Air conditioner control method, device and system and air conditioner

Abstract

The invention discloses an air conditioner control method, device and system and an air conditioner, and belongs to the field of air conditioners. When the air conditioner enters the pet mode, pet information is obtained, the pet information at least comprises pet behavior data and pet physiological data, then the pet information is input into a pre-trained temperature sensitivity model, model output is obtained, then the target temperature is obtained based on the model output, and the air conditioner is controlled to reach the target temperature. According to the scheme, the temperature of the air conditioner is controlled by combining the behavior data and the physiological data of the pet, the air conditioner can provide more accurate temperature according to behaviors and physiology of the pet, manual adjustment is not needed, and the user experience is improved.

Description

Technical Field

[0001] The present application relates to the technical field of air conditioning, and in particular, to an air conditioning control method, device and system, and an air conditioner. Background Art

[0002] Current smart air conditioning systems are already able to sense whether there are people in the room based on sensor technology and make automatic adjustments accordingly. For example, when no one is in the room, the air conditioner automatically switches to energy-saving mode. Some high-end air conditioners also have the function of sensing the human body through infrared sensors to achieve more precise temperature control. However, existing technologies mainly focus on human comfort adjustment, while ignoring the presence of indoor pets.

[0003] Currently, there is no smart air conditioning technology specifically designed for pet sensing and adjustment on the market. Although some air conditioners may allow users to manually adjust the temperature to suit the needs of pets, these methods rely on the user's active operation and lack automation and intelligent processing. Existing technical solutions mainly control the operation of air conditioners by manually setting the temperature or time period, but cannot automatically identify the pet's sensitivity to temperature and make corresponding adjustments. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the present application provides an air conditioning control method, device, system and air conditioner to solve the problem that some technical solutions mainly control the operation of the air conditioner by manually setting the temperature or time period, but cannot automatically identify the pet's sensitivity to temperature and then make corresponding adjustments.

[0005] The technical solution adopted by this application to solve its technical problem is:

[0006] In a first aspect, an air conditioning control method is provided, comprising:

[0007] When entering the pet mode, obtaining pet information, the pet information including pet behavior data and pet physiological data;

[0008] Inputting the pet behavior data and the pet physiological data into a pre-trained temperature sensitivity model to obtain a model output;

[0009] A target temperature is obtained based on the model output, and the air conditioner is controlled to reach the target temperature.

[0010] Furthermore, the training step of the temperature sensitivity model includes:

[0011] Constructing an initial model and a training set, wherein the initial model includes an input layer, a hidden layer and an output layer, wherein the input layer includes a first input unit and a second input unit, wherein the hidden layer includes a first hidden layer and a second hidden layer, wherein the first input unit is connected to the first hidden layer, wherein the second input unit is connected to the second hidden layer, wherein the training sample includes a first training sample and a second training sample, wherein the first training sample is pet behavior data, and the second training sample is pet physiological data;

[0012] Inputting the first training sample into the first input unit to obtain a first output of the first hidden layer, and inputting the second training sample into the second input unit to obtain a second output of the second hidden layer;

[0013] The first output and the second output are weighted based on the first initial weight and the second initial weight to obtain a model output;

[0014] If the loss function is less than or equal to a threshold, updating the parameters of the first hidden layer and the second hidden layer and updating the first initial weight and the second initial weight until the loss function is greater than the threshold;

[0015] When the loss function is greater than the threshold, the current model is used as a trained temperature sensitivity model.

[0016] Further, the first hidden layer and the second hidden layer respectively include a plurality of neurons, and the step of inputting the first training sample into the first input unit to obtain a first output of the first hidden layer, and inputting the second training sample into the second input unit to obtain a second output of the second hidden layer includes:

[0017] Inputting the first training sample into the first input unit, setting the outputs of a random number of neurons in the first hidden layer to 0, and obtaining a first output of the first hidden layer;

[0018] And / or, inputting the second training sample into the second input unit, setting the outputs of a random number of neurons in the second hidden layer to 0, and obtaining a second output of the second hidden layer.

[0019] Furthermore, the pet information also includes the pet type, and also includes:

[0020] A target temperature is obtained according to the type of pet.

[0021] Further, obtaining the target temperature according to the pet type includes:

[0022] Obtaining a type coefficient corresponding to the pet type based on a preset corresponding relationship;

[0023] The product of the model output and the type coefficient is used as the target temperature.

[0024] Further, obtaining the target temperature according to the pet type includes:

[0025] Inputting the pet behavior data, the pet physiological data, and the pet type into a pre-trained temperature sensitivity model to obtain a model output;

[0026] The model output is used as the target temperature.

[0027] Furthermore, it also includes:

[0028] Check whether there is someone in the room;

[0029] If there is no one in the room, detect whether there is a pet;

[0030] If a pet is detected, the system will enter pet mode.

[0031] Furthermore, the step of inputting the pet behavior data and the pet physiological data into a pre-trained temperature sensitivity model includes:

[0032] Perform time series analysis on the pet behavior data and the pet physiological data to eliminate noise and outliers.

[0033] In a second aspect, an air conditioning control device is provided, comprising:

[0034] A pet information acquisition module, used to acquire pet information when entering pet mode, wherein the pet information includes pet behavior data and pet physiological data;

[0035] A model output acquisition module, used for inputting the pet behavior data and the pet physiological data into a pre-trained temperature sensitivity model to obtain a model output;

[0036] The air conditioner temperature control module is used to obtain a target temperature based on the model output and control the air conditioner to reach the target temperature.

