Intelligent air conditioner temperature control method and system

Abstract

The invention discloses an intelligent air conditioner temperature control method and system, and relates to the technical field of temperature control, and the intelligent air conditioner temperature control method comprises the steps that an initial air conditioner temperature control model is sent to a control cloud, and a target air conditioner temperature control model sent by the control cloud is received, indoor environment state information, indoor personnel information, air conditioner state information, energy storage battery information, electrical load information, power grid electricity price information and current time period information are obtained, and a target air conditioner temperature control strategy is determined through a target air conditioner temperature control model; the temperature adjustment value and the energy storage control information are determined based on the collected information through the target air conditioner temperature control strategy, intelligent air conditioner temperature control is conducted according to the temperature adjustment value and the energy storage control information, and the electricity utilization efficiency is comprehensively optimized in the air conditioner temperature control process; the load of a power grid is balanced on the premise of keeping the comfort of a user.

Description

Technical Field

[0001] The present application relates to the field of temperature control technology, and in particular to an intelligent air-conditioning temperature control method and system. Background Art

[0002] With the growth of global energy consumption and increasingly severe environmental problems, energy conservation and emission reduction have become important issues in social development. In particular, in building energy consumption, air conditioning systems, as one of the main energy-consuming equipment, account for a considerable proportion of energy consumption.

[0003] Existing solutions mainly control air conditioning temperature based on the static Predicted Mean Vote (PMV) indicator. The comfort of the indoor environment is calculated through a series of thermal comfort factors (such as temperature, humidity, wind speed, clothing thermal resistance, metabolic rate, etc.) to predict a person's thermal sensation in the environment, and then provide a certain degree of comfort control in the process of controlling the air conditioning temperature. However, when faced with real-time changing grid demand and user behavior patterns, it is unable to dynamically consider the balance between power load and human comfort.

[0004] How to comprehensively optimize electricity efficiency in the process of controlling air-conditioning temperature so as to balance the grid load while maintaining user comfort has become a problem to be solved.

[0005] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are prior art. Summary of the invention

[0006] The main purpose of this application is to provide an intelligent air-conditioning temperature control method and system, aiming to solve the technical problem of how to comprehensively optimize the power efficiency in the process of controlling the air-conditioning temperature so as to balance the grid load while maintaining user comfort.

[0007] To achieve the above objectives, the present application proposes an intelligent air-conditioning temperature control method, which is applied to an edge terminal and includes:

[0008] Sending an initial air-conditioning temperature control model to a control cloud, so that the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receiving the target air-conditioning temperature control model sent by the control cloud;

[0009] Obtain indoor environment status information, indoor personnel information, air conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information;

[0010] Determine a target air conditioning temperature control strategy through the target air conditioning temperature control model, and determine a temperature adjustment value and energy storage control information through the target air conditioning temperature control strategy based on the indoor environment status information, the indoor personnel information, the air conditioning status information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information;

[0011] Intelligent air conditioning temperature control is performed according to the temperature adjustment value and the energy storage control information.

[0012] In one embodiment, the step of determining the temperature adjustment value and the energy storage control information based on the indoor environment status information, the indoor personnel information, the air conditioning status information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information through the target air conditioning temperature control strategy includes:

[0013] Determine a comfort reward based on the indoor environment state information and the indoor personnel information;

[0014] Determining an energy efficiency reward based on the air conditioning status information and the indoor environment status information;

[0015] Determining a grid load balancing reward based on the power load information and current time period information;

[0016] Determine an electricity price optimization reward based on the energy storage battery information, the grid electricity price information and the current time period information;

[0017] The temperature adjustment value and energy storage control information are determined through the target air-conditioning temperature control strategy based on the comfort reward, the energy efficiency reward, the grid load balancing reward and the electricity price optimization reward.

[0018] In one embodiment, the step of determining a comfort reward based on the indoor environment state information and the indoor personnel information includes:

[0019] Obtain thermal comfort prediction model;

[0020] Determine indoor temperature information, indoor humidity information and indoor wind speed information based on the indoor environment status information;

[0021] Determine the number of people, clothing information and activity status of people based on the indoor people information;

[0022] Predicting a thermal comfort value based on the indoor temperature information, the indoor humidity information, the indoor wind speed information, the number of people information, the clothing information of the people, and the activity status of the people by using the thermal comfort prediction model;

[0023] A comfort reward is calculated based on the thermal comfort value by a comfort reward function.

[0024] In one embodiment, before the step of obtaining the thermal comfort prediction model, the method further includes:

[0025] Obtain thermal comfort dataset and initialize deep neural network model;

[0026] Initialize a particle swarm, and use the number of hidden layers, the number of neurons in each layer, the learning rate, and the activation function of the deep neural network model as particles of the particle swarm;

[0027] Performing fitness evaluation on particles in the particle swarm based on the thermal comfort data set to obtain a fitness evaluation result;

[0028] updating the particle swarm based on the fitness evaluation result until there is a particle in the particle swarm whose fitness evaluation result reaches a preset threshold, and taking the particle whose fitness evaluation result reaches the preset threshold as a target particle;

[0029] A thermal comfort prediction model is determined based on the target particles.

[0030] In one embodiment, before the step of sending the initial air conditioning temperature control model to the control cloud so that the control cloud generates a target air conditioning temperature control model based on the initial air conditioning temperature control model and receives the target air conditioning temperature control model sent by the control cloud, the step further includes:

[0031] Initialize the agent, evaluation network, target network, and experience replay pool;

[0032] Acquire local air-conditioning temperature control data, and update the experience replay pool of the agent based on the local air-conditioning temperature control data;

[0033] Acquire experience samples from the experience replay pool, and train and update the evaluation network based on the experience samples;

[0034] The target network is optimized based on the evaluation network to obtain an initial air conditioning temperature control model.

[0035] In one embodiment, the step of obtaining local air-conditioning temperature control data and updating the experience replay pool of the agent based on the local air-conditioning temperature control data includes:

[0036] Acquire local air-conditioning temperature control data, and define the current state of the agent based on the local air-conditioning temperature control data;

[0037] Taking temperature adjustment and energy storage charging and discharging control as the action space of the intelligent agent;

[0038] Construct the comfort reward function, energy efficiency reward function, grid load balance reward function and electricity price optimization reward function to obtain the instant reward function;

[0039] Determine, based on the instant reward function, a current reward obtained by the agent when executing the target action in the action space in the current state, and obtain a next state;

[0040] The current state, the target action, the current reward and the next state are used as current experience, and the experience replay pool of the agent is updated based on the current experience.

[0041] In one embodiment, the step of obtaining experience samples from the experience replay pool and training and updating the evaluation network based on the experience samples includes:

[0042] Calculating the predicted return of the experience sample based on the evaluation network;

[0043] Calculating a target return for the experience sample based on the target network;

[0044] Calculating a loss value of the evaluation network according to the predicted return and the target return;

[0045] The evaluation network is updated based on the loss value by a gradient descent method.

