Intelligent air conditioner control method, system and equipment based on federal learning and medium

Abstract

The invention relates to the field of household appliances, and provides an intelligent air conditioner control method, system and device based on federal learning and a medium. According to the environment data and the time information corresponding to the environment data, an air conditioner prediction operation mode is obtained through a global preference model; wherein the global preference model is obtained and issued by the aggregation server by utilizing federal learning on the basis of a local training model which participates in uploading of the client; according to the environment data, the time information corresponding to the environment data and the air conditioner prediction operation mode, an air flow distribution prediction result is obtained through an indoor simulation model created on the basis of computational fluid dynamics; and according to the airflow distribution prediction result, the flexible air guide mechanism is controlled to adjust the angle. The problems that a traditional air conditioner has the risk of user privacy leakage and is poor in multi-user adaptability are solved, and collaborative optimization of high efficiency, energy conservation, accurate airflow regulation and control and user privacy protection is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of household appliances, and in particular, to an intelligent air conditioner control method, system, device and medium based on federated learning. Background Art

[0002] An air conditioner is a device for regulating indoor temperature and humidity, and is widely used in various environments such as homes, offices, and commercial places. In modern office buildings, they usually have large areas and high pedestrian flows, and have relatively high requirements for indoor temperature and air quality. In hot summers and cold winters, air conditioners become key devices for regulating indoor temperature and creating a comfortable working environment.

[0003] Currently, the commonly used air conditioner control method is mainly to collect and analyze user usage data, including temperature preferences, usage time and frequency, etc., and then use these data to train an algorithm model to automatically adjust the operating state of the air conditioner to achieve the purpose of energy saving and comfort improvement.

[0004] However, in order to make the algorithm more accurate, a large amount of personal usage habits and preference data need to be collected, and this data contains sensitive personal information such as living habits and daily activity patterns. If the data is transmitted or stored improperly, it is easy to cause user privacy leakage and increase the risk of unauthorized access to the data. In addition, in a multi-user environment, the usage data of different users may be confused with each other, making it difficult for the system to distinguish the specific needs of individual users, which affects the accuracy and efficiency of the algorithm and at the same time fails to have sufficient dynamic adaptability to respond to the immediate needs of multiple users in real time. Summary of the Invention

[0005] The present invention provides an intelligent air conditioner control method, system, device and medium based on federated learning to solve the defects of the existing traditional air conditioners, such as the risk of user privacy leakage and poor multi-user adaptability, and to achieve the coordinated optimization of high energy efficiency, precise air flow regulation and user privacy protection.

[0006] The present invention provides an intelligent air conditioner control method based on federated learning, including: obtaining environmental data and the corresponding time information of the environmental data; according to the environmental data and the corresponding time information of the environmental data, using the global preference model to obtain the predicted operating mode of the air conditioner; wherein, the global preference model is obtained and distributed by the aggregation server based on the local training models uploaded by the participating clients using federated learning; according to the environmental data, the corresponding time information of the environmental data and the predicted operating mode of the air conditioner, using the indoor simulation model previously created based on computational fluid dynamics to obtain the predicted result of the air flow distribution; and controlling the flexible air guiding mechanism to adjust the angle according to the predicted result of the air flow distribution.

[0007] It should be noted that by obtaining environmental data and the corresponding time information of the environmental data, basic data is provided for subsequent analysis. Further, the aggregation server uses the global preference model obtained through federated learning based on the local training models uploaded by multiple clients to predict the optimal operating mode of the air conditioner, ensuring the protection of user privacy while improving the adaptability and generalization ability of the model among different users, thereby providing a more personalized air conditioner operating mode to meet the comfort requirements of different users. In addition, an indoor simulation model based on computational fluid dynamics is used to predict the indoor air flow distribution, so as to adjust the blade angles of the flexible air guiding mechanism, optimize the indoor air flow distribution, reduce energy waste, improve the energy efficiency of the air conditioner system, and achieve the coordinated optimization of high energy efficiency, precise air flow control, and user privacy protection.

[0008] According to the intelligent air conditioner control method based on federated learning provided by the present invention, before obtaining the predicted operating mode of the air conditioner by using the global preference model according to the environmental data and the corresponding time information of the environmental data, it includes: the aggregation server obtains the local model parameters of the participating clients, and the participating clients are used to represent the devices or organizations participating in model training; if there is one participating client, the aggregation server, according to the local model parameters, combines with the previously selected algorithm model to obtain the global preference model; if there are at least two participating clients, the aggregation server, according to the local model parameters of each participating client, uses the preset aggregation algorithm to obtain the global model parameters, and combines with the previously selected algorithm model to obtain the global preference model, and sends it to the corresponding air conditioner.

[0009] It should be noted that the aggregation server collects the local model parameters from multiple participating clients to avoid the transmission of user privacy data and the risk of user privacy leakage, and by considering the cases of a single model and multiple models, improves the generalization ability and accuracy of the model. On the basis of multiple models, by adopting the federated learning algorithm, under the premise of protecting user privacy, the user behaviors in different environments are integrated to realize the distributed training and optimization of multi-user preferences, enabling the model to learn more extensive user behavior patterns and improving the adaptability of the model among different users.

