Thermal comfort regulation and control method and device based on visual perception and global sensitivity analysis

Abstract

The invention relates to the technical field of intelligent building environment control and energy management, in particular to a thermal comfort regulation and control method and device based on visual perception and global sensitivity analysis, and constructs an attribute recognition model based on collaborative optimization of a lightweight visual perception network and an attention mechanism. High-precision dynamic capture of human body attributes is realized through the model; human body metabolic rate information is calculated based on the recognized human body attributes, then a thermal comfort model based on global sensitivity analysis and a hybrid intelligent algorithm is constructed, human body thermal comfort is calculated through the model, and finally working parameters of an air conditioner or a heat pump are adjusted based on the human body thermal comfort. The contradiction between environment parameter real-time response and energy consumption optimization is effectively solved, and an innovative solution is provided for building energy conservation and personalized thermal comfort management.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent building environment control and energy management, and particularly to a thermal comfort regulation method and device based on visual perception and global sensitivity analysis. Background Art

[0002] With the rapid development of the global social economy, while people pay attention to energy conservation, emission reduction, green and low-carbon, higher requirements are also put forward for the performance of air-conditioning systems and the comfort of living environments. Today's air-conditioning systems must create a good indoor living environment while saving energy and protecting the environment. Therefore, the energy-saving design and operation control of air-conditioning systems are particularly important.

[0003] With the deepening of the concepts of smart buildings and healthy living environments, precise regulation of the indoor environment faces multiple challenges: 1) Most data collection of subjects, environmental variables, and human variables in the venue is carried out through various monitoring devices. However, the method of wearing monitoring devices is not easy to promote, and traditional convolutional networks are difficult to balance fine-grained feature extraction and computational efficiency, resulting in low recognition accuracy of key attributes such as clothing thermal resistance; 2) Thermal comfort evaluation relies on empirical formulas, ignoring the multi-factor coupling effect, making it difficult to quantify the non-linear impact of variables such as season and metabolic rate on thermal perception, and having a lag in response to individual differences and environmental mutations; 3) The energy consumption control strategy is single, lacking the dynamic adjustment ability based on real-time thermal comfort feedback, and it is difficult to balance comfort and energy efficiency. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to propose a thermal comfort regulation method and device based on visual perception and global sensitivity analysis to solve the above technical problems.

[0005] Based on the above purpose, the present invention provides a thermal comfort regulation method based on visual perception and global sensitivity analysis, including: Obtain environmental information and indoor human image information; Input the indoor human image information into a pre-trained attribute recognition model to identify human attribute information; Calculate the metabolic rate information of the person in the indoor human image information based on the human attribute information; Input the environmental information, human attribute information, and metabolic rate information into a pre-trained thermal comfort model to calculate human thermal comfort; Adjust the working parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort.

[0006] As an optional implementation manner, the environmental information includes indoor temperature information, outdoor temperature information, indoor humidity information, and indoor wind speed information; The human body attribute information includes gender information, age information, exercise posture information, and clothing thermal resistance information; Based on the gender information, age information, and exercise posture information, identify the metabolic rate information of the person in the indoor person image information.

[0007] As an optional implementation manner, the training process of the attribute recognition model is as follows: Obtain multi-dimensional human body image data based on a standardized indoor office scenario; Preprocess the multi-dimensional human body image data; Construct an initial attribute recognition model with PP-LCNet as the backbone network; Input the preprocessed multi-dimensional human body image data into the initial attribute recognition model for convolution calculation to obtain a feature map; Input the feature map into a channel attention module and a spatial attention module for calculation respectively. The channel attention module is used to strengthen the clothing texture features in the feature map, and the spatial attention module is used to strengthen the recognition of the key areas of the human body in the feature map; Input the feature map after strengthening the clothing texture features and strengthening the recognition of the key areas of the human body into an average pooling layer for average pooling, and then input the feature map after average pooling into a fully connected layer for expansion to obtain an expanded feature map; Input the expanded feature map into a classifier to obtain image data with human body attribute labels; Calculate the loss function, and repeat the above training process until the calculation result of the loss function reaches a threshold, that is, the model converges, and obtain a trained attribute recognition model.

[0008] As an optional implementation manner, the preprocessing includes normalization processing, horizontal flipping, and random cropping.