[0037] In a third aspect, an air conditioning control system is provided, comprising:

[0038] at least one processor and at least one memory;

[0039] The memory stores executable instructions of the processor;

[0040] The processor is configured to execute the above-mentioned air conditioning control method.

[0041] In a fourth aspect, an air conditioner is provided, applying the above-mentioned air conditioner control method.

[0042] Beneficial effects:

[0043] The technical solution of the present application provides an air conditioning control method, device, system and air conditioning. When entering the pet mode, pet information is obtained, and the pet information includes at least pet behavior data and pet physiological data. The pet information is then input into a pre-trained temperature sensitivity model to obtain a model output, and then a target temperature is obtained based on the model output, and the air conditioning is controlled to reach the target temperature. The present application solution controls the temperature of the air conditioning in combination with the pet's behavior data and physiological data, so that the air conditioning can provide a more accurate temperature according to the pet's behavior and physiology, without manual adjustment, thereby improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] 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, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 is a flow chart of an air conditioning control method provided in an embodiment of the present application;

[0046] Figure 2 It is a schematic diagram of an initial model structure provided in an embodiment of the present application;

[0047] Figure 3 is a flow chart of a specific air conditioning control method provided in an embodiment of the present application;

[0048] Figure 4 is a schematic diagram of the structure of an air conditioning control device provided in an embodiment of the present application;

[0049] Figure 5 It is a schematic diagram of the structure of an air-conditioning control system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme of the present application is described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other implementation methods obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0051] Existing smart air-conditioning systems are unable to automatically sense the presence of pets indoors and adjust the environment according to their characteristics, resulting in the air-conditioning being turned off or switched to energy-saving mode when no one is around, ignoring the comfort and health needs of pets.

[0052] To solve this problem, refer to Figure 1 , the embodiment of the present application provides an air conditioning control method, comprising:

[0053] S11: When entering the pet mode, obtaining pet information, the pet information including pet behavior data and pet physiological data.

[0054] As a preferred implementation of the embodiment of the present application, whether to enter the pet mode is determined in the following manner:

[0055] Detect whether there is anyone in the room; if there is no one in the room, detect whether there is a pet; if a pet is detected, enter pet mode.

[0056] Alternatively, when receiving an instruction from the user to enter the pet mode, the pet mode is entered.

[0057] In this application, pet information is obtained based on various sensors that are set up (such as acceleration sensors, RGB thermal infrared cameras, heart rate and breathing monitors, etc.), but its dimensions include at least two dimensions of data, namely pet behavior data (such as resting, walking, running, etc.) and pet physiological data (such as heart rate and breathing rate).

[0058] S12: Inputting the pet behavior data and the pet physiological data into a pre-trained temperature sensitivity model to obtain a model output.

[0059] Since the acquired data may contain noise or outliers, in the embodiment of the present application, it is necessary to perform time series analysis on the pet behavior data and the pet physiological data to eliminate noise and outliers. Then, the features of the pet behavior data and the pet physiological data are extracted and then input into the pre-trained temperature sensitivity model. Of course, a vector extraction network can also be set in the temperature sensitivity model, so that the data with noise and outliers eliminated can be directly input into the temperature sensitivity model.

[0060] The training step of the temperature sensitivity model includes:

[0061] Construct the initial model and training set, such as Figure 2As shown, the initial model includes an input layer, a hidden layer and an output layer, the input layer includes a first input unit and a second input unit, the hidden layer includes a first hidden layer and a second hidden layer, the first input unit is connected to the first hidden layer, the second input unit is connected to the second hidden layer, the training sample includes a first training sample and a second training sample, the first training sample is pet behavior data, and the second training sample is pet physiological data; it should be noted that since both the pet behavior data and the pet physiological data include at least two types of data, if the pet behavior data and the pet physiological data are input through one input unit, it may not be possible to fit, or the fitting time is long. Therefore, the initial model used in this application is two input units, which are used to input pet behavior data and pet physiological data respectively, and the hidden layer is also divided into two parts, which are used to obtain an output according to the pet behavior data and the pet physiological data, which is equivalent to two small models. Then the two outputs are weighted. It can be understood that the first input unit or the second input unit can take multiple data as input at the same time. The first hidden layer and the second hidden layer respectively include multiple neurons, and the outputs of the first hidden layer and the second hidden layer are obtained based on the output of each neuron.

[0062] The first training sample is input into the first input unit to obtain a first output of the first hidden layer, and the second training sample is input into the second input unit to obtain a second output of the second hidden layer.

[0063] The model output is obtained by weighting the first output and the second output based on the first initial weight and the second initial weight; wherein, there is a preset range for the first initial weight and the second initial weight, because if the initial weight is set too extreme, it will cause the model to overfit. Model output = first output * first initial weight + second output * second initial weight.

[0064] If the loss function is less than or equal to the threshold, the parameters of the first hidden layer and the second hidden layer are updated, and the first initial weight and the second initial weight are updated until the loss function is greater than the threshold. The loss function can be selected according to actual needs. The embodiment of the present application adopts Mean Squared Error (MSE) as the loss function.

[0065] When the loss function is greater than the threshold, the current model is used as a trained temperature sensitivity model.

[0066] Among them, according to actual needs, the model output can be an optimal temperature, or an optimal temperature range (that is, the upper limit and lower limit of the output range), or the output of the optimal temperature and the optimal stable range.