[0046] In addition, to achieve the above purpose, the present application also proposes a smart air conditioner temperature control method, which is applied to control the cloud, and the method includes:

[0047] Receive the initial air conditioning temperature control model sent by the edge terminal;

[0048] Determining the contribution of the edge terminal based on the initial air conditioning temperature control model;

[0049] Performing weighted aggregation on the edge terminals based on the contribution to obtain a target air conditioning temperature control model;

[0050] The target air-conditioning temperature control model is sent to the edge terminal.

[0051] In one embodiment, the step of determining the contribution of the edge terminal based on the initial air conditioning temperature control model includes:

[0052] Eliminating the edge terminal from the edge terminal set to obtain an edge terminal subset;

[0053] Aggregating the initial air-conditioning temperature control models in the edge terminal set to obtain a first aggregated model;

[0054] Aggregating the initial air-conditioning temperature control models in the edge terminal subset to obtain a second aggregated model;

[0055] The contribution of the edge terminal is determined according to a performance evaluation difference between the first aggregation model and the second aggregation model.

[0056] In addition, to achieve the above objectives, the present application also proposes an intelligent air-conditioning temperature control system, which includes the control cloud mentioned above and multiple edge terminals mentioned above.

[0057] In addition, to achieve the above purpose, the present application also proposes an intelligent air-conditioning temperature control device, which is applied to an edge terminal. The intelligent air-conditioning temperature control device includes:

[0058] A model feedback module is used to send an initial air-conditioning temperature control model to a control cloud, so that the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receives the target air-conditioning temperature control model sent by the control cloud;

[0059] Information acquisition module, used to obtain indoor environment status information, indoor personnel information, air conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information;

[0060] an information calculation module, configured to determine a target air-conditioning temperature control strategy through the target air-conditioning temperature control model, and determine a temperature adjustment value and energy storage control information through the target air-conditioning temperature control strategy based on the indoor environment status information, the indoor personnel information, the air-conditioning status information, the energy storage battery information, the power load information, the grid electricity price information, and the current time period information;

[0061] The temperature control module is used to perform intelligent air conditioning temperature control according to the temperature adjustment value and the energy storage control information.

[0062] In addition, to achieve the above purpose, the present application also proposes an intelligent air-conditioning temperature control device, which is applied to control the cloud, and the intelligent air-conditioning temperature control device includes:

[0063] A model receiving module, used to receive the initial air conditioning temperature control model sent by the edge terminal;

[0064] A weight allocation module, used to determine the contribution of the edge terminal based on the initial air-conditioning temperature control model;

[0065] A model aggregation module, used to perform weighted aggregation on the edge terminals based on the contribution to obtain a target air conditioning temperature control model;

[0066] The model distribution module is used to send the target air-conditioning temperature control model to the edge terminal.

[0067] In addition, to achieve the above-mentioned purpose, the present application also proposes an intelligent air-conditioning temperature control device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the intelligent air-conditioning temperature control method as described above.

[0068] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the intelligent air conditioning temperature control method described above are implemented.

[0069] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the intelligent air conditioning temperature control method as described above are implemented.

[0070] One or more technical solutions proposed in this application have at least the following technical effects:

[0071] By sending an initial air-conditioning temperature control model to the control cloud, the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receives the target air-conditioning temperature control model sent by the control cloud, obtains indoor environment status information, indoor personnel information, air-conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information, determines a target air-conditioning temperature control strategy through the target air-conditioning temperature control model, and uses the target air-conditioning temperature control strategy to control the temperature of the energy storage battery based on the indoor environment status information, the indoor personnel information, the air-conditioning status information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information. The temperature adjustment value and the energy storage control information are determined based on the battery information, the power load information, the power grid electricity price information and the current time period information. Intelligent air-conditioning temperature control is performed according to the temperature adjustment value and the energy storage control information. The air-conditioning control model is optimized through federated learning, and the air-conditioning temperature and power consumption strategy are dynamically optimized through intelligent control combined with multi-dimensional information (such as indoor environment, personnel activities, electricity prices, etc.). Under the premise of ensuring user comfort, the energy efficiency of the air conditioner is optimized, the burden on the power grid is reduced, and the electricity price fluctuation and the energy storage system are used to further reduce the electricity cost, thereby achieving comprehensive optimization of the power consumption efficiency in the process of controlling the air-conditioning temperature. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0073] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0074] Figure 1 This is a flow chart of the first embodiment of the intelligent air-conditioning temperature control method applied to the edge terminal of the present application;

[0075] Figure 2 This is a cloud-edge collaborative system architecture diagram of the first embodiment of the intelligent air-conditioning temperature control method applied to edge terminals of this application;

[0076] Figure 3 This is a flow chart of a second embodiment of the intelligent air-conditioning temperature control method applied to an edge terminal of the present application;

[0077] Figure 4 This is a reinforcement learning scenario diagram of the second embodiment of the intelligent air-conditioning temperature control method applied to the edge terminal of the present application;

[0078] Figure 5 This is a flow chart of the first embodiment of the intelligent air-conditioning temperature control method applied to control the cloud in this application;

[0079] Figure 6 This is a schematic diagram of the module structure of an intelligent air-conditioning temperature control identification device applied to an edge terminal according to an embodiment of the present application;

[0080] Figure 7 This is a schematic diagram of the module structure of an intelligent air-conditioning temperature control identification device applied to control the cloud in an embodiment of the present application;

[0081] Figure 8 Schematic diagram of the device structure of the hardware operating environment involved in the driving behavior recognition method in the embodiment of the present application.

[0082] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0083] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0084] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0085] The main solution of the embodiment of the present application is: by sending an initial air-conditioning temperature control model to the control cloud, so that the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receives the target air-conditioning temperature control model sent by the control cloud, obtains indoor environmental status information, indoor personnel information, air-conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information, determines a target air-conditioning temperature control strategy through the target air-conditioning temperature control model, and determines a temperature adjustment value and energy storage control information based on the indoor environmental status information, the indoor personnel information, the air-conditioning status information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information through the target air-conditioning temperature control strategy, and performs intelligent air-conditioning temperature control according to the temperature adjustment value and the energy storage control information.

[0086] In this embodiment, for ease of description, the following description is made by taking identification of the intelligent air-conditioning temperature control system as the execution subject.

[0087] Since the existing scheme is mainly based on the static Predicted Mean Vote (PMV) indicator for air conditioning temperature control, the comfort of the indoor environment is calculated through a series of thermal comfort factors (such as temperature, humidity, wind speed, clothing thermal resistance, metabolic rate, etc.) to predict a person's thermal sensation in the environment, and then provide a certain degree of comfort control in the process of controlling the air conditioning temperature. However, when faced with real-time changing grid demand and user behavior patterns, it is unable to dynamically consider the balance between power load and human comfort.

[0088] How to comprehensively optimize electricity efficiency in the process of controlling air-conditioning temperature so as to balance the grid load while maintaining user comfort has become a problem to be solved.

[0089] This application provides a solution that optimizes the air-conditioning control model through federated learning, and combines multi-dimensional information (such as indoor environment, human activities, electricity prices, etc.) through intelligent control to achieve dynamic optimization of air-conditioning temperature and power consumption strategy. While ensuring user comfort, it optimizes the energy efficiency of air conditioning and reduces the burden on the power grid. It also uses electricity price fluctuations and energy storage systems to further reduce electricity costs, thereby achieving comprehensive optimization of electricity efficiency in the process of controlling air-conditioning temperature.