[0009] As an optional implementation manner, the training process of the thermal comfort model is as follows: Obtain a historical thermal comfort data set, which includes historical gender information, historical age information, historical clothing thermal resistance information, historical metabolic rate information, historical indoor temperature information, historical outdoor temperature, historical indoor humidity information, and historical indoor wind speed information; Preprocess the data in the historical thermal comfort data set; Construct an initial thermal comfort model based on an FNN network. The initial thermal comfort model uses gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, relative humidity, and wind speed as input feature parameters, and uses thermal comfort target as the output feature parameter; Input the preprocessed thermal comfort dataset into the initial thermal comfort model for training, adjust the model parameters until the model converges, and obtain the trained thermal comfort model.

[0010] As an alternative implementation, the preprocessing of the data in the historical thermal comfort dataset includes: Remove outliers, fill in missing data, and process columns with non-numeric labels using one-hot encoding.

[0011] As an alternative implementation, the method for selecting the input feature parameters of the thermal comfort model is: Obtain a historical dataset, where the historical dataset includes historical season information season, historical age information age, historical gender information gender, historical person's height information ht, historical weight information wt, historical indoor temperature information ta, historical operative temperature information top, historical radiant temperature information tr, historical outdoor temperature information tg, historical indoor humidity information rh, historical indoor wind speed information vel, historical human metabolic rate information met, historical clothing insulation information clo, historical outdoor air temperature information t_out, historical outdoor humidity information rh_out, historical fan status information fan, historical window status information window, and historical door status information door; Preprocess the data information in the historical dataset; Construct an initial Cubist model, where the initial Cubist model uses season information, age information, gender information, person's height information, weight information, indoor temperature information, operative temperature information, radiant temperature information, outdoor temperature information, indoor humidity information, indoor wind speed information, human metabolic rate information, clothing insulation information, outdoor air temperature information, outdoor humidity information, fan status information, window status information, and door status information as input feature parameters, and uses thermal sensation as the output feature parameter; Input the preprocessed historical dataset into the initial Cubist model to train the model, adjust the model parameters until the model converges, and obtain the trained Cubist model; Adopt the Sobol sequence sampling method to extract multiple groups of feature sample combinations from the historical thermal comfort dataset; Input multiple groups of the feature sample combinations into the trained Cubist model respectively to obtain the predicted thermal comfort values output by the model; According to the variance decomposition theory, based on each predicted thermal comfort value output by the model, calculate the main effect index S1 and the total effect index ST of the corresponding feature sample combination, and evaluate the influence degree of various factors on thermal comfort based on the calculation results; Based on the evaluation results, gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, relative humidity, and wind speed are selected as the input characteristic parameters of the thermal comfort model.

[0012] As an alternative implementation, the preprocessing of the data information in the historical data set includes: For the missing historical indoor temperature information ta, historical working temperature information top, historical radiant temperature information tr, historical outdoor temperature information tg, historical indoor humidity information rh, historical outdoor air temperature information t_out, historical outdoor humidity information rh_out, historical clothing thermal resistance information clo, historical fan status information fan, historical window status information window, historical door status information door, and historical season information season in the historical data set, fill them with the mode according to the season when filling; Process the historical gender information gender in the historical data set with one-hot encoding; For the missing historical age information age, historical personnel height information ht, historical weight information wt, historical human metabolic rate information met, and historical indoor wind speed information vel in the historical data set, fill them with the average value.

[0013] As an alternative implementation, the adjustment of the working parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort includes: Take the difference between the actual thermal comfort value obtained by calculating based on the human thermal comfort and the target thermal comfort value; If the difference is outside the preset range, adjust the working parameters of the air conditioner or heat pump to make the actual thermal comfort value approach the target thermal comfort value.

[0014] Corresponding to the thermal comfort control method, the present invention also provides a thermal comfort control device based on visual perception and global sensitivity analysis, including: An acquisition module for acquiring environmental information and indoor human image information; An identification module for inputting the indoor human image information into a pre-trained attribute identification model to identify human attribute information; A first calculation module for calculating the metabolic rate information of the person in the indoor human image information based on the human attribute information; A second calculation module for inputting the environmental information, human attribute information, and metabolic rate information into a pre-trained thermal comfort model to calculate human thermal comfort; An adjustment module for adjusting the working parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort.