[0010] According to the intelligent air conditioner control method based on federated learning provided by the present invention, the aggregation server, according to the local model parameters of each participating client, uses the preset aggregation algorithm to obtain the global model parameters, including: the aggregation server performs weighted averaging on the local model parameters of each participating client to generate the global model parameters; or, the aggregation server averages the local model parameters of each participating client to generate the global model parameters.

[0011] It should be noted that by assigning different weights to different clients to better adapt to the differences between clients, it helps to improve the accuracy of the global model, and the weight strategy can be adjusted according to different scenarios and requirements to optimize the model performance; or, the local model parameters of all participating clients are directly averaged without considering the differences between clients to generate global model parameters, which reduces the complexity and computational cost of the algorithm, ensures that the parameters of each client are treated equally, fully considers the contributions of all participants, and reduces communication overhead.

[0012] According to the present invention, a smart air conditioner control method based on federated learning is provided. The aggregation server obtains the local model parameters of the participating clients, including: the aggregation server sends the initial model parameters to the pre-determined participating clients; the aggregation server receives the local model parameters encrypted and uploaded by each participating client based on its local training data, or monitors the local training of each participating client, sends a parameter acquisition request after the participating client completes local training, and receives the encrypted local model parameters returned by the corresponding participating client based on the corresponding parameter acquisition request; wherein, the local model parameters are obtained by the participating client using the historical environmental data and historical time corresponding to the historical operation adjustment records in its local training data as the input data for training, and using the historical operation adjustment records in its local training data as the labels for training, and performing offline training on the corresponding local model.

[0013] It should be noted that the aggregation server sends the initial model parameters to the pre-determined participating clients to ensure that all clients start training from the same starting point, thereby ensuring the consistency and comparability of the model, and using the local data of each participating client for training to improve the personalized adaptation ability of the model, while reducing the data transmission requirements. After the client completes local training, it uploads the encrypted local model parameters to the aggregation server, protecting user privacy while reducing the communication burden. The aggregation server receives the encrypted local model parameters and uses an aggregation algorithm to integrate these parameters. By integrating the data characteristics of different participating clients, the generalization ability and computational efficiency of the global model are improved. Through distributed training using the computing resources of multiple participating clients, load balancing is achieved, and it is easy to expand and can incorporate more clients, enhancing the adaptability and flexibility of the model.

[0014] According to the present invention, an intelligent air conditioner control method based on federated learning is provided. Before obtaining the predicted air flow distribution result by using the indoor simulation model previously created based on computational fluid dynamics according to the environmental data, the time information corresponding to the environmental data, and the predicted operation mode of the air conditioner, it includes: based on computational fluid dynamics (CFD), constructing an indoor geometric model, meshing the indoor geometric model, and creating a computational network; on the grids of the computational network, setting corresponding boundary conditions, and combining with a preset fluid dynamics model to construct an indoor simulation model.

[0015] It should be noted that by using CFD to construct an accurate geometric model to ensure the geometric consistency between the simulation model and the actual indoor environment, laying a foundation for subsequent accurate analysis, meshing the indoor geometric model to create a computational network to capture fine air flow characteristics through reasonable meshing, improving the accuracy of the simulation results, and setting boundary conditions on the grids of the computational network according to the actual physical situation, enabling the indoor simulation model to more realistically reflect the interaction between the indoor and outdoor, improving the reliability of the prediction, and then combining with a preset fluid dynamics model to construct an indoor simulation model, so that the indoor simulation model can not only provide high-precision air flow distribution prediction, but also provide strong support for the design, optimization, and intelligent control of the air conditioner system, ensuring the performance and efficiency of the system in practical applications.

[0016] According to the present invention, an intelligent air conditioner control method based on federated learning is provided. According to the predicted air flow distribution result, controlling the flexible air deflector mechanism to adjust the angle includes: determining the target blade curvature according to the predicted air flow distribution result, and generating a driving signal according to the target blade curvature; controlling the flexible air deflector mechanism to adjust the corresponding blade curvature according to the driving signal.

[0017] It should be noted that according to the predicted air flow distribution result, potential air flow problems are identified in advance, so as to calculate the target blade curvature required to achieve the ideal air flow distribution according to the prediction result, ensure the accuracy of the blade curvature, contribute to achieving the best air flow distribution, and generating a driving signal according to the target blade curvature to control the flexible air deflector mechanism to adjust the corresponding blade curvature, thereby more finely adjusting the air flow, improving the comfort and energy efficiency of the indoor environment, reducing manual intervention, and enhancing the intelligent level of the system.