[0090] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, an intelligent air conditioning temperature control system, etc. The following takes the intelligent air conditioning temperature control system as an example to illustrate this embodiment and the following embodiments.

[0091] Based on this, the embodiment of the present application provides an intelligent air conditioning temperature control method, which is applied to an edge terminal, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the intelligent air-conditioning temperature control method of the present application.

[0092] In this embodiment, the intelligent air conditioning temperature control method includes steps S10 to S40:

[0093] Step S10, sending an initial air-conditioning temperature control model to a control cloud, so that the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receives the target air-conditioning temperature control model sent by the control cloud;

[0094] It should be understood that the intelligent air conditioning temperature control method of the present application is proposed based on the cloud-edge collaboration background and serverless architecture. The system architecture under the cloud-edge collaboration background includes a cloud application platform layer, an edge platform layer, and an edge node layer. Figure 2 The cloud-edge collaborative system architecture diagram of the first embodiment of the intelligent air conditioning temperature control method applied to edge terminals in this application. The edge node layer includes multiple edge nodes, each edge node is connected to multiple devices or sensors (such as intelligent air conditioners, temperature sensors, humidity sensors, cameras, and battery monitoring systems, etc.) deployed at the user's location, and the edge node is used to collect environmental status data and other information that affects air conditioning control and perform specific air conditioning control tasks such as temperature adjustment and energy efficiency management; the edge platform layer is located between the cloud and the edge node, and includes multiple edge terminals corresponding to the edge nodes. The edge terminal is used to collect real-time data from the edge node, and the data collected by the edge node is learned through reinforcement learning to build an initial air conditioning temperature control model; the cloud application platform layer corresponds to the control cloud, which is connected to multiple edge terminals, and is used to receive the initial air conditioning temperature control model sent from multiple edge terminals, and aggregates multiple initial air conditioning temperature control models through federated learning to obtain a global optimal target air conditioning temperature control model, and the target air conditioning temperature control model is sent to multiple edge terminals.

[0095] In addition, it should be understood that under the serverless architecture, the management and operation and maintenance of the server are entirely the responsibility of the cloud service provider (such as AWS, Azure, Google Cloud, etc.), and the developer does not need to intervene directly. The application of the serverless architecture will automatically expand and shrink according to the traffic, and the cloud provider will dynamically allocate computing resources according to the number of requests without the need for manual adjustment by the developer. Specifically, at the edge platform layer, the edge terminal processes data and tasks from the edge node. Through load distribution, it can ensure that the computing tasks are evenly distributed to multiple edge terminals, rather than concentrated on a certain node, which can avoid single-point overload, improve the computing power of the entire system, and ensure that the computing needs at different time periods and locations are effectively responded to. The cloud application platform layer will intelligently assign tasks to the nodes with the best performance according to the computing resources, network bandwidth, latency and other indicators of each edge terminal, avoid excessive consumption of resources by some nodes, and ensure the optimization of system resource utilization. At the same time, through containerization technology, applications can be packaged into containers, deployed and run between different edge terminals, and container warehouses are used to manage these container images, store different application versions and service dependencies, and ensure that containers can be quickly and flexibly migrated and deployed between edge terminals. When the computing load of an edge terminal increases, more container instances can be dynamically deployed or tasks can be migrated to terminals with stronger computing power to ensure that computing needs are met.

[0096] It should be noted that before building the initial air-conditioning temperature control model, the control cloud will first initialize a global model, and send the global model containing the model parameters (such as weights and biases, etc.) and structure (such as the layer type of the neural network, the layer connection method, the number of layer nodes, and the activation function, etc.) to each edge terminal. Each edge terminal will first train and learn the global model based on the local data collected in real time through a reinforcement learning algorithm near the data source to obtain the initial air-conditioning temperature control model.

[0097] Step S20, obtaining indoor environment status information, indoor personnel information, air conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information;

[0098] It should be understood that the indoor environmental status information includes environmental data such as temperature, humidity, and wind speed that affect the human body's perception of thermal comfort, which can be obtained through sensors arranged indoors. Indoor personnel information can specifically be indoor personnel image information, which can be obtained through cameras arranged indoors. Energy storage battery information includes the charging power and charge state of the energy storage battery, which can be obtained through the battery monitoring system. The power load information includes the current power load of the power grid and the maximum power load of the power grid, which can be determined through the information of the smart grid system and the smart meter. The power grid electricity price information can be obtained through the real-time electricity price provided by the power company or the smart grid system. The current time period information is obtained through the clock in the system of the edge device. The electricity price of the current time period (such as time-of-use electricity price, peak electricity price, valley electricity price, etc.) can be determined by combining the power grid electricity price information and the current time period information.

[0099] Step S30, determining a target air-conditioning temperature control strategy through the target air-conditioning temperature control model, and determining a temperature adjustment value and energy storage control information through the target air-conditioning temperature control strategy based on the indoor environment state information, the indoor personnel information, the air-conditioning state information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information;

[0100] It should be noted that after receiving the target air-conditioning temperature control model sent by the control cloud, the evaluation network in the initial air-conditioning temperature control model will be optimized according to the gradient information of the target air-conditioning temperature control model, and the reward return for performing a certain action (such as temperature adjustment and energy storage charging and discharging control) under a given state (the state obtained according to the indoor environment state information, indoor personnel information, air-conditioning state information, energy storage battery information, power load information, grid electricity price information and current time period information) is determined according to the evaluation network, so that the action with the highest reward return is used as the next action through the target air-conditioning temperature control strategy to obtain the temperature adjustment value and energy storage control information.

[0101] In a feasible implementation, step S30 may include steps S31 to S35:

[0102] Step S31, determining a comfort reward based on the indoor environment state information and the indoor personnel information;

[0103] It should be understood that the comfort reward is used to ensure that the indoor temperature remains within a range that is comfortable for the human body. By processing the indoor environmental status information and the indoor occupant information and converting the non-numerical data therein into quantitative data, the key data that affects the thermal comfort of the indoor occupants can be obtained. The thermal comfort value of the human body is determined based on the key data through the thermal comfort calculation formula or the thermal comfort prediction model, and the comfort reward is determined based on the thermal comfort value through the comfort reward function. Among them, the comfort reward function has been determined when the initial air conditioning temperature control model is obtained.

[0104] In a feasible implementation, step S31 may include steps A11 to A15:

[0105] Step A11, obtaining a thermal comfort prediction model;

[0106] It should be understood that by calculating these key information through the thermal comfort prediction model, the thermal comfort prediction value of indoor occupants can be obtained, and the comfort reward can be calculated based on the thermal comfort prediction value to ensure the comfort experience of indoor occupants during the temperature control process of intelligent air conditioning.

[0107] In a feasible implementation, before step A11, it may include: obtaining a thermal comfort data set and initializing a deep neural network model; initializing a particle swarm, and using the number of hidden layers, the number of layer neurons, the learning rate, and the activation function of the deep neural network model as particles of the particle swarm; performing fitness evaluation on the particles in the particle swarm based on the thermal comfort data set to obtain a fitness evaluation result; updating the particle swarm based on the fitness evaluation result until there are particles in the particle swarm whose fitness evaluation results reach a preset threshold, and using the particles whose fitness evaluation results reach the preset threshold as target particles; and determining a thermal comfort prediction model based on the target particles.