[0015] Advantages of the present invention: The present invention constructs an attribute recognition model based on the collaborative optimization of a lightweight visual perception network and an attention mechanism, and realizes high-precision dynamic capture of human attributes through this model; and based on the recognized human attributes, calculates the human metabolic rate information, and then constructs a thermal comfort model based on global sensitivity analysis and a hybrid intelligent algorithm, and calculates the human thermal comfort through this model. Finally, based on the human thermal comfort, adjusts the working parameters of the air conditioner or heat pump, effectively solving the contradiction between real-time response of environmental parameters and energy consumption optimization, and providing an innovative solution for building energy conservation and personalized thermal comfort management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 Schematic flow chart of the thermal comfort regulation method according to the embodiment of the present invention; Figure 2 Calculation results obtained through sensitivity analysis in the embodiment of the present invention; Figure 3 Schematic flow chart of the thermal comfort regulation device according to the embodiment of the present invention; Figure 4 Schematic diagram of the device principle of the specific embodiment of the present invention; Figure 5 Floor plan of the building of the specific embodiment of the present invention; Figure 6 Building physical model diagram of the specific embodiment of the present invention; Figure 7 Energy consumption comparison chart between the embodiment of the present invention and the prior art. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the following further details the present invention in conjunction with specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. The terms such as "connected" or "linked" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right" are only used to represent relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0020] As an embodiment of the present invention, a thermal comfort regulation method based on visual perception and global sensitivity analysis is provided, including: Obtain environmental information and indoor human image information; Input the indoor human image information into a pre-trained attribute recognition model to recognize human attribute information; Calculate the metabolic rate information of the person in the indoor human image information based on the human attribute information; Input the environmental information, human attribute information and metabolic rate information into a pre-trained thermal comfort model to calculate human thermal comfort; Adjust the working parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort.

[0021] In the present invention, an attribute recognition model based on the collaborative optimization of a lightweight visual perception network and an attention mechanism is constructed to achieve high-precision dynamic capture of human attributes through this model; and based on the recognized human attributes, the human metabolic rate information is calculated, and then a thermal comfort model based on global sensitivity analysis and a hybrid intelligent algorithm is constructed to calculate human thermal comfort through this model. Finally, the working parameters of the air conditioner or heat pump are adjusted based on the human thermal comfort, effectively solving the contradiction between real-time response of environmental parameters and energy consumption optimization, and providing an innovative solution for building energy conservation and personalized thermal comfort management.

[0022] As a specific implementation manner of the present invention, as Figure 1 shown, a thermal comfort regulation method based on visual perception and global sensitivity analysis is provided, including: S100. Obtain environmental information and indoor human image information; Among them, the environmental information includes indoor temperature information, outdoor temperature information, indoor humidity information, and indoor wind speed information.

[0023] S200. Input the indoor human image information into a pre-trained attribute recognition model to recognize human attribute information; Among them, the human attribute information includes gender information, age information, motion posture information, and clothing thermal resistance information.

[0024] S300. Calculate the metabolic rate information of the person in the indoor human image information based on the human attribute information; Among them, the metabolic rate information of the person in the indoor human image information is recognized based on the gender information, age information, and motion posture information.

[0025] S400. Input the environmental information, human attribute information, and metabolic rate information into a pre-trained thermal comfort model to calculate human thermal comfort; S500. Adjust the working parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort.

[0026] As an optional implementation manner, the training process of the attribute recognition model is as follows: Obtain multi-dimensional human image data based on a standardized indoor office scenario; Preprocess the multi-dimensional human image data. Optionally, the preprocessing includes normalization, horizontal flipping, and random cropping to ensure the robust recognition ability of the model for complex human attributes; Construct an initial attribute recognition model with PP-LCNet as the backbone network; Input the preprocessed multi-dimensional human image data into the initial attribute recognition model for convolution calculation to obtain a feature map; Input the feature map into a channel attention module and a spatial attention module for calculation. The channel attention module is used to strengthen the clothing texture features in the feature map, and the spatial attention module is used to strengthen the recognition of the key areas of the human body in the feature map; Input the feature map after strengthening the clothing texture features and strengthening the recognition of the key areas of the human body into an average pooling layer for average pooling, and then input the feature map after average pooling into a fully connected layer for expansion to obtain an expanded feature map; Input the expanded feature map into a classifier to obtain image data with human attribute labels; Calculate the loss function, and repeat the above training process until the calculation result of the loss function reaches a threshold, that is, the model converges, and obtain a trained attribute recognition model.

[0027] Optionally, the dataset construction of the present invention is based on a standardized indoor office scenario, and a camera is used to collect multi-dimensional human body images. The experimental samples cover more than 30 subjects. Through a multi-batch combined experimental design, the system collects human body image data of different genders, multiple postures (turning / standing / sitting), and variable clothing thermal resistance states (including the behavior of putting on and taking off down jackets). To enhance the data representation ability, a standardized action paradigm (including axial rotation, body posture transformation, etc.) is implemented during the construction process, and data augmentation strategies such as random cropping are adopted to ensure the robust recognition ability of the model for complex human attributes.