[0067] In practice, the trained model may have an overfitting problem in which the fitting effect on the training data is good, but the effect on the actual data is poor. The inventors found that this is because some features in the training data may always appear in combination, which leads to the problem of overfitting in the training. Therefore, as a preferred implementation method of the embodiment of the present application, the first training sample is input into the first input unit, and the outputs of a random number of neurons in the first hidden layer are set to 0 to obtain the first output of the first hidden layer;

[0068] And / or, inputting the second training sample into the second input unit, setting the outputs of a random number of neurons in the second hidden layer to 0, and obtaining a second output of the second hidden layer.

[0069] Since the outputs of a random number of neurons are set to 0, that is, some features are actively ignored during each training to avoid the appearance of certain feature combinations, the overfitting effect of the model can be greatly reduced.

[0070] S13: Obtaining a target temperature based on the model output, and controlling the air conditioner to reach the target temperature.

[0071] Optionally, the model output is directly used to obtain the target temperature.

[0072] In practical applications, different pets have different optimal temperatures. In addition, even if they are the same animal, their optimal temperatures will be different if their specific breeds are different. Therefore, as a preferred implementation method of the present application, the pet information also includes the type of pet, and also includes: obtaining the target temperature according to the type of pet. The type of pet can be known based on the information input by the user, or based on image recognition technology. It is not limited here. In addition, the type of pet can only include general types, such as cats and dogs, or specific breeds, such as tabby cats, hairless cats, Shiba Inus, poodles, etc.

[0073] In one embodiment, obtaining the target temperature according to the pet type includes:

[0074] Based on the preset corresponding relationship, the type coefficient corresponding to the pet type is obtained; the product of the model output and the type coefficient is used as the target temperature. That is, a coefficient is set for each type of pet in advance. When the pet type is obtained, the target temperature can be obtained by multiplying the model output by the coefficient. This solution is relatively simple and the training model is relatively low cost.

[0075] In another embodiment, obtaining the target temperature according to the pet type includes:

[0076] The pet behavior data, the pet physiological data and the pet type are input into the pre-trained temperature sensitivity model to obtain a model output, and the model output is used as the target temperature. That is, the pet type needs to be considered when training the model.

[0077] If the desired model is only for one target type of pet, the training steps and model structure are the same as those of the above embodiment, except that when constructing the training set, the data to be collected are the behavioral data and physiological data of the target type.

[0078] If the model you want to obtain needs to meet the needs of multiple types of pets, you can either train one for each type, or take the pet type as an input during training and input it into the model for training.

[0079] Specifically, the pet behavior data and the corresponding pet species are input as one sample, and the physiological data of the same pet and the corresponding pet species are input as another sample. In this way, the overall structure of the model does not need to be changed, that is, there are still two input units, but each input unit has an additional pet species input.

[0080] In another method, the input layer of the model needs to be set to three input units, which input the pet physiological data, pet behavior data and pet type respectively. Similarly, the hidden layer is also divided into three, namely Figure 2 The third input unit and the third hidden layer are added to the structure. After the pet type is input into the third input unit, the third output of the third hidden layer can be obtained, and then weighted according to the third initial weight. That is, model output = model output = first output * first initial weight + second output * second initial weight + third output * third initial weight. The subsequent training process is the process of adjusting the three hidden layers and the three initial weights, which will not be described in detail here.

[0081] The air conditioning control method provided in the embodiment of the present application obtains pet information when entering the pet mode, and the pet information includes at least pet behavior data and pet physiological data, and then inputs the pet information into a pre-trained temperature sensitivity model to obtain a model output, and then obtains a target temperature based on the model output, and controls the air conditioning to reach the target temperature. The present application scheme controls the temperature of the air conditioning in combination with the pet's behavior data and physiological data, so that the air conditioning can provide a more accurate temperature according to the pet's behavior and physiology, without manual adjustment, thereby improving the user experience.

[0082] In order to more specifically explain the present application scheme, a more specific air conditioning control method is provided below, which is applied to the intelligent air conditioning system, wherein the intelligent air conditioning system includes but is not limited to the intelligent air conditioning, the intelligent sensor of the pet, the server, and the APP; the devices communicate and transmit data through wireless communication. The intelligent air conditioning system supports OTA updates and upgrades, continuously optimizes the pet perception algorithm and temperature adjustment logic, and adapts to more diverse pet needs.

[0083] Among them, the smart air-conditioning system can be understood as a smart home system.

[0084] Air conditioners, pet smart sensors, APPs, and servers form an intelligent air conditioning system, that is, a smart home system. The air conditioner is connected to the server, various pet smart sensors are connected to the server, and the APP is connected to the server; the server side is used as a central control scheduling process to process data and instructions between various devices. For example, the air conditioner feeds back the data uploaded by the air conditioner to the server, the server side analyzes and processes it, and the server side sends instructions to the air conditioner and sends information to the APP; the pet smart sensor feeds back the data uploaded by the sensor to the server, the server side analyzes and processes it, and the server side sends instructions to the air conditioner.

[0085] The intelligent air conditioner includes but is not limited to a sensing module unit, a main control unit, and a wireless module unit; the sensing module unit senses whether there are people in the environment; the main control unit is responsible for logical scheduling and control; and the wireless module unit performs wireless communication.

[0086] Pet smart sensors include accelerometers, RGB thermal infrared cameras, heart rate and respiration monitors, etc. Accelerometers capture the activity level and posture changes of pets. By monitoring the acceleration changes of pets, it can be determined whether the pet is resting, walking, running, etc.; RGB thermal infrared cameras capture real-time images and video data of pets, and analyze pet behavior patterns, such as eating, resting, playing, etc.; thermal infrared cameras can obtain clear images at night or in low light conditions, and can identify the pet's body surface temperature; heart rate and respiration monitors monitor the pet's heart rate and respiratory rate in real time, and communicate with the system to provide real-time data.