[0108] It should be noted that the thermal comfort dataset needs to include indoor temperature information (unit: °C), indoor humidity information (unit: %), indoor wind speed information (unit: m / s), average clothing thermal resistance (unit: m2·K / W) and average metabolism (unit: W / m 2 ), and the thermal comfort values ​​corresponding to these data. The thermal comfort value is an estimated PMV value. Considering that the parameters needed to calculate the PMV value according to the PMV formula (specifically including the human body's metabolic rate, the human body's external mechanical work, air temperature, clothing outer surface temperature, average radiation temperature, water vapor partial pressure, human clothing area coefficient and convection heat transfer coefficient) are relatively complex and some are difficult to obtain, the solution of this application selects indoor temperature information, indoor humidity information, indoor wind speed information, average clothing thermal resistance and average metabolism as key information for predicting thermal comfort values. The deep neural network model is trained by a thermal comfort data set containing these key information to obtain a thermal comfort prediction model to estimate the PMV value.

[0109] Specifically, a deep neural network model can be initialized first, and the deep neural network model can be trained based on the thermal comfort data set by the particle swarm optimization algorithm to obtain the thermal comfort prediction model. The number of particles in each particle swarm is the same, and each particle represents a set of hyperparameters of the deep neural network model: the number of hidden layers, the number of neurons in the layer, the learning rate and the activation function. The thermal comfort data set is divided into a training set, a validation set and a test set. After the deep neural network model is trained by the training set, the fitness value of the particle is calculated according to the performance indicators of the model on the validation set, such as the mean squared error (MSE), to obtain the fitness evaluation result. According to the fitness evaluation result, the position and speed of each particle in the particle swarm are iteratively updated to search for a better hyperparameter combination, and an updated particle swarm is obtained until there are particles in the updated particle swarm whose fitness evaluation results reach the fitness threshold (i.e., the preset threshold). The particles whose fitness evaluation results reach the preset threshold are taken as target particles, and the target particles represent the optimal hyperparameter combination. The thermal comfort prediction model determined based on the target particles is verified by the test set, and the thermal comfort prediction model is output after ensuring the generalization ability of the model.

[0110] Step A12, determining indoor temperature information, indoor humidity information and indoor wind speed information based on the indoor environmental status information;

[0111] It should be understood that the edge terminal will determine the indoor temperature information, indoor humidity information and indoor wind speed information from the indoor environment status information collected by the edge node, and obtain the indoor temperature (unit: ° C), indoor humidity (unit: %) and indoor wind speed (unit: m / s) of the current period. Among them, the indoor temperature information can be obtained through the temperature sensor connected to the edge node, the indoor humidity information can be obtained through the humidity sensor connected to the edge node, and the indoor wind speed information can be obtained through the wind speed sensor connected to the edge node.

[0112] Step A13, determining the number of people, clothing information and activity status of people based on the indoor people information;

[0113] It should be noted that indoor personnel information can be obtained through panoramic images collected by indoor cameras. The edge node will identify the personnel in the panoramic image through the personnel detection model to obtain the number of personnel information, personnel clothing information and personnel activity status. The personnel detection model can be constructed based on the number detection model, clothing detection model and activity status detection model. The number detection model can be obtained by training the target detection model. Through the target detection model, the bounding box of each individual in the image can be identified, and then the number of personnel information can be obtained by statistics. The clothing detection model can be constructed by combining the instance segmentation model and the classification model, wherein the instance segmentation model is used to separate the human body from the background and identify the clothing area worn, and the classification model is used to perform clothing type detection on the identified clothing area, and then obtain the clothing information of the person. The key points of the human body of indoor personnel (such as head, shoulders, elbows, knees, etc.) can be detected through the activity status detection model, and the activity status of the human body can be obtained by analyzing the key point positions of the human body.

[0114] Step A14, predicting a thermal comfort value based on the indoor temperature information, the indoor humidity information, the indoor wind speed information, the number of people information, the clothing information of the people, and the activity status of the people by using the thermal comfort prediction model;

[0115] It should be understood that the clothing thermal resistance of the corresponding person can be estimated based on the clothing information of the person, and the clothing thermal resistance information can be obtained. The metabolic rate of the corresponding person can be estimated based on the human activity state, and the metabolic information can be obtained. The average clothing thermal resistance in the room can be calculated by combining the clothing thermal resistance information and the number of people information, and the average metabolism can be calculated by combining the metabolic information and the number of people information. The indoor temperature information, indoor humidity information, indoor wind speed information, average clothing thermal resistance and average metabolism are input into the thermal comfort prediction model to obtain the predicted thermal comfort value.

[0116] Step A15: Calculate a comfort reward based on the thermal comfort value using a comfort reward function.

[0117] It should be noted that the thermal comfort value is an estimated PMV value, which quantifies the thermal comfort of the environment through a numerical value from -3 to +3. Specifically, -3 corresponds to very cold, -2 corresponds to cold, -1 corresponds to slightly cold, 0 corresponds to comfortable, +1 corresponds to slightly hot, +2 corresponds to hot, and +3 corresponds to very hot. The comfort reward R can be calculated based on the thermal comfort value. comfort , the formula of the comfort reward function is as follows:

[0118] R comfort =-PMV|

[0119] The comfort reward function has been determined when the initial temperature control model of the air conditioner is obtained. The smaller the absolute value of the PMV value, the greater the comfort reward.

[0120] Step S32, determining an energy efficiency reward based on the air conditioning state information and the indoor environment state information;

[0121] It should be noted that energy efficiency rewards are used to minimize energy consumption while ensuring indoor comfort. The air conditioner status information can be collected through the built-in control system and sensors of the smart air conditioner. The instantaneous power consumption of the air conditioner and the air conditioner target temperature set by the user can be determined based on the air conditioner status information. The indoor temperature can be determined based on the indoor environment status information. The energy efficiency reward R can be calculated based on the instantaneous power consumption of the air conditioner, the air conditioner target temperature and the indoor temperature through the energy efficiency reward function. efficiency , the formula of energy efficiency reward function is as follows:

[0122]

[0123] Where, P is the instantaneous power consumption of the air conditioner (unit: kW); T set is the target temperature set by the user (unit: °C); T room is the indoor temperature (unit: °C); γ is a positive constant used to adjust the degree of influence of temperature difference on energy efficiency reward. The energy efficiency reward function has been determined when the initial temperature control model of the air conditioner is obtained. The smaller the temperature difference, the more energy efficiency reward; the smaller the power consumption, the more energy efficiency reward.