[0028] In terms of the selection of the annotation format, the same annotation format as PA100k (Pedestrian Attribute Dataset) is adopted, that is, the picture name is on the left and the attribute features are on the right, and the attribute features are represented by a segment of 0 / 1 numbers. And it is divided into a training set, a validation set, and a test set according to 6:2:2.

[0029] The embodiment of the present invention is based on the Baidu PaddlePaddle platform to train a deep learning model. PP-LCNet is used as the backbone network to construct an attribute recognition model. The learning rate is set to 0.001, and the number of iterations is 50. It not only retains the high-precision characteristics of the currently popular backbone feature extraction network but also reduces the computational amount. At the same time, the CBAM attention network that combines convolution and attention mechanisms is adopted to analyze the image from both spatial and channel aspects. The spatial attention focuses on the key areas of the human body, and the channel attention strengthens the clothing texture features. By adopting this attention network to actively learn the important information in the low-level features and fuse it with the high-level features, the model can obtain both low-level fine-grained information and high-level semantic information, thereby improving the accuracy.

[0030] For the convenience of understanding, the following provides a specific embodiment to illustrate the specific training process of the attribute recognition model: Use a Hikvision network camera (model DS-2CD3386FWDA4-LS) to collect RGB images of indoor personnel, and normalize them into a tensor format (244*244 in size with 3 channels, that is, 244*244*3). Subsequently, data augmentation is performed through horizontal flipping and random cropping to improve the robustness of the model; The input image passes through a 3*3 standard convolution (Stride=2) in the first layer, and the output feature tensor is 112*112*16, extracting low-order edge and texture features; The feature map sequentially passes through 3 depthwise separable convolution modules (each module has 2 convolution layers). Each depthwise separable convolution kernel is 3*3, and the number of channels doubles in turn. At this time, the size of the feature image is 14*14*256; Enter the last depthwise separable convolution module, the convolution kernel size is 5*5, and the stride is 2. At this time, the size of the feature map is 7*7*512; The CBAM attention module is added, and the feature map enters the channel attention module and the spatial attention module respectively. In the channel attention module, the weights in the channel dimension are learned. First, the spatial dimension is compressed to generate a channel description vector with a feature map size of 1*1*512. Subsequently, the maximum activation value of each channel is extracted. The two pooling results are respectively input into the same multi-layer perceptron (MLP) and the Sigmoid activation function is adopted to obtain the channel weights, strengthening the clothing texture features. Through the spatial attention module, the feature information is compressed to 7*7*1. After max pooling, the two pooling results are concatenated to obtain the spatial attention weights, enhancing the ability to identify key body regions; The image information after attention enhancement is input into the average pooling layer and enters the fully connected layer for expansion. At this time, the size is 1*1*1280; Finally, the human body attribute labels are output through the classifier (Softmax classification layer); The loss function is calculated. The weighted cross-entropy loss function is adopted, and then backpropagation is performed to calculate the gradients of each layer and update the weight information until the model converges.

[0031] Through the above method, the neural network converges quickly within 50 iterations. The training loss decline curve is smooth, and the accuracy of the validation set is steadily improved; the model accuracy obtained by the final model is f1: 0.96547, acc: 0.7818, prec: 0.96846, recall: 0.96249, meeting the high-precision requirements of the building energy-saving system.

[0032] As an alternative implementation, the training process of the thermal comfort model is as follows: Obtain the historical thermal comfort dataset, which includes historical gender information, historical age information, historical clothing thermal resistance information, historical metabolic rate information, historical indoor temperature information, historical outdoor temperature, historical indoor humidity information, and historical indoor wind speed information; Preprocess the data in the historical thermal comfort dataset, including: removing outliers, filling in missing data, and processing non-numeric marked columns with one-hot encoding; Construct an initial thermal comfort model based on the FNN network. The initial thermal comfort model uses gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, relative humidity, and wind speed as input feature parameters, and the thermal comfort target as the output feature parameter; Input the preprocessed thermal comfort dataset into the initial thermal comfort model for training, adjust the model parameters until the model converges, and obtain the trained thermal comfort model.