[0018] According to the present invention, an intelligent air conditioner control method based on federated learning is provided. The flexible air deflector mechanism is formed by using shape memory alloy (SMA) or electroactive polymer (EAP); the method further includes: monitoring the air flow speed and direction in the room; evaluating whether it conforms to the predicted air flow distribution result according to the air flow speed and direction; based on the fact that the air flow speed and direction do not conform to the predicted air flow distribution result, determining the corresponding blade curvature to be adjusted, and generating a driving signal according to the corresponding blade curvature to be adjusted to control the flexible air deflector mechanism to adjust the corresponding blade curvature to be adjusted.

[0019] It should be noted that by adopting a shape memory alloy (SMA) or an electroactive polymer (EAP), the driving signal can be quickly responded to, so as to achieve stepless continuous adjustment of the curvature of the air guiding vane, accurately control the angle and shape of the air guiding vane, meet the requirements of complex air flow distribution, and improve comfort and energy efficiency; through real-time monitoring and evaluation, the non-uniformity of air flow distribution can be quickly identified and corrected, a more uniform and comfortable indoor environment can be provided, and the user experience can be improved. In addition, by optimizing the air flow distribution, energy waste can be reduced and the energy efficiency of the air conditioning system can be improved. In addition, the monitoring of the indoor air flow speed and direction can be achieved through an air flow sensor, which is not further limited here.

[0020] The present invention also provides an intelligent air conditioning control system based on federated learning, including: a data acquisition module for acquiring environmental data and the time information corresponding to the environmental data; a preference prediction module for obtaining a predicted operating mode of the air conditioner by using a global preference model according to the environmental data and the time information corresponding to the environmental data, wherein the global preference model is obtained and distributed by an aggregation server based on local training models uploaded by participating clients by using federated learning; an air flow prediction module for obtaining a predicted result of air flow distribution by using an indoor simulation model previously created based on computational fluid dynamics according to the environmental data, the time information corresponding to the environmental data, and the predicted operating mode of the air conditioner; and a air guiding control module for controlling a flexible air guiding mechanism to adjust the angle according to the predicted result of air flow distribution.

[0021] It should be noted that the data acquisition module acquires environmental data and the time information corresponding to the environmental data to provide basic data for subsequent analysis, and further, the preference prediction module uses the global preference model obtained by federated learning by the aggregation server based on local training models uploaded by multiple clients to predict the optimal operating mode of the air conditioner, ensuring the adaptability and generalization ability of the model among different users while protecting the privacy of users, so as to provide a more personalized air conditioning operating mode, meet the comfort requirements of different users, and the air flow prediction module uses an indoor simulation model based on computational fluid dynamics to predict the indoor air flow distribution, so as to adjust the blade angle of the flexible air guiding mechanism through the air guiding control module, optimize the indoor air flow distribution, reduce energy waste, and improve the energy efficiency of the air conditioning system, realizing the collaborative optimization of high energy efficiency, accurate air flow regulation and user privacy protection.

[0022] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the steps of any one of the above-mentioned intelligent air conditioning control methods based on federated learning are implemented.

[0023] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the intelligent air conditioner control method based on federated learning as described in any one of the above are implemented. Description of the Drawings

[0024] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0025] Figure 1 is a schematic flowchart of the intelligent air conditioner control method based on federated learning provided by the present invention; Figure 2 is a schematic structural diagram of the intelligent air conditioner control system based on federated learning provided by the present invention; Figure 3 is a schematic structural diagram of the electronic device provided by the present invention. Detailed Embodiments

[0026] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0027] The following will describe Figure 1 a schematic flowchart of an intelligent air conditioner control method based on federated learning of the present invention. The method includes: S11. Obtain environmental data and the time information corresponding to the environmental data; S12. According to the environmental data and the time information corresponding to the environmental data, use the global preference model to obtain the predicted operating mode of the air conditioner; wherein, the global preference model is obtained and distributed by the aggregation server based on the local training models uploaded by the participating clients using federated learning; S13. According to the environmental data, the time information corresponding to the environmental data, and the predicted operating mode of the air conditioner, use the indoor simulation model previously created based on computational fluid dynamics to obtain the predicted result of the air flow distribution; S14. Control the flexible air deflector mechanism to adjust the angle according to the predicted result of the air flow distribution.

[0028] It should be noted that the step numbers "S1N" in this specification do not represent the sequence of the intelligent air conditioner control method based on federated learning. The following specifically describes the intelligent air conditioner control method based on federated learning of the present invention.

[0029] Step S11, obtain environmental data and the time information corresponding to the environmental data.

[0030] It should be added that the environmental data includes indoor temperature, outdoor temperature, indoor humidity, outdoor humidity, air quality, and user distribution data. Obtaining the environmental data includes: using temperature and humidity sensors to collect indoor temperature, indoor humidity, outdoor temperature, and outdoor humidity; using a PM2.5 / CO2 sensor to detect air quality; using an infrared thermal imager to detect the location and density of people to obtain user distribution data.

[0031] In addition, after obtaining the environmental data and the time information corresponding to the environmental data, it includes: based on the Kalman filter algorithm, preprocessing the environmental data and the time information corresponding to the environmental data to remove noise data.