[0124] Step S33, determining a power grid load balancing reward based on the power load information and the current time period information;

[0125] It should be noted that the grid load balancing reward is used to ensure that grid consumption is reduced during high-load periods to avoid grid overload. According to the smart grid, the historical grid load information can be obtained. According to the historical grid load information, combined with the current weather information, the maximum grid load value of the day is predicted to obtain the maximum grid load. According to the smart meter, the user's current load can be obtained. According to the current load, combined with the current time period information and the historical grid load information, the grid load in the current time period is predicted to obtain the current grid load. The grid load balancing reward R can be calculated based on the maximum grid load and the current grid load through the grid load balancing reward function. grid , the formula of the grid load balancing reward function is as follows:

[0126]

[0127] Among them, L max is the maximum power load of the power grid (unit: kW), L currentis the current power load of the power grid (unit: kW). The power grid load balance reward function has been determined when the initial temperature control model of the air conditioner is obtained. The smaller the current power load of the power grid, the greater the reward.

[0128] Step S34, determining an electricity price optimization reward based on the energy storage battery information, the grid electricity price information and the current time period information;

[0129] It should be noted that the electricity price optimization reward is used to ensure that the energy storage battery is charged during the period of low electricity price and discharged during the period of high electricity price, so as to reduce the cost of peak electricity price. The energy storage battery information includes the energy storage battery charging power, the energy storage battery discharging power and the energy storage battery power state. According to the current time period information and the power grid electricity price information (such as peak and valley electricity price information), the electricity price of the current time period can be obtained, and it can be determined whether the current time period belongs to the peak load period or the valley load period. The electricity price optimization reward R can be calculated based on the energy storage battery charging power, the energy storage battery discharging power, the energy storage battery power state, the electricity price of the current time period and the current time period information. electricity_price , the formula of the electricity price optimization reward function is as follows:

[0130]

[0131] 0≤E battery ≤E max

[0132] Among them, P charge is the charging power of the energy storage battery (unit: kW); P discharge is the discharge power of the energy storage battery (unit: kW); C is the electricity price in the current period (unit: yuan / kWh); E battery is the energy storage battery power state (unit: kWh); E battery is the maximum capacity of the energy storage battery (unit: kWh), 0≤E battery ≤E max It is used to constrain the power of the energy storage battery to prevent the battery from being overcharged or over-discharged. The electricity price optimization reward function has been determined when the initial temperature control model of the air conditioner is obtained. The greater the discharge of the energy storage battery during the peak period, the greater the electricity price optimization reward, and the greater the charge of the energy storage battery during the valley period, the greater the electricity price optimization reward.

[0133] Step S35, determining the temperature adjustment value and energy storage control information based on the comfort reward, the energy efficiency reward, the grid load balancing reward and the electricity price optimization reward through the target air conditioning temperature control strategy.

[0134] It should be understood that the total reward return can be calculated based on the comfort reward, energy efficiency reward, grid load balancing reward and electricity price optimization reward through the instant reward function. The instant reward function R tIt can be obtained by weighting the reward functions of each part. The calculation formula is as follows:

[0135] R t =β 1 ·R efficiency +β 2 ·R grid +β 3 ·R electricity_price +β 4 ·R comfort

[0136] Among them, β 1、 β 2、 β 3、 β 4、 The instant reward function can guide the edge terminal to adopt the optimal air conditioning temperature control solution that can find a balance between energy saving, balancing grid load, reducing electricity costs and ensuring comfort.

[0137] It should be noted that after obtaining the indoor environment status information, indoor personnel information, air conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information, the current state will be determined based on the indoor environment status information, indoor personnel information, air conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information, and the next state after each action (such as temperature adjustment and energy storage charging and discharging control) is performed in the current state is predicted, and the next state with the highest reward return after executing the action is determined through the target air conditioning temperature control strategy, and the action is executed according to the need to reach the next state to obtain the temperature adjustment value and energy storage control information.

[0138] Step S40: performing intelligent air conditioning temperature control according to the temperature adjustment value and the energy storage control information.

[0139] It should be understood that after obtaining the temperature adjustment value and the energy storage control information, the edge terminal will send the temperature adjustment value and the energy storage control information to the corresponding edge node, and the edge node will generate corresponding temperature adjustment instructions and energy storage control instructions according to the temperature adjustment value and the energy storage control information, so that the energy storage battery can draw electricity from the power grid for charging or discharging according to the energy storage control instruction to drive the operation of the smart air conditioner, thereby enabling the smart air conditioner to adjust the indoor temperature to the target temperature according to the temperature adjustment instruction during operation, and the target temperature is determined according to the indoor temperature information collected by the temperature sensor and the temperature adjustment value carried by the temperature adjustment instruction.

[0140] This embodiment proposes an intelligent air-conditioning temperature control method applied to an edge terminal, by sending an initial air-conditioning temperature control model to a control cloud, so that the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receives the target air-conditioning temperature control model sent by the control cloud, obtains indoor environmental status information, indoor personnel information, air-conditioning status information, energy storage battery information, power load information, power grid electricity price information and current time period information, determines a target air-conditioning temperature control strategy through the target air-conditioning temperature control model, and uses the target air-conditioning temperature control strategy to control the temperature of the indoor environment based on the indoor personnel information. , the air-conditioning status information, the energy storage battery information, the power load information, the power grid electricity price information and the current time period information determine the temperature adjustment value and the energy storage control information, and intelligent air-conditioning temperature control is performed according to the temperature adjustment value and the energy storage control information. The air-conditioning control model is optimized through federated learning, and the air-conditioning temperature and power consumption strategy are dynamically optimized through intelligent control combined with multi-dimensional information (such as indoor environment, personnel activities, electricity prices, etc.). Under the premise of ensuring user comfort, the energy efficiency of the air conditioner is optimized, the burden on the power grid is reduced, and the electricity price fluctuation and the energy storage system are used to further reduce the electricity cost, thereby realizing the comprehensive optimization of the power consumption efficiency in the process of controlling the air-conditioning temperature.

[0141] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 3 Before step S10, the intelligent air conditioning temperature control method further includes steps S01 to S04:

[0142] Step S01, initializing the agent, evaluation network, target network and experience replay pool;

[0143] It should be noted that the agent is responsible for performing operations, i.e., actions, that interact with the environment (i.e., intelligent air conditioning temperature control). When the agent is initialized, its state space and action space are defined. The state space is a collection of all possible states that the environment may be in at any given time. A state (such as indoor environment state information, indoor personnel information, air conditioning state information, energy storage battery information, power load information, grid electricity price information, and current time period information, etc.) is the description of the environment of the agent at a certain moment. The action space defines all possible actions that the agent can perform in each state (such as adjusting the air conditioning temperature, energy storage charging and discharging control). The evaluation network is used to evaluate the current action of the agent and predict the long-term return (i.e., Q value) of the current action based on the reward function. The target network is a copy of the evaluation network, and its weights are updated after a certain time interval to provide a more stable learning signal to avoid drastic fluctuations during the training process of the evaluation network. The experience replay pool is used to store the experience data of the interaction between the agent and the environment (i.e., the combination of state, action, reward, and next state), which is used for the training of the evaluation network.

[0144] Step S02, obtaining local air-conditioning temperature control data, and updating the experience replay pool of the agent based on the local air-conditioning temperature control data;

[0145] It should be noted that the local air-conditioning temperature control data is the real-time collected indoor environment status information, indoor personnel information, air-conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information. After initializing the agent, evaluation network and experience replay pool, the current state of the agent is determined based on the local air-conditioning temperature control data according to the state space of the defined agent, an action is selected from the action space to interact with the environment, the environment state is updated according to the selected action, and after calculating the execution action, the current reward and next state are obtained and the experience replay pool of the agent is updated.