[0087] APP users can enter the pet's specific information (such as breed, age, health status, etc.) through the APP. The system will perform personalized temperature adjustment based on this information, and the operating parameters of the air conditioner can be adjusted and controlled through the APP.

[0088] The server deploys a pet temperature sensitivity recognition algorithm model. The temperature sensitivity model uses a deep neural network (DNN) model to analyze the pet's comfortable environment temperature range through model recognition.

[0089] Temperature Sensitivity Model Scheme

[0090] 1. Data Collection

[0091] 1.1 Behavioral Data Collection

[0092] Sensor network: Various smart sensors are deployed in the pet's activity space, including accelerometers, RGB thermal infrared cameras, heart rate and respiration monitors, etc. Accelerometers are used to capture the pet's activity level, cameras are used to analyze its behavior patterns and detect the pet's body temperature distribution, and heart rate and respiration monitors provide real-time data on physiological parameters. The pet's behavioral characteristic data at different temperatures are recorded, such as activity frequency, posture changes, eating and drinking behavior, etc.

[0093] 1.2 Physiological data collection

[0094] Respiratory Rate and Heart Rate: Use wearable sensors to monitor your pet’s respiratory rate and heart rate.

[0095] Body surface temperature: The temperature distribution and overall body temperature of pets are monitored through thermal imaging technology. The fluctuation of body temperature can reflect the adaptation of pets in different temperature environments.

[0096] 2. Data analysis and sensitivity calculation

[0097] 2.1 Data Preprocessing

[0098] Time series analysis: Perform time series analysis on the collected behavioral and physiological data to remove noise and outliers.

[0099] Feature extraction: Extract key features from behavioral and physiological data, such as average activity level, heart rate variability, respiratory rate fluctuation, body surface temperature gradient, etc., for subsequent sensitivity analysis.

[0100] 2.2 Temperature sensitivity function

[0101] Temperature-behavior response relationship: Statistically analyze the correspondence between behavioral data and temperature, and construct a temperature-behavior response curve (TBRC). TBRC describes the changes in the activity level of pets at different temperatures. At higher temperatures, the activity level may drop significantly, while at moderate temperatures, the activity frequency reaches a peak. It can be expressed by the following formula (the functional relationship obtained by fitting the output data of the trained model):

[0102]

[0103] Where, T is the ambient temperature, T opt is the optimal temperature, α and β are parameters fitted according to the data. TBRC(T): The behavioral response value at temperature T, similar to the ordinate y, represents the activity level or behavioral intensity of the pet.

[0104] T: temperature, similar to the horizontal axis x, representing different ambient temperatures.

[0105] α: Proportional factor, which controls the maximum behavioral response value of the curve (equivalent to the peak value of the curve) and is set according to the activity characteristics of the specific pet.

[0106] β: Sensitivity parameter, affecting the distribution width of the curve on temperature, that is, the sensitivity of the pet to temperature changes. A higher β means that the pet is more sensitive to temperature changes and the activity level is more dependent on temperature.

[0107] T opt : Optimal temperature, which is the temperature at which your pet's activity level is the highest (or reaches a comfort temperature). This is the temperature at which an individual is most active and most comfortable.

[0108] The correspondence between the horizontal and vertical coordinates in this response curve:

[0109] Horizontal axis (x): temperature T, indicating the temperature value of the current environment.

[0110] Vertical axis (y): TBRC (T), that is, the behavioral response value of the pet at temperature T, reflecting the activity level. The higher the activity level, the more suitable the temperature is for the pet; the decrease in activity level means that the temperature deviates from the optimal range for the pet.

[0111] TBRC constructs the activity response curve of pets under different temperature conditions based on behavioral data. The main technical advantages include:

[0112] Enhanced model sensitivity: By associating temperature changes with the pet's activity level, the system can respond more quickly to changes in ambient temperature. For example, when the temperature rises and the activity level decreases, TBRC can trigger the system to lower the temperature in advance. This sensitive regulation based on activity patterns can avoid discomfort or safety risks caused by high temperatures in a timely manner.

[0113] Improve the accuracy and efficiency of air conditioning: Using TBRC can help the air conditioning system accurately locate the activity habits of pets at different temperatures. The model can accurately identify the temperature range suitable for pets, allowing the system to avoid unnecessary adjustments while maintaining comfort and saving energy.

[0114] Support for abnormal behavior detection: TBRC can serve as a reference for behavior monitoring and identify abnormal behavior. When the activity level is significantly lower than the expected value of TBRC, the system can indicate possible health risks. For example, low activity levels in extreme temperatures may indicate environmental temperature discomfort or health problems, triggering additional alarms or interventions.

[0115] Without TBRC support, the air conditioning system will not be able to accurately predict the behavior changes of pets at different temperatures and can only adjust based on a single temperature setting. This may result in: abnormally reduced activity levels or overexcitement of pets at extreme temperatures, leading to discomfort or health risks. The temperature regulation of the system is too mechanical and cannot achieve dynamic comfort optimization. It may fail to adjust in advance at high temperatures, causing pets to overheat.

[0116] Physiological response relationship: Statistically calculate the corresponding relationship between physiological data (respiration, heart rate, body temperature) and temperature, and construct the Temperature-Physiological Response Curve (TPRC) to describe the impact of temperature on physiological parameters such as respiratory rate, heart rate, and body temperature. Through multiple regression analysis, the sensitivity of temperature to these parameters can be obtained. It can be expressed by the following formula:

[0117] TPRC(T)=γ·log(1+δ(TT opt ))

[0118] Among them, γ and δ are fitting parameters.