[0033] The present invention constructs a thermal comfort prediction model based on a forward neural network (FNN), and uses the ASHRAE GTDB-II standard dataset for training. For the algorithm optimization part, a hybrid intelligent optimization strategy is designed: Given that the particle swarm optimization (PSO) has the advantages of fast convergence speed, strong local optimization ability and simple parameter adjustment, PSO is used to achieve fast convergence in the initial stage of training; Given that the gravitational search algorithm (GSA) has strong global exploration ability and high solution space traversal, GSA is introduced in the later stage of training for fine search. Through the synergistic effect of the two-stage optimization algorithm, the prediction accuracy of the FNN model for the indoor environment thermal comfort is significantly improved.

[0034] For the convenience of understanding, a specific embodiment is provided below to illustrate the specific training process of the thermal comfort model: Obtain a historical thermal comfort dataset, which includes historical gender information, historical age information, historical clothing thermal resistance information, historical metabolic rate information, historical indoor temperature information, historical outdoor temperature, historical indoor humidity information, and historical indoor wind speed information; Preprocess the data in the dataset, including removing outliers, filling in missing data, and processing columns with non-numeric labels using one-hot encoding. A total of the above 8 parameters are selected, including: gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, relative humidity, and wind speed as input features; The thermal comfort target is used as the output, including 7 classifications: -3, -2, -1, 0, 1, 2, 3.

[0035] Perform Min—Max normalization processing on the input features of the dataset, and divide the dataset into a training set, a validation set, and a test set according to 6:2:2.

[0036] Construct an FNN network, with 8 input nodes and 7 output nodes, and use the Softmax function to output the probability distribution; In the design of the middle hidden layer, 3 hidden layers are selected, with 20 neurons in each layer, and the Relu activation function is used. The loss function uses the cross-entropy loss function.

[0037] Expand all the weights and biases of the FNN into vectors, and each particle represents a set of network parameters. The loss function of the validation set is used as the fitness value function of the optimization algorithm to guide the particle update.

[0038] Initialize the FNN network weight parameters, input the data into the FNN network; calculate the network output and the loss function; update the particle velocity and position according to the optimization algorithm; repeat this process until the maximum number of iterations. Finally, the prediction accuracy of the model is 0.88, meeting the actual application requirements.

[0039] As an alternative implementation, the method for selecting the input characteristic parameters of the thermal comfort model is as follows: Obtain a historical dataset, where the historical dataset includes historical season information "season", historical age information "age", historical gender information "gender", historical personnel height information "ht", historical weight information "wt", historical indoor temperature information "ta", historical operative temperature information "top", historical radiant temperature information "tr", historical outdoor temperature information "tg", historical indoor humidity information "rh", historical indoor wind speed information "vel", historical human metabolic rate information "met", historical clothing insulation information "clo", historical outdoor air temperature information "t_out", historical outdoor humidity information "rh_out", historical fan status information "fan", historical window status information "window", and historical door status information "door"; Preprocess the data information in the historical dataset, including: for the missing historical indoor temperature information "ta", historical operative temperature information "top", historical radiant temperature information "tr", historical outdoor temperature information "tg", historical indoor humidity information "rh", historical outdoor air temperature information "t_out", historical outdoor humidity information "rh_out", historical clothing insulation information "clo", historical fan status information "fan", historical window status information "window", historical door status information "door", and historical season information "season" in the historical dataset, perform mode filling by season during filling; Process the historical gender information "gender" in the historical dataset using one-hot encoding; Perform average filling for the missing historical age information "age", historical personnel height information "ht", historical weight information "wt", historical human metabolic rate information "met", and historical indoor wind speed information "vel" in the historical dataset.

[0040] Construct an initial Cubist model, where the initial Cubist model uses season information, age information, gender information, personnel height information, weight information, indoor temperature information, operative temperature information, radiant temperature information, outdoor temperature information, indoor humidity information, indoor wind speed information, human metabolic rate information, clothing insulation information, outdoor air temperature information, outdoor humidity information, fan status information, window status information, and door status information as input characteristic parameters, and uses thermal comfort "thermal_sensation" as the output characteristic parameter; Input the preprocessed historical dataset into the initial Cubist model to train the model, and adjust the model parameters until the model converges to obtain a trained Cubist model; Adopt the Sobol sequence sampling method to extract multiple groups of characteristic sample combinations from the historical thermal comfort dataset; Input multiple groups of the combined feature samples into the trained Cubist model respectively to obtain the predicted values of thermal comfort output by the model; According to the variance decomposition theory, based on each predicted value of thermal comfort output by the model, calculate the main effect index S1 and the total effect index ST of the corresponding combined feature samples, and evaluate the influence degree of various factors on thermal comfort based on the calculation results; Select gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, relative humidity, and wind speed as the input feature parameters of the thermal comfort model based on the evaluation results.