[0032] Step S12, according to the environmental data and the time information corresponding to the environmental data, use the global preference model to obtain the predicted operating mode of the air conditioner; wherein, the global preference model is obtained by the aggregation server based on the local training models uploaded by the participating clients using federated learning and then distributed.

[0033] In this embodiment, according to the environmental data and the time information corresponding to the environmental data, using the global preference model to obtain the predicted operating mode of the air conditioner includes: inputting the environmental data and the time information corresponding to the environmental data into the global preference model for preference prediction to determine the corresponding predicted operating mode of the air conditioner according to the predicted user preferences.

[0034] In an alternative embodiment, before obtaining the predicted operating mode of the air conditioner according to the environmental data and the time information corresponding to the environmental data, it includes: the aggregation server obtains the local model parameters of the participating clients, where the participating clients are used to represent the devices or organizations participating in model training; if there is one participating client, the aggregation server, according to the local model parameters, combines with the previously selected algorithm model to obtain the global preference model; if there are at least two participating clients, the aggregation server, according to the local model parameters of each participating client, uses a preset aggregation algorithm to obtain the global model parameters, combines with the previously selected algorithm model to obtain the global preference model, and distributes it to the corresponding air conditioner.

[0035] It should be noted that the aggregation server collects local model parameters from multiple participating clients to avoid the transmission of user privacy data and the risk of user privacy leakage. By considering the cases of a single model and multiple models, the generalization ability and accuracy of the model are improved. Based on multiple models, by adopting the federated learning algorithm, under the premise of protecting user privacy, the user behaviors in different environments are integrated to achieve distributed training and optimization of multi-user preferences, enabling the model to learn a wider range of user behavior patterns and improving the adaptability of the model among different users.

[0036] Specifically, the aggregation server obtains the local model parameters of the participating clients, including: the aggregation server sends the initial model parameters to the pre-determined participating clients; receives the locally trained and encrypted uploaded local model parameters of each participating client, or the aggregation server monitors the local training of each participating client, sends a parameter acquisition request after the participating client completes local training, and receives the encrypted local model parameters returned by the corresponding participating client based on the corresponding parameter acquisition request; wherein, the local model parameters are obtained by the participating client using the historical environment data and historical time corresponding to the historical operation adjustment records in its local training data as the input data for training, and using the historical operation adjustment records in its local training data as the labels for training, and performing offline training on the corresponding local model. It should be noted that after receiving the encrypted local model parameters, the aggregation server also needs to decrypt them and then aggregate them to obtain the global preference model.

[0037] It should be noted that the aggregation server sends the initial model parameters to the pre-determined participating clients to ensure that all clients start training from the same starting point, thus ensuring the consistency and comparability of the models, and using the local data of each participating client for training to improve the personalized adaptation ability of the model, while reducing the data transmission requirements. After the client completes local training, it uploads the encrypted local model parameters to the aggregation server, reducing the communication burden while protecting user privacy. The aggregation server receives the encrypted local model parameters and uses the aggregation algorithm to integrate these parameters. By integrating the data characteristics of different participating clients, the generalization ability and computational efficiency of the global model are improved. Through distributed training using the computing resources of multiple participating clients, load balancing is achieved, and it is easy to expand, and more clients can be included, enhancing the adaptability and flexibility of the model.

[0038] In addition, the local model can be an existing network built into the training device, which usually includes a network structure, or it can be other networks specified by the user, such as decision trees, random forests, or other neural networks, etc. Specifically, it can be selected according to the actual design requirements of the user, and no further limitation is made here.

[0039] Furthermore, each participating client also uses differential privacy technology to protect the privacy of local data.

[0040] More specifically, each participating client is also used to: monitor whether the historical operation adjustment records in the local training data are updated, and based on the update, obtain the corresponding updated operation adjustment records, corresponding environmental data, and time, and use a preset incremental learning algorithm to perform online updates on the corresponding local model.

[0041] It should be noted that the participating client tracks the air-conditioning operation adjustment records in real time to facilitate the timely identification of changes in user preferences or environmental conditions, and uses a preset incremental learning algorithm for online learning to update the model without retraining the entire model, saving computing resources and improving efficiency, ensuring that the model can learn from the latest data, and guaranteeing the timeliness and accuracy of the local model. In addition, the preset incremental learning algorithm can be selected according to actual online update requirements and design goals. For example, it can be online gradient descent to avoid retraining with all data. Furthermore, during the process of online updating the model, higher weights can be assigned to recent data according to the timeliness of adjustment behaviors. The specific weights can be determined according to prior experience or experimental configurations and are not further limited here. In addition, the aggregation server obtains global model parameters based on the local model parameters of each participating client using a preset aggregation algorithm, including: the aggregation server performs weighted averaging on the local model parameters of each participating client to generate global model parameters; or, the aggregation server averages the local model parameters of each participating client to generate global model parameters.