[0146] In a feasible implementation, step S02 may include: obtaining local air-conditioning temperature control data, and defining the current state of the intelligent agent based on the local air-conditioning temperature control data; using temperature adjustment and energy storage charging and discharging control as the action space of the intelligent agent; constructing a comfort reward function, an energy efficiency reward function, a power grid load balancing reward function, and an electricity price optimization reward function to obtain an immediate reward function; based on the immediate reward function, determining the current reward obtained by the intelligent agent when performing the target action in the action space in the current state, and obtaining the next state; using the current state, the target action, the current reward, and the next state as the current experience, and updating the experience replay pool of the intelligent agent based on the current experience.

[0147] Specifically, after obtaining the local air-conditioning temperature control data, the indoor environment status information, indoor personnel information, air-conditioning status information, energy storage battery information, power load information, power grid electricity price information and current time period information will be collected in real time. After processing the real-time collected information, the indoor temperature, indoor humidity, indoor wind speed, average clothing thermal resistance, average metabolism, air-conditioning instantaneous power consumption, air-conditioning target temperature, maximum power load of the power grid, current power load of the power grid, energy storage battery charging power, energy storage battery discharging power, energy storage battery power status, current time period electricity price and current time period information are obtained, which are defined as the current state of the intelligent agent. Temperature adjustment and energy storage charging and discharging control are used as the action space of the intelligent agent. The instant reward function is constructed by combining the comfort reward function, energy efficiency reward function, power grid load balancing reward function and electricity price optimization reward function. Based on the instant reward function, the current reward R obtained by the intelligent agent when performing a certain action (i.e., target action) in the action space in the current state can be calculated. t , and get the next state, change the current state S t Target action A t 、Current reward R t and the next state S t+1 As current experience (S t ,A t ,R t ,S t+1 ), store the current experience into the experience replay pool to update the agent’s experience replay pool.

[0148] Step S03, obtaining experience samples from the experience replay pool, and training and updating the evaluation network based on the experience samples;

[0149] It should be understood that after the experience replay pool is updated, small batches are randomly drawn from the experience replay pool periodically to train the evaluation network to obtain an updated evaluation network.

[0150] In a feasible implementation, step S03 may include: calculating the predicted return of the experience sample based on the evaluation network; calculating the target return of the experience sample based on the target network; calculating the loss value of the evaluation network according to the predicted return and the target return; and updating the evaluation network based on the loss value by gradient descent method.

[0151] Specifically, in each iterative update process, a small batch of samples (S t ,A t ,R t ,S t+1 ), S t Input evaluation network, output in state S tThe Q value of each action is the predicted reward of the evaluation network. During the training process, the evaluation network weights are updated based on the loss function value of the predicted reward calculated by the evaluation network and the target reward calculated by the target reward using the gradient descent method. For edge terminal k, its loss function value L k The calculation formula of (θ) is as follows:

[0152]

[0153] Among them, θ is the parameter of the evaluation network or the target network; S t is the state at time step t; A t is the target action selected at time step t; R t is the immediate reward obtained at time step t; S t+1 is the next state at time step t; η is the discount factor, which is used to indicate the importance of future rewards and its value range is between [0,1]. Used to represent the target network in state S t The target reward of the optimal action selected under t ,A t ; θ) is used to represent the state S according to the evaluation network t Take target action A t The predicted return value at time .

[0154] Step S04: optimizing the target network based on the evaluation network to obtain an initial air conditioning temperature control model.

[0155] Please note that, please refer to Figure 4 , Figure 4 This is a reinforcement learning scenario diagram for the second embodiment of the intelligent air conditioning temperature control method for edge terminals of this application. After the evaluation network is regularly iterated and updated, the parameters of the evaluation network are copied to the target network to improve stability. The parameter θ of the target network is a copy of the parameter θ of the evaluation network, and synchronization is performed every fixed number of steps (such as every 1000 steps), that is, the parameters of the evaluation network are copied to the target network to obtain the initial air conditioning temperature control model.

[0156] Before obtaining the initial air-conditioning temperature control model, this implementation initializes the intelligent agent, the evaluation network, the target network and the experience replay pool, obtains local air-conditioning temperature control data, and updates the experience replay pool of the intelligent agent based on the local air-conditioning temperature control data, obtains experience samples from the experience replay pool, and trains and updates the evaluation network based on the experience samples, optimizes the target network based on the evaluation network, and obtains the initial air-conditioning temperature control model. Through the reinforcement learning algorithm, based on the real-time collected local air-conditioning temperature control data, the intelligent agent continuously learns from the real-time collected data, and combines the real-time feedback of the environment and the user to continuously update and optimize its control strategy to obtain the initial air-conditioning temperature control model, so that the obtained initial air-conditioning temperature control model can make the best intelligent air-conditioning temperature control decision when facing different environments and different needs.

[0157] In addition, the embodiment of the present application provides an intelligent air conditioning temperature control method, and the driving behavior recognition method is applied to control the cloud, please refer to Figure 5 , Figure 5 This is a flow chart of the first embodiment of the intelligent air-conditioning temperature control method applied to control the cloud in the present application.

[0158] In this embodiment, the intelligent air conditioning temperature control method applied to control the cloud includes steps S50 to S80:

[0159] Step S50, receiving the initial air conditioning temperature control model sent by the edge terminal;

[0160] It should be noted that before receiving the initial air-conditioning temperature control model sent by the edge terminal, the control cloud will first initialize a global model, and send the global model containing the model parameters (such as weights and biases, etc.) and structure (such as the layer type of the neural network, layer connection method, number of layer nodes, and activation function, etc.) to each edge terminal, so that each edge terminal can train and learn the global model based on the local data collected in real time through the reinforcement learning algorithm near the data source, obtain the initial air-conditioning temperature control model and feed it back to the edge terminal.

[0161] Step S60, determining the contribution of the edge terminal based on the initial air conditioning temperature control model;

[0162] It should be understood that after receiving the initial air-conditioning temperature control model sent by the edge terminal, the control cloud will feedback the target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning. Among them, federated learning is a distributed machine learning framework that does not require the local data collected by the edge node to be transmitted to the central server through the edge terminal. It only needs to send the initial air-conditioning temperature control model trained at the edge terminal to the control cloud. The control cloud will aggregate the model updates (such as gradient or parameter updates) of multiple edge terminals to obtain the globally optimized target air-conditioning temperature control model. This solution evaluates the contribution of each edge terminal to determine the weight of the model update of each edge terminal during the aggregation process.

[0163] In a feasible implementation, step S60 may include: removing the edge terminal from the edge terminal set to obtain an edge terminal subset; aggregating the initial air-conditioning temperature control model in the edge terminal set to obtain a first aggregate model; aggregating the initial air-conditioning temperature control model in the edge terminal subset to obtain a second aggregate model; and determining the contribution of the edge terminal according to the performance evaluation difference between the first aggregate model and the second aggregate model.