[0041] The present invention discloses a thermal comfort evaluation method based on the ASHRAE GTDB-II dataset. Through data preprocessing (mode / mean filling, One-hot encoding), Cubist model construction and parameter optimization (MSE = 1.03), combined with Sobol global sensitivity analysis (1000 samples), key influencing factors such as season, temperature and humidity are determined to achieve accurate prediction and control of the building thermal environment.

[0042] For the convenience of understanding, the following provides a specific embodiment to illustrate the selection method of the input feature parameters of the thermal comfort model: Obtain the historical dataset as ASHRAE GTDB-II (ASHRAE Global Thermal Comfort DatabaseII), where the historical dataset includes historical season information season, historical age information age, historical gender information gender, historical person height information ht, historical weight information wt, historical indoor temperature information ta, historical working temperature information top, historical radiant temperature information tr, historical outdoor temperature information tg, historical indoor humidity information rh, historical indoor wind speed information vel, historical human metabolic rate information met, historical clothing thermal resistance information clo, historical outdoor air temperature information t_out, historical outdoor humidity information rh_out, historical fan status information fan, historical window status information window, and historical door status information door; among them, tg is the temperature obtained by the dataset producer from the meteorological bureau at that time and place, and t_out is the temperature obtained by the dataset producer by installing sensors outdoors; Perform data preprocessing on the data and fill in the missing data. Specifically, when filling in types such as temperature, humidity, and clothing thermal resistance, fill in the mode according to the season; for the column gender with non-numeric marks, perform one-hot encoding; and fill in the average for data such as age, height, weight, and metabolic rate that has nothing to do with the climate season.

[0043] Construct an initial Cubist model, where the initial Cubist model takes seasonal information, age information, gender information, personnel height information, weight information, indoor temperature information, working temperature information, radiant temperature information, outdoor temperature information, indoor humidity information, indoor wind speed information, human metabolic rate information, clothing thermal resistance information, outdoor air temperature information, outdoor humidity information, fan status information, window status information, and door status information as input feature parameters, and takes thermal comfort (thermal_sensation) as the output feature parameter; Divide the preprocessed data into a training set and a test set at a ratio of 8:2.

[0044] Use the training set to train the initial Cubist model, and adjust the model parameters to optimize the model performance. Determine the best parameter combination through cross-validation to make the model achieve a better fitting effect. Use the test set data to evaluate the trained Cubist model, and calculate the mean square error (MSE) of the model as the evaluation index. The final result is 1.03288, proving that the model has high accuracy.

[0045] Perform Sobol global sensitivity analysis using the constructed Cubist model, where committees is set to 1, neighbors is set to 0, rules is set to 99, maxdepth is set to 1, and mincases is set to 10. Adopt Sobol sequence sampling to generate 1000 input feature sample combinations for subsequent sensitivity analysis calculations. For each sampled input feature sample combination, calculate the corresponding output prediction value through the Cubist model. According to the variance decomposition theory, calculate the main effect index (S1) and total effect index (ST) of each input feature. The calculation results are as Figure 2 shown. The figure shows that the total effect and main effect of eight factors, namely gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, indoor humidity, and indoor wind speed, are greater than other factors. Therefore, select gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, indoor humidity, and indoor wind speed as the calculation parameters of the thermal comfort model.

[0046] As an alternative implementation, adjusting the working parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort includes: Subtract the actual thermal comfort value obtained from the calculation of the human thermal comfort from the target thermal comfort value; If the difference is outside the preset range, adjust the working parameters of the air conditioner or heat pump to make the actual thermal comfort value approach the target thermal comfort value.

[0047] Corresponding to the above-mentioned thermal comfort control method, the present invention further provides a thermal comfort control device based on visual perception and global sensitivity analysis, as Figure 3 shown, including: An acquisition module 10, configured to obtain environmental information and indoor human image information; An identification module 20, configured to input the indoor human image information into a pre-trained attribute recognition model to identify human attribute information; A first calculation module 30, configured to calculate the metabolic rate information of the person in the indoor human image information based on the human attribute information; A second calculation module 40, configured to input the environmental information, human attribute information, and metabolic rate information into a pre-trained thermal comfort model to calculate human thermal comfort; An adjustment module 50, configured to adjust the working parameters of an air conditioner or a heat pump based on the calculation result of the human thermal comfort.

[0048] Those skilled in the art should understand that the technical effects of the thermal comfort control device are the same as those of the thermal comfort control method, and will not be elaborated herein.