[0119] TPRC(T): Physiological response value at temperature T (similar to the vertical axis y), reflecting the physiological parameter response of the pet, the intensity of physiological responses such as heart rate and respiratory rate at different temperatures.

[0120] T: Temperature, as the horizontal axis x, represents the ambient temperature.

[0121] γ: Proportional factor, which controls the maximum physiological response value of the curve and is related to the physiological characteristics of the pet.

[0122] δ: Physiological sensitivity parameter, affecting the sensitivity of the curve to temperature changes.

[0123] T opt : The optimal temperature, that is, the pet's physiological state is closest to the ideal state at this temperature.

[0124] Correspondence between horizontal and vertical coordinates

[0125] In this physiological response curve:

[0126] Abscissa (x): temperature T, representing different ambient temperature values.

[0127] Vertical axis (y): TPRC (T), that is, the physiological response value of the pet at temperature T, reflecting the response degree of physiological parameters such as heart rate and respiratory rate.

[0128] TPRC describes the physiological load of pets at different temperatures based on physiological data (such as respiratory rate, heart rate, etc.). The main technical advantages include:

[0129] Enhanced personalization of temperature control strategies: TPRC can provide specific physiological temperature control needs for pets of different types, sizes and ages. For example, some dogs may experience an increase in heart rate more quickly in high temperatures, while cats are more heat tolerant. TPRC allows the system to dynamically adjust the temperature based on physiological characteristics to meet personalized needs.

[0130] Achieve efficient control with dynamic adjustment: By using physiological load as an important indicator of temperature control, the system can monitor changes in the pet's comfort at any time. TPRC provides real-time physiological feedback, enabling the system to actively regulate temperature, avoid excessive physiological load under high or low temperature conditions, and reduce potential health risks.

[0131] Accurately calculate and optimize the comfort zone: TPRC can quantify the physiological tolerance range of pets and form an accurate calculation basis for the optimal temperature range. For example, through physiological response data, it is known that the heart rate and respiratory rate increase significantly when the temperature exceeds a certain threshold. The model can clearly define the physiological comfort zone of pets. In this way, the physiological balance of pets can be maintained without unnecessary adjustments, reducing the operating cost of the system.

[0132] Without TPRC, the accuracy of the temperature sensitivity model will be greatly reduced because the system will lack effective monitoring of changes in physiological load. Its disadvantages may include: the model cannot identify the physiological stress caused by high or low temperatures, resulting in the air conditioner not responding in time, which may affect the health of the pet. It cannot distinguish the physiological effects of different temperatures on different pets, resulting in insufficient personalization of temperature regulation, affecting the overall comfort of the pet.

[0133] 2.3 Temperature sensitivity index calculation

[0134] Comprehensive Sensitivity Index: The behavioral and physiological response curves are integrated into the Comprehensive Sensitivity Index (CSI) by weighted average method. CSI can be expressed as:

[0135] CSI=ω1·TBRC(T)+ω2·TPRC(T)

[0136] Among them, ω1 and ω2 are the weight coefficients of behavioral and physiological responses.

[0137] For some high-activity pets (such as dogs), their behavioral data may be more critical to temperature sensitivity. You can set ω1 = 0.6 and ω2 = 0.4. The ratio of each weight is adjusted between 0.3 and 0.7 (in practice, if the initial weight is not in this range, the training is more seriously overfitting) to meet the personalized needs of different types of pets. Through different weight ratios, CSI can find a balance between behavioral and physiological data. Behavioral responses are more sensitive to temperature, while physiological data reflect the direct impact of temperature on health. Therefore, reasonable weighting can avoid the deviation caused by a single dimension.

[0138] Weighted average method technical effect:

[0139] Improved accuracy: The CSI weighted average model provides weight adjustment space for different data sources, which can better ensure the accuracy of temperature control in dynamic monitoring.

[0140] Personalized adaptation: By setting a weight ratio suitable for pet characteristics, CSI achieves personalized temperature control, which meets the flexible control requirements of smart air conditioners.

[0141] Long-term tuning: As data accumulates, it can be gradually fine-tuned to make the system more intelligent and adaptive. This tuning process enhances the robustness of the temperature control system and can achieve optimal temperature comfort assessment in long-term applications.

[0142] 3. Temperature Sensitivity Model

[0143] 3.1 Model Structure

[0144] Model framework: When building the temperature sensitivity model, a deep neural network (DNN) model is used. The model input includes pet behavior data, physiological data and other characteristics, and the output is the optimal comfortable temperature range.

[0145] Input layer: The input layer contains multiple feature vectors, representing activity frequency, heart rate, respiration, body surface temperature, etc. Each feature vector is extracted through an independent fully connected layer.

[0146] Hidden layer: The hidden layer of the model uses the ReLU activation function, and the output of the neurons in the hidden layer is randomly set to 0 during training to prevent overfitting. The outputs of multiple hidden layers are weighted averaged to obtain the comprehensive sensitivity index CSI.

[0147] Output layer: The output layer generates the optimal comfortable temperature range for pets (T min ,T opt ,T max ), where T opt is the optimal temperature, T min and T max These are the minimum and maximum comfort temperatures, respectively.

[0148] 3.2 Model training and optimization

[0149] Data annotation and training set: Use historical data and expert annotations to build a training set. The data in the training set includes the pet’s behavior and physiological response at different temperatures, as well as the temperature range annotated by experts based on actual experience.