[0164] Specifically, the calculation formula of the contribution of the edge terminal is as follows:

[0165]

[0166] Wherein, N is the set of all edge terminals, i.e., the edge terminal set, including n edge terminals; M is the subset excluding edge terminal k, i.e., the other edge terminals in the edge terminal set; v(M) represents the performance evaluation of subset S during global aggregation (such as model-based accuracy, error, etc.); v(M∪{k} represents the performance evaluation of the aggregation result after adding edge terminal k to subset M; |M| represents the size of subset M, and |N| represents the total number of edge terminals.

[0167] It should be noted that for the edge terminal set M∪{k}, the edge terminal models in the subset M∪{k} can be aggregated together through the federated averaging algorithm to obtain the first aggregate model. M The performance indicators of the model (such as loss value) are calculated to obtain v(M∪{k}), which represents the performance of the average aggregation of the model under the edge terminal set. Similarly, for the edge terminal subset M, the edge terminal models in the subset M can be aggregated together through the federated average algorithm to obtain the second aggregate model θ M, the performance indicators (such as loss value) of the second aggregation model are calculated to obtain v(M), which represents the performance of the model under the cooperation of the edge terminals in subset M. The additional contribution of edge terminal k after joining subset M can be calculated through v(M∪{k})-v(M), that is, v(M∪{k})-v(M).

[0168] Step S70, performing weighted aggregation on the edge terminals based on the contribution to obtain a target air conditioning temperature control model;

[0169] It should be understood that after traversing all possible edge terminal subsets M, the contribution φ of all edge terminals will be obtained. k According to the contribution, the updates of all initial air conditioning temperature control models in the edge terminal set N can be weighted and aggregated to obtain global update information. The initialized global model is iteratively trained according to the global update information to obtain the target air conditioning temperature control model. The update submitted by each edge terminal k is the gradient The formula for weighted aggregation is as follows:

[0170]

[0171] in, is the weighted global model update, which means that all edge terminals are updated according to their contribution φ k Weighted results.

[0172] Step S80: sending the target air-conditioning temperature control model to the edge terminal.

[0173] It should be understood that the target air-conditioning temperature control model is obtained by weighted aggregation of the model updates of the edge terminals, and the target air-conditioning temperature control model is sent to the edge terminals as feedback to the initial air-conditioning temperature control model.

[0174] The present embodiment provides an intelligent air-conditioning temperature control method applied to a control cloud, receiving an initial air-conditioning temperature control model sent by an edge terminal, determining the contribution of the edge terminal based on the initial air-conditioning temperature control model, performing weighted aggregation on the edge terminal based on the contribution to obtain a target air-conditioning temperature control model, and sending the target air-conditioning temperature control model to the edge terminal. In a federated learning process, by evaluating the contribution of the edge terminal, the contribution of each terminal is quantified, and the model update provided by the edge terminal is weightedly aggregated according to the quantified contribution, thereby obtaining a globally optimized target air-conditioning temperature control model, so that high-quality edge terminals can play a greater role in improving the global model, while low-quality or non-compliant terminals will have less impact on the global model, thereby improving the accuracy and stability of the overall model.

[0175] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the intelligent air-conditioning temperature control method of the present application. More simple transformations based on this technical concept are all within the protection scope of the present application.

[0176] The present application also proposes an intelligent air-conditioning temperature control system, which includes the control cloud mentioned above and multiple edge terminals mentioned above.

[0177] Since the intelligent air-conditioning temperature control system adopts all the technical solutions of all the above embodiments, it has at least all the beneficial effects brought by the technical solutions of the above embodiments, which will not be described one by one here.

[0178] This application also proposes an intelligent air conditioning temperature control device, please refer to Figure 6 , applied to an edge terminal, the intelligent air-conditioning temperature control device comprises:

[0179] The model feedback module 10 is used to send an initial air-conditioning temperature control model to a control cloud, so that the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receives the target air-conditioning temperature control model sent by the control cloud;

[0180] The information acquisition module 20 is used to obtain indoor environment status information, indoor personnel information, air conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information;

[0181] An information calculation module 30 is used to determine a target air-conditioning temperature control strategy through the target air-conditioning temperature control model, and determine a temperature adjustment value and energy storage control information based on the indoor environment state information, the indoor personnel information, the air-conditioning state information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information through the target air-conditioning temperature control strategy;

[0182] The temperature control module 40 is used to perform intelligent air conditioning temperature control according to the temperature adjustment value and the energy storage control information.

[0183] This application also proposes an intelligent air conditioning temperature control device, please refer to Figure 7 , applied to control the cloud, the intelligent air conditioning temperature control device comprises:

[0184] The model receiving module 50 is used to receive the initial air conditioning temperature control model sent by the edge terminal;

[0185] A weight allocation module 60, configured to determine the contribution of the edge terminal based on the initial air conditioning temperature control model;

[0186] A model aggregation module 70, configured to perform weighted aggregation on the edge terminals based on the contribution to obtain a target air conditioning temperature control model;

[0187] The model distribution module 80 is used to send the target air-conditioning temperature control model to the edge terminal.

[0188] The intelligent air-conditioning temperature control device provided by the present application adopts the intelligent air-conditioning temperature control method in the above embodiment, which can solve the technical problem of how to comprehensively optimize the power efficiency in the process of controlling the air-conditioning temperature, so as to balance the load of the power grid while maintaining the comfort of the user. Compared with the prior art, the beneficial effects of the intelligent air-conditioning temperature control device provided by the present application are the same as the beneficial effects of the intelligent air-conditioning temperature control method provided by the above embodiment, and the other technical features in the intelligent air-conditioning temperature control device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0189] The present application provides an intelligent air-conditioning temperature control device, which includes: at least one processor; and a memory that is communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the intelligent air-conditioning temperature control method in the above-mentioned embodiment one.

[0190] Reference below Figure 8 , which shows a schematic diagram of the structure of an intelligent air-conditioning temperature control device suitable for implementing the embodiment of the present application. The intelligent air-conditioning temperature control device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The intelligent air-conditioning temperature control device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0191] like Figure 8As shown, the intelligent air conditioning temperature control device may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the intelligent air conditioning temperature control device are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the intelligent air conditioning temperature control device to communicate with other devices wirelessly or wired to exchange data. Although the figure shows an intelligent air conditioning temperature control device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided instead.

[0192] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0193] The intelligent air-conditioning temperature control device provided by the present application adopts the intelligent air-conditioning temperature control method in the above embodiment, which can solve the technical problem of how to comprehensively optimize the power efficiency in the process of controlling the air-conditioning temperature, so as to balance the load of the power grid while maintaining the comfort of the user. Compared with the prior art, the beneficial effects of the intelligent air-conditioning temperature control device provided by the present application are the same as the beneficial effects of the intelligent air-conditioning temperature control method provided by the above embodiment, and the other technical features in the intelligent air-conditioning temperature control device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0194] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0195] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0196] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the intelligent air-conditioning temperature control method in the above-mentioned embodiment.

[0197] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0198] The computer-readable storage medium may be included in the intelligent air-conditioning temperature control device; or may exist independently without being assembled into the intelligent air-conditioning temperature control device.