[0042] It should be noted that by assigning different weights to different clients to better adapt to the differences between clients, it helps to improve the accuracy of the global model, and the weight strategy can be adjusted according to different scenarios and requirements to optimize the model performance; or, directly average the local model parameters of all participating clients without considering the differences between clients to generate global model parameters, reducing the complexity and computational cost of the algorithm, ensuring that the parameters of each client are treated equally, fully considering the contributions of all participants, and reducing communication overhead.

[0043] Step S13: Based on the environmental data, the time information corresponding to the environmental data, and the predicted air-conditioning operation mode, use the indoor simulation model created in advance based on computational fluid dynamics to obtain the predicted result of the airflow distribution.

[0044] In an alternative embodiment, before obtaining the predicted airflow distribution result by using the indoor simulation model previously created based on computational fluid dynamics according to the environmental data, the time information corresponding to the environmental data, and the predicted operation mode of the air conditioner, it includes: constructing an indoor geometric model based on computational fluid dynamics (CFD), performing mesh division on the indoor geometric model to create a computational network; setting corresponding boundary conditions on the meshes of the computational network, and combining with a preset fluid dynamics model to construct an indoor simulation model.

[0045] It should be noted that an accurate geometric model is constructed through CFD to ensure the geometric consistency between the simulation model and the actual indoor environment, laying a foundation for subsequent accurate analysis, and performing mesh division on the indoor geometric model to create a computational network to capture subtle airflow characteristics through reasonable mesh division, improving the accuracy of the simulation results. Also, on the meshes of the computational network, boundary conditions are set according to the actual physical situation, enabling the indoor simulation model to more realistically reflect the interaction between the indoor and outdoor environments, improving the reliability of the prediction. Then, in combination with the preset fluid dynamics model, an indoor simulation model is constructed, enabling the indoor simulation model to not only provide high-precision predicted airflow distribution but also provide strong support for the design, optimization, and intelligent control of the air conditioning system, ensuring the performance and efficiency of the system in actual applications.

[0046] In addition, the boundary conditions include inlet conditions, outlet conditions, and other conditions such as wall conditions, symmetry planes, or periodic boundary conditions. The inlet conditions can include the operation mode of the air conditioner, environmental data, and the corresponding time. The outlet conditions can include the predicted airflow distribution result. The wall conditions can include no-slip or slip conditions and thermal boundary conditions, etc. The specific boundary condition settings can be configured according to the actual physical situation to ensure that these conditions are consistent with the actual physical situation, and no further limitation is made here.

[0047] Correspondingly, when obtaining the predicted airflow distribution result by using the indoor simulation model previously created based on computational fluid dynamics according to the environmental data, the time information corresponding to the environmental data, and the predicted operation mode of the air conditioner, the indoor simulation model solves the fluid dynamics equation to obtain the dynamic behavior of the indoor airflow, and monitors the simulation process to ensure numerical stability and reach the convergence standard to obtain the predicted airflow distribution result.

[0048] Step S14, controlling the flexible air guiding mechanism to adjust the angle according to the predicted airflow distribution result.

[0049] In this embodiment, controlling the flexible air guiding mechanism to adjust the angle according to the predicted airflow distribution result includes: determining the target blade curvature according to the predicted airflow distribution result, and generating a driving signal according to the target blade curvature; controlling the flexible air guiding mechanism to adjust the corresponding blade curvature according to the driving signal.

[0050] It should be noted that according to the predicted results of the air flow distribution, potential air flow problems are identified in advance, so that according to the predicted results, the target blade curvature required to achieve the ideal air flow distribution is calculated to ensure the accuracy of the blade curvature, which helps to achieve the best air flow distribution. And a driving signal is generated according to the target blade curvature to control the flexible air guiding mechanism to adjust the corresponding blade curvature, so as to more finely regulate the air flow, improve the comfort and energy efficiency of the indoor environment, reduce manual intervention, and enhance the intelligent level of the system.

[0051] In an alternative embodiment, the flexible air guiding mechanism is made of shape memory alloy SMA or electroactive polymer EAP.

[0052] It should be noted that by using shape memory alloy SMA or electroactive polymer EAP, it can quickly respond to the driving signal, realize the stepless continuous adjustment of the curvature of the air guiding blade, accurately control the angle and shape of the air guiding blade, meet the requirements of complex air flow distribution, and improve comfort and energy efficiency.

[0053] In addition, a hydrophobic nano - coating is formed on the blade surface to reduce the accumulation of condensate water, improve the air flow stability, and the blade is driven by a micro - electrode to achieve precise control of the bending angle of the blade (0° - 120°).

[0054] In an alternative embodiment, the method further includes: monitoring the air flow speed and direction in the room; evaluating whether it conforms to the predicted results of the air flow distribution according to the air flow speed and direction; based on the fact that the air flow speed and direction do not conform to the predicted results of the air flow distribution, determining the corresponding blade curvature to be adjusted, and generating a driving signal according to the corresponding blade curvature to be adjusted to control the flexible air guiding mechanism to adjust the corresponding blade curvature to be adjusted.