[0150] Loss function: Use Mean Squared Error (MSE) as the loss function to measure the difference between the model's predicted value and the actual value. The goal of the model is to minimize MSE, thereby improving the accuracy of the prediction.

[0151]

[0152] Among them, L is the loss function, N is the number of training samples, and are the optimal, minimum and maximum temperatures predicted by the model. The formula variables are analyzed as follows:

[0153] i: represents the index number of the sample, which is used to refer to the i-th sample in the training data. For a data set containing N samples, i takes values ​​from 1 to N.

[0154] The (i) in the upper right corner indicates the data of the i-th sample. These exponential symbols indicate the temperature value T min ,T opt ,T max are the actual value and predicted value for the i-th sample.

[0155] T opt (i) , T min (i) , T max (i) : Represents the actual values ​​of the optimal temperature, minimum temperature, and maximum temperature of the i-th sample. These values ​​are the target values ​​that the model hopes to achieve by adjusting the temperature.

[0156] The predicted value of the i-th sample, that is, the temperature-adjusted prediction result calculated by the model.

[0157] Overall explanation

[0158] The formula represents the model loss function L, which measures the deviation between the predicted value and the actual value of each sample through the mean square error (MSE). For each sample i, the sum of the squares of the difference between the predicted value and the actual value of its optimal, minimum, and maximum temperatures is calculated, and then the average value of all samples is taken to minimize the value of L as the optimization goal. This method helps to improve the prediction accuracy of the model under different temperature requirements, so that the temperature adjustment is more in line with the optimal comfort requirements of the pet.

[0159] The main advantages of using MSE in this solution include:

[0160] Precise adaptability: Compared with the Mean Absolute Error (MAE), MSE is more severe in penalizing large errors. This is especially important for air conditioning temperature control systems, because the system cannot tolerate comfort deviations caused by large temperature differences. The penalty for large errors makes the model more inclined to fine-tune temperature control to ensure a comfortable environment.

[0161] Easy to optimize and converge: MSE accumulates the squared errors of all samples, making it perform better in the gradient descent method and more compatible with the optimization process of deep learning models. Compared with other non-square penalty loss functions (such as MAE), MSE is more accurate in reflecting the overall trend of errors and is suitable for air-conditioning systems that require high precision and small tolerances.

[0162] Applicable to continuous output: MSE is suitable for the prediction of continuous output, especially in the temperature prediction scenario. MSE can directly calculate the temperature error, which is concise, effective and highly adaptable. If a binary or multi-classification loss function (such as the cross entropy loss function) is used, it is difficult to accurately predict and optimize continuous variables such as temperature.

[0163] Consistency with system tuning: The square property of MSE is consistent with the high error requirements of the temperature control system. By minimizing MSE, the air conditioning system can more quickly adapt to the real-time needs of pets and continuously improve temperature control comfort.

[0164] If MAE or other loss functions without square penalties are used, the model will be less sensitive to large errors, resulting in an inability to respond quickly when large errors occur. This may cause large deviations in air-conditioning temperature adjustment and affect the actual comfort experience of the pet.

[0165] If a non-continuous or non-differentiable loss function is used, such as some classification-based losses, it will be difficult to ensure convergence during optimization, and the training process will be more complicated, which is not conducive to model stability.

[0166] Failure to use MSE may cause the model training objectives to be dispersed and unable to focus on reducing errors, thereby reducing the adaptability and generalization ability of the model in different temperature scenarios, making it difficult for the air-conditioning system to adapt to the control needs of various ambient temperatures.

[0167] Model evaluation and tuning: The generalization ability of the model is evaluated through cross-validation, and the model structure and hyperparameters are adjusted according to the results. At the same time, the model is evaluated using validation sets and test sets to ensure its stability and accuracy on different data sets.

[0168] Specific control methods, such as Figure 3 As shown:

[0169] The system first detects the activity status in the room through the air conditioning sensing system. If the system detects that there is no one, it enters the pet activity detection. If the detection senses the pet activity, it will automatically switch to the "pet mode" (that is, the air conditioning first determines whether there is a person, and then uses the smart sensor to monitor and identify whether there is a pet when there is no one. That is, after the air conditioning detects that there is no one in the room, the air conditioning feedback is sent to the smart air conditioning system, and the smart air conditioning system starts to link with the pet smart sensor to monitor whether the pet exists; if it is linked with the pet wearable sensor, the pet wearable sensor has feedback to the smart air conditioning system, which means that the pet exists, and the smart air conditioning system feedbacks that the air conditioning switches to pet mode). At this time, the smart sensor in the pet activity space starts to continuously monitor the pet's behavioral data and physiological data, and communicates the data to the system server in real time. The cloud server temperature sensitivity model analyzes the pet's environmental temperature sensitivity in real time, and the model outputs the optimal temperature value suitable for the pet's current state. The cloud server sends the optimal temperature value to the air conditioning master control, and the air conditioning master control adjusts the temperature value of the air conditioning according to the optimal temperature value to ensure that the pet is in the best comfortable temperature range; if it enters the pet activity detection, the detection senses no pet activity, the air conditioning maintains the current mode (energy saving or shutdown). If the presence of a person is detected, the current mode continues.

[0170] In addition, users can manually input the specific information of their pets (such as breed, age, health status, etc.), and the system will adjust the temperature according to this information. At the same time, users can intervene in the adjustment of the system. Users can adjust the operating parameters of the air conditioner according to their understanding of the pet to meet the needs of the pet in a comfortable environment.

