target air conditioner, and the control strategy of the target air conditioner includes: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the maximum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the minimum temperature in the adjustable temperature set that is greater than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

6. The flexible control method for air conditioning load based on edge computing as described in claim 1, characterized in that, If the target air conditioner is operating in heating mode, determining the controlled temperature of the target air conditioner based on the temperature-comfort index mapping table, the current operating parameters of the target air conditioner, and the control strategy of the target air conditioner includes: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the minimum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the maximum temperature value in the adjustable temperature set that is less than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

7. A flexible air conditioning load control method based on edge computing as described in any one of claims 1-6, characterized in that, The comfort prediction model, constructed based on a neural network and trained using several historical control reference data, includes: An initial comfort prediction model is constructed based on a neural network. The initial comfort prediction model includes one input layer, three hidden layers, and one output layer. The activation function of the hidden layer is the GELU function. The comfort index of various public places at different times was obtained through questionnaires; Data on indoor occupants, indoor temperature and humidity, outdoor temperature and humidity, air conditioning air volume, air conditioning outlet water temperature, and air conditioning return water temperature of each of the aforementioned public places at different time periods are obtained, and then several historical control reference data for each of the aforementioned public places are constructed. The historical control reference data and the comfort index are matched and aligned by time period and location, and a training dataset is constructed using the comfort index as the data label. The initial comfort prediction model is trained based on the training dataset and a preset loss function. The model parameters of the initial comfort prediction model are adjusted through error backpropagation to obtain the comfort prediction model.

8. A flexible air conditioning load control system based on edge computing, characterized in that, It includes an acquisition module, a comfort prediction module, a temperature control module, and a control module; The acquisition module is used to acquire the current control reference data corresponding to the target air conditioner at the current control time. The current control reference data includes the current environmental data, the current air conditioner operating parameters, and the current number of people indoors. The comfort prediction module is used to input the current control reference data and the preset adjustable range of air conditioning temperature into the preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable range of air conditioning temperature based on the current control reference data; wherein, the comfort prediction model is constructed based on a neural network and trained based on several historical control reference data; The temperature control determination module is used to determine the temperature control of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner. The control strategy includes the adjustable comfort range and target power of the target air conditioner. The control module is used to send the regulated temperature to the target air conditioner, thereby adjusting the air conditioner outlet water temperature of the target air conditioner at the next regulation time.

9. The flexible air conditioning load control system based on edge computing as described in claim 8, characterized in that, The comfort prediction module inputs the current control reference data and the preset adjustable air conditioning temperature range into a preset comfort prediction model, so that the comfort prediction model generates a temperature-comfort index mapping table corresponding to the adjustable air conditioning temperature range based on the current control reference data, including: Based on the adjustable range of the air conditioner temperature and the preset temperature step, several air conditioner outlet water temperature adjustment values ​​are generated. Each of the air conditioner outlet water temperature adjustment values ​​is combined with the current control reference data to generate corresponding model input data. Each of the model input data is input into the comfort prediction model so that the comfort prediction model generates a comfort index corresponding to each of the air conditioner outlet water temperature adjustment values. By combining the various air conditioner outlet water temperature adjustment values ​​with the various comfort indices, the temperature-comfort index mapping table is obtained.

10. The flexible air conditioning load control system based on edge computing as described in claim 8, characterized in that, If the target air conditioner is operating in cooling mode, the temperature control module determines the control temperature of the target air conditioner based on the temperature-comfort index mapping table, the current air conditioner operating parameters of the target air conditioner, and the control strategy of the target air conditioner, including: Based on the adjustable comfort range and the temperature-comfort index mapping table, determine the set of adjustable temperatures for the target air conditioner; The current comfort index of the target air conditioner is determined based on the current outlet water temperature of the target air conditioner and the temperature-comfort index mapping table; Determine whether the current comfort index is less than the minimum comfort level in the adjustable comfort range. If so, determine the maximum temperature in the adjustable temperature set as the target air conditioner's adjustable temperature. If the current comfort index is not less than the minimum comfort level in the adjustable comfort range, then it is determined whether the real-time power of the target air conditioner is greater than the target power. If so, the minimum temperature in the adjustable temperature set that is greater than the current outlet water temperature is determined as the adjustable temperature of the target air conditioner.

Images