[0055] It should be noted that through real - time monitoring and evaluation, the non - uniformity of the air flow distribution can be quickly identified and corrected, providing a more uniform and comfortable indoor environment and improving the user experience. In addition, by optimizing the air flow distribution, energy waste is reduced and the energy efficiency of the air - conditioning system is improved. In addition, the monitoring of the air flow speed and direction in the room can be realized by an air flow sensor, which is not further limited here.

[0056] In summary, the embodiments of the present invention obtain environmental data and the corresponding time information of the environmental data to provide basic data for subsequent analysis, and further use an aggregation server to obtain a global preference model through federated learning based on local training models uploaded by multiple clients, and predict the optimal operating mode of the air conditioner, ensuring the adaptability and generalization ability of the model among different users while protecting user privacy, thereby providing a more personalized air conditioner operating mode to meet the comfort requirements of different users, and using an indoor simulation model based on computational fluid dynamics to predict the indoor air flow distribution, so as to adjust the blade angles of the flexible air guiding mechanism, optimize the indoor air flow distribution, reduce energy waste, improve the energy efficiency of the air conditioning system, and achieve the coordinated optimization of high energy efficiency, precise air flow control and user privacy protection.

[0057] Next, the intelligent air conditioner control device based on federated learning provided by the present invention will be described. The intelligent air conditioner control device based on federated learning described below can be correspondingly referred to the federated learning-based intelligent air conditioner control method described above.

[0058] Figure 2 FIG. shows a schematic structural diagram of an intelligent air conditioner control system based on federated learning. The system includes: A data acquisition module 21, which acquires environmental data and the corresponding time information of the environmental data; A preference prediction module 22, which obtains a predicted operating mode of the air conditioner according to the environmental data and the corresponding time information of the environmental data by using the global preference model; wherein, the global preference model is obtained and distributed by an aggregation server through federated learning based on local training models uploaded by participating clients; An air flow prediction module 23, which obtains a predicted air flow distribution result according to the environmental data, the corresponding time information of the environmental data and the predicted operating mode of the air conditioner by using an indoor simulation model created in advance based on computational fluid dynamics; An air guiding control module 24, which controls the flexible air guiding mechanism to adjust the angle according to the predicted air flow distribution result.

[0059] In this embodiment, the environmental data includes indoor temperature, outdoor temperature, indoor humidity, outdoor humidity, air quality and user distribution data. The data acquisition module 21 is used for: collecting indoor temperature, indoor humidity, outdoor temperature and outdoor humidity by using temperature and humidity sensors; detecting air quality by using PM2.5 / CO2 sensors; detecting the positions and densities of personnel by using an infrared thermal imager to obtain user distribution data.

[0060] In addition, the system further includes: a preprocessing module, which preprocesses the environmental data and the corresponding time information of the environmental data based on the Kalman filter algorithm to remove noise data after obtaining the environmental data and the corresponding time information of the environmental data.

[0061] A preference prediction module 22, configured to: input environmental data and the time information corresponding to the environmental data into a global preference model for preference prediction, so as to determine a corresponding predicted operation mode of the air conditioner according to the predicted user preference.

[0062] In an optional embodiment, the system further includes: a model distribution module. Before obtaining the predicted operation mode of the air conditioner by using the global preference model according to the environmental data and the time information corresponding to the environmental data, the aggregation server obtains the local model parameters of the participating clients, where the participating clients are used to represent devices or organizations participating in model training; if there is one participating client, the aggregation server, according to the local model parameters, combines with the previously selected algorithm model to obtain the global preference model; if there are at least two participating clients, the aggregation server, according to the local model parameters of each participating client, uses a preset aggregation algorithm to obtain the global model parameters, and combines with the previously selected algorithm model to obtain the global preference model, and distributes it to the corresponding air conditioner.

[0063] Specifically, the model distribution module is further configured to: the aggregation server sends the initial model parameters to the pre-determined participating clients; receive the local model parameters encrypted and uploaded by each participating client based on its local training data, or the aggregation server monitors the local training of each participating client, and sends a parameter acquisition request after the participating client completes local training, and receives the encrypted local model parameters returned by the corresponding participating client based on the corresponding parameter acquisition request; where the local model parameters are obtained by the participating client using the historical environmental data and historical time corresponding to the historical operation adjustment records in its local training data as the input data for training, and using the historical operation adjustment records in its local training data as the label for training, and performing offline training on the corresponding local model. It should be noted that after receiving the encrypted local model parameters, the aggregation server also needs to decrypt them and then aggregate them to obtain the global preference model.

[0064] In addition, each participating client also uses differential privacy technology to protect the privacy of local data.

[0065] Furthermore, each participating client is further configured to: monitor whether the historical operation adjustment records in the local training data are updated, and based on the update, obtain the corresponding updated operation adjustment records, corresponding environmental data and time, and use a preset incremental learning algorithm to perform online update on the corresponding local model.