[0049] As Figure 4 shown, an embodiment of the present invention further provides a schematic diagram of a device applying this solution. In this figure, a camera and a temperature and humidity sensor are used for data acquisition, and then the algorithm of the embodiment of the present invention is used for data processing.

[0050] To facilitate the understanding of the technical solution of the present invention by those skilled in the art, an embodiment with a specific application scenario is provided below. Embodiment

[0051] The research object is the first-floor hall of a building in a university in Tianjin, and its floor plan is as Figure 5 shown. According to the floor plan, the indoor dimension table of the building as shown in Table 1 is obtained.

[0052] Table 1 Indoor dimension table of the building

[0053] The hall is connected to the kitchen, equipment room, bathroom, and an office. There is a staircase leading to the second floor in the northwest corner of the room. There are a total of four independent fan coil units (heat pump output) in the hall. The size of the fan coil units is 0.90 m × 0.50 m × 0.20 m, and the bottom of them is 0.3 m from the ground. The four fan coil units are respectively located on the south side of the two windows on the east wall, the west side of the window on the south wall, and near the entrance on the west wall of the hall. When the fan coil units are turned on, the air supply outlet size is 0.90 m × 0.10 m. According to the actual situation of the fan coil blades, the air supply angle is set at 45° obliquely upward. The air return opening of the fan coil unit is located at the bottom, with a size of 0.90 m × 0.10 m. There is a conference table with a size of 4.50 m × 3.00 m × 1.40 m opposite the kitchen. For the convenience of the experiment, the staircase is simplified to a wall, and the space where the equipment room, bathroom, and office are connected to the hall is not modeled. The physical model of the office is as Figure 6 shown.

[0054] This experiment was carried out in this building. The personnel attribute information was collected in real time through the monitoring cameras deployed on the roof, and the environmental parameters were obtained by using the temperature sensors distributed indoors and outdoors synchronously. Three students were invited to participate in the experiment, which was carried out in two days (the outdoor meteorological conditions were basically the same during the experiment), and the daily monitoring period was from 10:00 to 18:00. On the first day, the natural state operation mode was adopted (which means not using the present invention to control the heat pump and air conditioner, and the indoor personnel remotely control the temperature of the air conditioner and heat pump according to their usual living habits), and no control strategy was applied to the system; on the second day, the intelligent regulation system of the present invention was enabled. The experimental data show that ( Figure 7 ): In the two operation modes, the hourly power consumption after the system is enabled is significantly lower than that in the natural state. Through quantitative analysis, after adopting the intelligent regulation system, the daily average comprehensive energy consumption is reduced by 5.6%, verifying the actual energy-saving effect of the present invention on buildings.

[0055] Those of ordinary skill in the art should understand that the discussion of any of the above embodiments is only exemplary, and is not intended to imply that the scope of the present invention (including the claims) is limited to these examples; under the idea of the present invention, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of the present invention as described above, which are not provided in detail for the sake of brevity.

[0056] The present invention is intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A thermal comfort regulation method based on visual perception and global sensitivity analysis, characterized in that, Including: Obtain environmental information and indoor human image information; Input the indoor human image information into a pre-trained attribute recognition model to recognize human attribute information; Calculate the metabolic rate information of the person in the indoor human image information based on the human attribute information; Input the environmental information, human attribute information, and metabolic rate information into a pre-trained thermal comfort model to calculate human thermal comfort; Adjust the working parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort; The training process of the attribute recognition model is as follows: Obtain multi-dimensional human image data based on a standardized indoor office scenario; Preprocess the multi-dimensional human image data; Construct an initial attribute recognition model with PP-LCNet as the backbone network; Input the preprocessed multi-dimensional human image data into the initial attribute recognition model for convolution calculation to obtain a feature map; Input the feature map into a channel attention module and a spatial attention module for calculation. The channel attention module is used to strengthen the clothing texture features in the feature map, and the spatial attention module is used to strengthen the recognition of the key areas of the human body in the feature map; Input the feature map after strengthening the clothing texture features and strengthening the recognition of the key areas of the human body into an average pooling layer for average pooling, and then input the feature map after average pooling into a fully connected layer for expansion to obtain an expanded feature map; Input the expanded feature map into a classifier to obtain image data with human attribute labels; Calculate the loss function, and repeat the above training process until the calculation result of the loss function reaches a threshold, that is, the model converges, and obtain a trained attribute recognition model.

2. The thermal comfort regulation method based on visual perception and global sensitivity analysis according to claim 1, characterized in that The environmental information includes indoor temperature information, outdoor temperature information, indoor humidity information, and indoor wind speed information; The human attribute information includes gender information, age information, movement posture information, and clothing thermal resistance information; Identify the metabolic rate information of the person in the indoor human image information based on the gender information, age information, and movement posture information.