[0199] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the intelligent air-conditioning temperature control device, the intelligent air-conditioning temperature control device: sends an initial air-conditioning temperature control model to the control cloud, so that the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receives the target air-conditioning temperature control model sent by the control cloud; obtains indoor environmental status information, indoor personnel information, air-conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information; determines a target air-conditioning temperature control strategy through the target air-conditioning temperature control model, and determines a temperature adjustment value and energy storage control information through the target air-conditioning temperature control strategy based on the indoor environmental status information, the indoor personnel information, the air-conditioning status information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information; and performs intelligent air-conditioning temperature control according to the temperature adjustment value and the energy storage control information.

[0200] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0201] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0202] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0203] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned intelligent air-conditioning temperature control method, and can solve the technical problem of how to comprehensively optimize the power efficiency in the process of controlling the air-conditioning temperature, so as to balance the load of the power grid while maintaining the comfort of the user. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the intelligent air-conditioning temperature control method provided by the above-mentioned embodiment, and will not be repeated here.

[0204] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned intelligent air-conditioning temperature control method when executed by a processor.

[0205] The computer program product provided by this application can solve the technical problem of how to comprehensively optimize the power efficiency in the process of controlling the air conditioner temperature so as to balance the load of the power grid while maintaining the comfort of the user. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as those of the intelligent air conditioner temperature control method provided by the above embodiment, and will not be described in detail here.

[0206] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. An intelligent air conditioning temperature control method, characterized in that: The intelligent air-conditioning temperature control method is applied to an edge terminal, and the method comprises: Sending an initial air-conditioning temperature control model to a control cloud, so that the control cloud feeds back a target air-conditioning temperature control model based on the initial air-conditioning temperature control model through federated learning, and receiving the target air-conditioning temperature control model sent by the control cloud; Obtain indoor environment status information, indoor personnel information, air conditioning status information, energy storage battery information, power load information, grid electricity price information and current time period information; Determine a target air conditioning temperature control strategy through the target air conditioning temperature control model, and determine a temperature adjustment value and energy storage control information through the target air conditioning temperature control strategy based on the indoor environment status information, the indoor personnel information, the air conditioning status information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information; Intelligent air conditioning temperature control is performed according to the temperature adjustment value and the energy storage control information.

2. The method according to claim 1, characterized in that The step of determining the temperature adjustment value and the energy storage control information based on the indoor environment status information, the indoor personnel information, the air conditioning status information, the energy storage battery information, the power load information, the grid electricity price information and the current time period information through the target air conditioning temperature control strategy includes: Determine a comfort reward based on the indoor environment state information and the indoor personnel information; Determining an energy efficiency reward based on the air conditioning status information and the indoor environment status information; Determining a grid load balancing reward based on the power load information and current time period information; Determine an electricity price optimization reward based on the energy storage battery information, the grid electricity price information and the current time period information; The temperature adjustment value and energy storage control information are determined through the target air-conditioning temperature control strategy based on the comfort reward, the energy efficiency reward, the grid load balancing reward and the electricity price optimization reward.

3. The method according to claim 2, characterized in that The step of determining a comfort reward based on the indoor environment state information and the indoor personnel information comprises: Obtain thermal comfort prediction model; Determine indoor temperature information, indoor humidity information and indoor wind speed information based on the indoor environment status information; Determine the number of people, clothing information and activity status of people based on the indoor people information; Predicting a thermal comfort value based on the indoor temperature information, the indoor humidity information, the indoor wind speed information, the number of people information, the clothing information of the people, and the activity status of the people by using the thermal comfort prediction model; A comfort reward is calculated based on the thermal comfort value by a comfort reward function.

4. The method according to claim 3, characterized in that Before the step of obtaining the thermal comfort prediction model, the method further includes: Obtain thermal comfort dataset and initialize deep neural network model; Initialize a particle swarm, and use the number of hidden layers, the number of neurons in each layer, the learning rate, and the activation function of the deep neural network model as particles of the particle swarm; Performing fitness evaluation on particles in the particle swarm based on the thermal comfort data set to obtain a fitness evaluation result; updating the particle swarm based on the fitness evaluation result until there is a particle in the particle swarm whose fitness evaluation result reaches a preset threshold, and taking the particle whose fitness evaluation result reaches the preset threshold as a target particle; A thermal comfort prediction model is determined based on the target particles.

5. The method according to claim 1, characterized in that Before the step of sending the initial air conditioning temperature control model to the control cloud so that the control cloud targets the air conditioning temperature control model based on the initial air conditioning temperature control model and receives the target air conditioning temperature control model sent by the control cloud, the step further includes: Initialize the agent, evaluation network, target network, and experience replay pool; Acquire local air-conditioning temperature control data, and update the experience replay pool of the agent based on the local air-conditioning temperature control data; Acquire experience samples from the experience replay pool, and train and update the evaluation network based on the experience samples; The target network is optimized based on the evaluation network to obtain an initial air conditioning temperature control model.

6. The method according to claim 5, characterized in that The step of obtaining local air-conditioning temperature control data and updating the experience replay pool of the agent based on the local air-conditioning temperature control data includes: Acquire local air-conditioning temperature control data, and define the current state of the agent based on the local air-conditioning temperature control data; Taking temperature adjustment and energy storage charging and discharging control as the action space of the intelligent agent; Construct the comfort reward function, energy efficiency reward function, grid load balance reward function and electricity price optimization reward function to obtain the instant reward function; Determine, based on the instant reward function, a current reward obtained by the agent when executing the target action in the action space in the current state, and obtain a next state; The current state, the target action, the current reward and the next state are used as current experience, and the experience replay pool of the agent is updated based on the current experience.

7. The method according to claim 5, characterized in that The step of obtaining experience samples from the experience replay pool and training and updating the evaluation network based on the experience samples includes: Calculating the predicted return of the experience sample based on the evaluation network; Calculating a target return for the experience sample based on the target network; Calculating a loss value of the evaluation network according to the predicted return and the target return; The evaluation network is updated based on the loss value by a gradient descent method.

8. An intelligent air conditioning temperature control method, characterized in that: The intelligent air-conditioning temperature control method is applied to control the cloud, and the method comprises: Receive the initial air conditioning temperature control model sent by the edge terminal; Determining the contribution of the edge terminal based on the initial air conditioning temperature control model; Performing weighted aggregation on the edge terminals based on the contribution to obtain a target air conditioning temperature control model; The target air-conditioning temperature control model is sent to the edge terminal.

9. The method according to claim 8, characterized in that The step of determining the contribution of the edge terminal based on the initial air conditioning temperature control model includes: Eliminating the edge terminal from the edge terminal set to obtain an edge terminal subset; Aggregating the initial air-conditioning temperature control models in the edge terminal set to obtain a first aggregated model; Aggregating the initial air-conditioning temperature control models in the edge terminal subset to obtain a second aggregated model; The contribution of the edge terminal is determined according to a performance evaluation difference between the first aggregation model and the second aggregation model.

10. An intelligent air conditioning temperature control system, characterized in that: The intelligent air-conditioning temperature control system comprises: an edge terminal and multiple control clouds, wherein the steps of the intelligent air-conditioning temperature control method as described in any one of claims 1 to 7 are executed on the edge terminal, and the steps of the intelligent air-conditioning temperature control method as described in any one of claims 8 to 9 are executed on the multiple control clouds.

Images