[0066] In addition, the model distribution module is further configured to: the aggregation server performs weighted averaging on the local model parameters of each participating client to generate global model parameters; or the aggregation server performs averaging on the local model parameters of each participating client to generate global model parameters.

[0067] In an alternative embodiment, the system further includes a simulation model construction module. Before obtaining the predicted air flow distribution result by using the indoor simulation model previously created based on computational fluid dynamics according to the environmental data, the time information corresponding to the environmental data, and the predicted operation mode of the air conditioner, a indoor geometric model is constructed based on computational fluid dynamics (CFD), and the indoor geometric model is meshed to create a computational network. Corresponding boundary conditions are set on the grids of the computational network, and an indoor simulation model is constructed in combination with a preset fluid dynamics model.

[0068] The air deflector control module 24 includes: a signal generation unit that determines a target blade curvature according to the predicted air flow distribution result and generates a drive signal according to the target blade curvature; and an air deflector control unit that controls the flexible air deflector mechanism to adjust the corresponding blade curvature according to the drive signal.

[0069] In an alternative embodiment, the flexible air deflector mechanism is made of shape memory alloy (SMA) or electroactive polymer (EAP).

[0070] In addition, a hydrophobic nano-coating is formed on the blade surface to reduce the accumulation of condensed water, improve the air flow stability, and the blade is driven by a microelectrode to achieve precise control of the blade bending angle (0° - 120°).

[0071] In an alternative embodiment, the system further includes: a monitoring module that monitors the air flow speed and direction in the room; an evaluation module that evaluates whether it conforms to the predicted air flow distribution result according to the air flow speed and direction; and an adjustment module that determines the corresponding blade curvature to be adjusted based on the fact that the air flow speed and direction do not conform to the predicted air flow distribution result, and generates a drive signal according to the corresponding blade curvature to be adjusted to control the flexible air deflector mechanism to adjust the corresponding blade curvature to be adjusted.

[0072] In summary, in the embodiment of the present invention, the data acquisition module acquires environmental data and the time information corresponding to the environmental data to provide basic data for subsequent analysis. Further, the preference prediction module uses the global preference model obtained by federated learning based on the local training models uploaded by multiple clients by the aggregation server to predict the best operation mode of the air conditioner, ensuring the adaptability and generalization ability of the model among different users while protecting user privacy, thereby providing a more personalized air conditioner operation mode to meet the comfort requirements of different users. And the air flow prediction module uses an indoor simulation model based on computational fluid dynamics to predict the indoor air flow distribution, so as to adjust the blade angle of the flexible air deflector mechanism through the air deflector control module, optimize the indoor air flow distribution, reduce energy waste, improve the energy efficiency of the air conditioner system, and achieve the coordinated optimization of high energy efficiency, precise air flow regulation, and user privacy protection.

[0073] Figure 3 Schematically shows a physical structure diagram of an electronic device, such asFigure 3 As shown in Figure 3 , the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340. Among them, the processor 310, the communications interface 320, and the memory 330 complete communication with each other through the communication bus 340. The processor 310 may call the logical instructions in the memory 330 to execute an intelligent air conditioner control method based on federated learning. The method includes: obtaining environmental data and the time information corresponding to the environmental data; according to the environmental data and the time information corresponding to the environmental data, using a global preference model to obtain a predicted air conditioner operation mode; wherein, the global preference model is obtained and distributed by an aggregation server based on local training models uploaded by participating clients using federated learning; according to the environmental data, the time information corresponding to the environmental data, and the predicted air conditioner operation mode, using an indoor simulation model previously created based on computational fluid dynamics to obtain a predicted result of air flow distribution; and according to the predicted result of air flow distribution, controlling a flexible air deflector mechanism to adjust the angle.

[0074] In addition, when the logical instructions in the above-mentioned memory 330 can be implemented in the form of software function units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc that can store program codes.

[0075] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the intelligent air conditioner control method based on federated learning provided by each of the above methods. The method includes: obtaining environmental data and time information corresponding to the environmental data; according to the environmental data and the time information corresponding to the environmental data, using a global preference model to obtain a predicted operation mode of the air conditioner; wherein, the global preference model is obtained and distributed by an aggregation server based on local training models uploaded by participating clients using federated learning; according to the environmental data, the time information corresponding to the environmental data, and the predicted operation mode of the air conditioner, using an indoor simulation model created in advance based on computational fluid dynamics to obtain a predicted result of air flow distribution; and according to the predicted result of air flow distribution, controlling a flexible air deflector mechanism to adjust the angle.

[0076] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute the intelligent air conditioner control method based on federated learning provided by each of the above. The method includes: obtaining environmental data and time information corresponding to the environmental data; according to the environmental data and the time information corresponding to the environmental data, using a global preference model to obtain a predicted operation mode of the air conditioner; wherein, the global preference model is obtained and distributed by an aggregation server based on local training models uploaded by participating clients using federated learning; according to the environmental data, the time information corresponding to the environmental data, and the predicted operation mode of the air conditioner, using an indoor simulation model created in advance based on computational fluid dynamics to obtain a predicted result of air flow distribution; and according to the predicted result of air flow distribution, controlling a flexible air deflector mechanism to adjust the angle.