[0171] System temperature regulation strategy basis

[0172] For different pet breeds, the air conditioning system will automatically adjust the indoor temperature based on the temperature sensitivity model. The following are some temperature adjustment strategies for some pet breeds (including but not limited to these pet breeds):

[0173] For example, Shiba Inu: Shiba Inu is a breed with thicker hair, which is usually more adaptable to cold environments, but is prone to heatstroke in hot environments. Therefore, in summer, the air conditioning temperature should be kept low (20-22℃) and the wind speed can be moderate to ensure that the Shiba Inu does not feel overheated. In winter, the air conditioning is kept at 18-20℃.

[0174] For example, Pugs have short nasal cavities, which can make it difficult for them to breathe in high temperatures. Pugs are very sensitive to temperature changes. In summer, the air conditioning temperature should be kept between 22-24°C, and the wind speed can be lower to avoid direct wind. In winter, the temperature should be appropriately raised to 20-22°C to keep them comfortable.

[0175] For example, poodles have curly and dense hair, which is good at keeping heat, but they are prone to overheating in high temperature environments. In summer, the air conditioning temperature should be maintained at 21-23℃, and the wind speed can be adjusted to medium to help dissipate heat. In winter, the temperature should be maintained between 18-20℃.

[0176] For example, pet cats: Cats usually like warm environments, but too high temperatures may also cause them discomfort. Most domestic cats feel comfortable in an environment of 22-25℃, so the air conditioning temperature can be kept at around 24℃ in summer with a lower wind speed. In winter, it can be kept between 20-22℃.

[0177] The specific air-conditioning control method provided in the embodiment of the present application automatically switches to "pet mode" when it detects that there is no one indoors but there is a pet. The pet's sensitivity to the ambient temperature is identified by analyzing the pet's characteristics, and a comfortable ambient temperature value is recommended based on a sensitivity recommendation algorithm. The air conditioner dynamically adjusts the indoor temperature based on the recommended temperature value to ensure that the pet is in a comfortable and safe environment.

[0178] Based on the same inventive concept, Figure 4 As shown, the present application also provides an air conditioning control device 40, including:

[0179] The pet information acquisition module 41 is used to acquire pet information when entering the pet mode, and the pet information includes pet behavior data and pet physiological data. It should be noted that it detects whether there is anyone in the room; if there is no one in the room, it detects whether there is a pet; if it is detected that there is a pet, it enters the pet mode.

[0180] The model output acquisition module 42 is used to input the pet behavior data and the pet physiological data into the pre-trained temperature sensitivity model to obtain the model output.

[0181] First, the pet behavior data and the pet physiological data are subjected to time series analysis to eliminate noise and outliers, and then input into a pre-trained temperature sensitivity model.

[0182] The training step of the temperature sensitivity model includes:

[0183] Constructing an initial model and a training set, wherein the initial model includes an input layer, a hidden layer and an output layer, wherein the input layer includes a first input unit and a second input unit, wherein the hidden layer includes a first hidden layer and a second hidden layer, wherein the first input unit is connected to the first hidden layer, wherein the second input unit is connected to the second hidden layer, wherein the training sample includes a first training sample and a second training sample, wherein the first training sample is pet behavior data, and the second training sample is pet physiological data;

[0184] Inputting the first training sample into the first input unit to obtain a first output of the first hidden layer, and inputting the second training sample into the second input unit to obtain a second output of the second hidden layer;

[0185] The first output and the second output are weighted based on the first initial weight and the second initial weight to obtain a model output;

[0186] If the loss function is less than or equal to a threshold, updating the parameters of the first hidden layer and the second hidden layer and updating the first initial weight and the second initial weight until the loss function is greater than the threshold;

[0187] When the loss function is greater than the threshold, the current model is used as a trained temperature sensitivity model.

[0188] In addition, the first hidden layer and the second hidden layer respectively include a plurality of neurons, and the step of inputting the first training sample into the first input unit to obtain a first output of the first hidden layer, and inputting the second training sample into the second input unit to obtain a second output of the second hidden layer includes:

[0189] Inputting the first training sample into the first input unit, setting the outputs of a random number of neurons in the first hidden layer to 0, and obtaining a first output of the first hidden layer;

[0190] And / or, inputting the second training sample into the second input unit, setting the outputs of a random number of neurons in the second hidden layer to 0, and obtaining a second output of the second hidden layer.

[0191] The air conditioning temperature control module 43 is used to obtain a target temperature based on the model output and control the air conditioning to reach the target temperature.

[0192] As a preferred implementation of the embodiment of the present application, the pet information also includes the pet type, and also includes:

[0193] A target temperature is obtained according to the type of pet.

[0194] In one embodiment, obtaining the target temperature according to the pet type includes:

[0195] Obtaining a type coefficient corresponding to the pet type based on a preset corresponding relationship;

[0196] The product of the model output and the type coefficient is used as the target temperature.

[0197] In another embodiment, obtaining the target temperature according to the pet type includes:

[0198] Inputting the pet behavior data, the pet physiological data, and the pet type into a pre-trained temperature sensitivity model to obtain a model output;

[0199] The model output is used as the target temperature.

[0200] The air conditioning control device provided in the embodiment of the present application obtains pet information when entering the pet mode, and the pet information includes at least pet behavior data and pet physiological data, and then inputs the pet information into the pre-trained temperature sensitivity model to obtain the model output, and then obtains the target temperature based on the model output, and controls the air conditioning to reach the target temperature. The present application scheme controls the temperature of the air conditioning in combination with the pet's behavior data and physiological data, so that the air conditioning can provide a more accurate temperature according to the pet's behavior and physiology, without manual adjustment, thereby improving the user experience.

[0201] Based on the same inventive concept, Figure 5 As shown, the present application also provides an air conditioning control system 50, including:

[0202] at least one processor 51 and at least one memory 52;

[0203] The memory stores executable instructions of the processor;

[0204] The processor is configured to execute the air conditioning control method provided in the above embodiment.

[0205] The air conditioning control system provided by the embodiment of the present application stores executable instructions of the processor in the memory. When the executable instructions are executed, the processor can obtain pet information when entering the pet mode. The pet information includes at least pet behavior data and pet physiological data. Then the pet information is input into the pre-trained temperature sensitivity model to obtain the model output. Then the target temperature is obtained based on the model output, and the air conditioner is controlled to reach the target temperature. The present application scheme controls the temperature of the air conditioner in combination with the pet's behavior data and physiological data, so that the air conditioner can provide a more accurate temperature according to the pet's behavior and physiology without manual adjustment, thereby improving the user experience.

[0206] Based on the same inventive concept, the present application also provides an air conditioner, which applies the air conditioner control method provided in the above embodiment.

[0207] The air conditioner provided in the embodiment of the present application can obtain pet information when entering the pet mode by applying the air conditioner control method provided in the above embodiment, and the pet information includes at least pet behavior data and pet physiological data, and then input the pet information into the pre-trained temperature sensitivity model to obtain the model output, and then obtain the target temperature based on the model output, and control the air conditioner to reach the target temperature. The solution of the present application controls the temperature of the air conditioner in combination with the pet's behavior data and physiological data, so that the air conditioner can provide a more accurate temperature according to the pet's behavior and physiology, without manual adjustment, thereby improving the user experience.

[0208] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0209] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

Claims

1. An air conditioning control method, characterized in that: include: When entering the pet mode, obtaining pet information, the pet information including pet behavior data and pet physiological data; Inputting the pet behavior data and the pet physiological data into a pre-trained temperature sensitivity model to obtain a model output; A target temperature is obtained based on the model output, and the air conditioner is controlled to reach the target temperature.

2. The method according to claim 1, characterized in that: The training steps of the temperature sensitivity model include: Constructing an initial model and a training set, wherein the initial model includes an input layer, a hidden layer and an output layer, wherein the input layer includes a first input unit and a second input unit, wherein the hidden layer includes a first hidden layer and a second hidden layer, wherein the first input unit is connected to the first hidden layer, wherein the second input unit is connected to the second hidden layer, wherein the training sample includes a first training sample and a second training sample, wherein the first training sample is pet behavior data, and the second training sample is pet physiological data; Inputting the first training sample into the first input unit to obtain a first output of the first hidden layer, and inputting the second training sample into the second input unit to obtain a second output of the second hidden layer; The first output and the second output are weighted based on the first initial weight and the second initial weight to obtain a model output; If the loss function is less than or equal to a threshold, updating the parameters of the first hidden layer and the second hidden layer and updating the first initial weight and the second initial weight until the loss function is greater than the threshold; When the loss function is greater than the threshold, the current model is used as a trained temperature sensitivity model.

3. The method according to claim 2, characterized in that: The first hidden layer and the second hidden layer respectively include a plurality of neurons, and the step of inputting the first training sample into the first input unit to obtain a first output of the first hidden layer, and inputting the second training sample into the second input unit to obtain a second output of the second hidden layer includes: Inputting the first training sample into the first input unit, setting the outputs of a random number of neurons in the first hidden layer to 0, and obtaining a first output of the first hidden layer; And / or, inputting the second training sample into the second input unit, setting the outputs of a random number of neurons in the second hidden layer to 0, and obtaining a second output of the second hidden layer.

4. The method according to claim 1, characterized in that: The pet information also includes the pet type and: A target temperature is obtained according to the type of pet.

5. The method according to claim 4, characterized in that: The step of obtaining the target temperature according to the pet type includes: Obtaining a type coefficient corresponding to the pet type based on a preset corresponding relationship; The product of the model output and the type coefficient is used as the target temperature.

6. The method according to claim 4, characterized in that: The step of obtaining the target temperature according to the pet type includes: Inputting the pet behavior data, the pet physiological data, and the pet type into a pre-trained temperature sensitivity model to obtain a model output; The model output is used as the target temperature.

7. The method according to claim 1, characterized in that Also includes: Check whether there is someone in the room; If there is no one in the room, detect whether there is a pet; If a pet is detected, the system will enter pet mode.

8. The method according to claim 1, characterized in that: The step of inputting the pet behavior data and the pet physiological data into a pre-trained temperature sensitivity model comprises: Perform time series analysis on the pet behavior data and the pet physiological data to eliminate noise and outliers.

9. An air conditioning control device, characterized in that: include: A pet information acquisition module, used to acquire pet information when entering pet mode, wherein the pet information includes pet behavior data and pet physiological data; A model output acquisition module, used for inputting the pet behavior data and the pet physiological data into a pre-trained temperature sensitivity model to obtain a model output; The air conditioner temperature control module is used to obtain a target temperature based on the model output and control the air conditioner to reach the target temperature.

10. An air conditioning control system, characterized in that: include: at least one processor and at least one memory; The memory stores executable instructions of the processor; The processor is configured to execute the method according to any one of claims 1 to 8.

11. An air conditioner, characterized in that: The method according to any one of claims 1 to 8 is used.

Images