[0077] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0078] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent air conditioning control method based on federated learning, characterized in that: include: Acquire environmental data and time information corresponding to the environmental data; According to the environmental data and the time information corresponding to the environmental data, the predicted operation mode of the air conditioner is obtained by using a global preference model; wherein the global preference model is obtained and issued by the aggregation server based on the local training model uploaded by the participating client by using federated learning; According to the environmental data, the time information corresponding to the environmental data and the predicted operation mode of the air conditioner, using the indoor simulation model previously created based on computational fluid dynamics, an air flow distribution prediction result is obtained; According to the airflow distribution prediction result, the flexible air guide mechanism is controlled to adjust the angle.

2. The intelligent air conditioning control method based on federated learning according to claim 1, characterized in that: Before obtaining the predicted operation mode of the air conditioner by using the global preference model according to the environmental data and the time information corresponding to the environmental data, the method includes: The aggregation server obtains local model parameters of participating clients, where the participating clients are used to represent devices or organizations participating in model training; If there is only one participating client, the aggregation server obtains a global preference model based on the local model parameters and the previously selected algorithm model; If there are at least two participating clients, the aggregation server obtains global model parameters based on the local model parameters of each participating client using a preset aggregation algorithm, and obtains a global preference model in combination with the previously selected algorithm model, and sends it to the corresponding air conditioner.

3. The intelligent air conditioning control method based on federated learning according to claim 2 is characterized in that: The aggregation server obtains global model parameters according to the local model parameters of each participating client using a preset aggregation algorithm, including: The aggregation server performs weighted averaging of the local model parameters of each participating client to generate a global model parameter; or The aggregation server averages the local model parameters of each participating client to generate a global model parameter.

4. The intelligent air conditioning control method based on federated learning according to claim 2, characterized in that: The aggregation server obtains local model parameters of participating clients, including: The aggregation server sends the initial model parameters to the predetermined participating clients; the aggregation server receives the local model parameters trained and encrypted uploaded by each participating client based on its local training data, or monitors the local training of each participating client, and sends a parameter acquisition request after the participating client completes the local training, and receives the encrypted local model parameters returned by the corresponding participating client based on the corresponding parameter acquisition request; wherein the local model parameters are obtained by the participating client using the historical environment data and historical time corresponding to the historical operation adjustment records in its local training data as input data for training, and the historical operation adjustment records in its local training data as labels for training, and performing offline training on the corresponding local model.

5. The intelligent air conditioning control method based on federated learning according to claim 1, characterized in that: Before obtaining the airflow distribution prediction result according to the environmental data, the time information corresponding to the environmental data and the predicted air conditioning operation mode, using the indoor simulation model previously created based on computational fluid dynamics, the method includes: Based on computational fluid dynamics (CFD), an indoor geometric model is constructed, and the indoor geometric model is meshed to create a computational network; On the grid of the computing network, corresponding boundary conditions are set, and an indoor simulation model is constructed in combination with a preset fluid dynamics model.

6. The intelligent air conditioning control method based on federated learning according to any one of claims 1 to 5, characterized in that: According to the airflow distribution prediction result, controlling the flexible air guide mechanism to adjust the angle includes: Determining a target blade curvature according to the airflow distribution prediction result, and generating a driving signal according to the target blade curvature; According to the driving signal, the flexible air guiding mechanism is controlled to adjust the curvature of the corresponding blade.

7. The intelligent air conditioning control method based on federated learning according to claim 6, characterized in that: The flexible air guide mechanism is formed by shape memory alloy SMA or electroactive polymer EAP; The method further comprises: Monitor airflow speed and direction in the room; According to the airflow speed and direction, evaluating whether it conforms to the airflow distribution prediction result; Based on the fact that the airflow velocity and direction do not conform to the airflow distribution prediction result, the corresponding blade curvature to be adjusted is determined, and according to the corresponding blade curvature to be adjusted, a driving signal is generated to control the flexible air guide mechanism to adjust the corresponding blade curvature to be adjusted.

8. An intelligent air conditioning control system based on federated learning, characterized in that: include: A data acquisition module, which acquires environmental data and time information corresponding to the environmental data; A preference prediction module, which obtains the predicted operation mode of the air conditioner by using a global preference model according to the environmental data and the time information corresponding to the environmental data; wherein the global preference model is obtained and issued by the aggregation server by using federated learning based on the local training model uploaded by the participating client; An airflow prediction module, which obtains an airflow distribution prediction result by using an indoor simulation model previously created based on computational fluid dynamics according to the environmental data, time information corresponding to the environmental data and the predicted air conditioning operation mode; The air guide control module controls the flexible air guide mechanism to adjust the angle according to the airflow distribution prediction result.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the intelligent air-conditioning control method based on federated learning as described in any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent air-conditioning control method based on federated learning as described in any one of claims 1 to 7 are implemented.

Images