3. The thermal comfort regulation method based on visual perception and global sensitivity analysis according to claim 1, characterized in that The preprocessing includes normalization, horizontal flipping, and random cropping.

4. The thermal comfort regulation method based on visual perception and global sensitivity analysis according to claim 2, wherein The training process of the thermal comfort model is as follows: Obtain a historical thermal comfort dataset, which includes historical gender information, historical age information, historical clothing thermal resistance information, historical metabolic rate information, historical indoor temperature information, historical outdoor temperature, historical indoor humidity information, and historical indoor wind speed information; Preprocess the data in the historical thermal comfort dataset; Construct an initial thermal comfort model based on the FNN network. The initial thermal comfort model takes gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, relative humidity, and wind speed as input feature parameters and the thermal comfort target as the output feature parameter; Input the preprocessed thermal comfort dataset into the initial thermal comfort model for training, and adjust the model parameters until the model converges to obtain a trained thermal comfort model.

5. The thermal comfort regulation method based on visual perception and global sensitivity analysis according to claim 4, wherein The preprocessing of the data in the historical thermal comfort dataset includes: Remove outliers, fill in missing data, and process non-numeric marked columns with one-hot encoding.

6. The thermal comfort control method based on visual perception and global sensitivity analysis according to claim 4, wherein The method for selecting the input characteristic parameters of the thermal comfort model is as follows: Obtain a historical data set, which includes historical season information season, historical age information age, historical gender information gender, historical personal height information ht, historical weight information wt, historical indoor temperature information ta, historical operative temperature information top, historical radiant temperature information tr, historical outdoor temperature information tg, historical indoor humidity information rh, historical indoor wind speed information vel, historical human metabolic rate information met, historical clothing thermal resistance information clo, historical outdoor air temperature information t_out, historical outdoor humidity information rh_out, historical fan status information fan, historical window status information window, and historical door status information door; Preprocess the data information in the historical data set; Construct an initial Cubist model, which uses season information, age information, gender information, personal height information, weight information, indoor temperature information, operative temperature information, radiant temperature information, outdoor temperature information, indoor humidity information, indoor wind speed information, human metabolic rate information, clothing thermal resistance information, outdoor air temperature information, outdoor humidity information, fan status information, window status information, and door status information as input characteristic parameters, and uses thermal comfort thermal_sensation as the output characteristic parameter; Input the preprocessed historical data set into the initial Cubist model to train the model, and adjust the model parameters until the model converges to obtain a trained Cubist model; Adopt the Sobol sequence sampling method to extract multiple groups of characteristic sample combinations from the historical thermal comfort data set; Input multiple groups of the characteristic sample combinations into the trained Cubist model respectively to obtain the predicted thermal comfort values output by the model; According to the variance decomposition theory, based on each predicted thermal comfort value output by the model, calculate the main effect index S1 and the total effect index ST of the corresponding characteristic sample combination, and evaluate the influence degree of various factors on thermal comfort based on the calculation results; Based on the evaluation results, select gender, age, clothing thermal resistance, metabolic rate, indoor air temperature, outdoor temperature, relative humidity, and wind speed as the input characteristic parameters of the thermal comfort model.

7. The thermal comfort regulation method based on visual perception and global sensitivity analysis according to claim 6, characterized in that, The preprocessing of the data information in the historical data set includes: For the missing historical indoor temperature information ta, historical operative temperature information top, historical radiant temperature information tr, historical outdoor temperature information tg, historical indoor humidity information rh, historical outdoor air temperature information t_out, historical outdoor humidity information rh_out, historical clothing thermal resistance information clo, historical fan status information fan, historical window status information window, historical door status information door, and historical season information season in the historical data set, fill them with the mode according to the season when filling; Process the historical gender information gender in the historical data set by one-hot encoding; Fill in the missing historical age information age, historical height information ht, historical weight information wt, historical human metabolic rate information met, and historical indoor wind speed information vel in the historical dataset with the average value.

8. The thermal comfort regulation method based on visual perception and global sensitivity analysis according to claim 1, characterized in that Adjust the operating parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort, including: Subtract the actual thermal comfort value obtained from the calculation of the human thermal comfort from the target thermal comfort value; If the difference is outside the preset range, adjust the operating parameters of the air conditioner or heat pump to make the actual thermal comfort value approach the target thermal comfort value.

9. A thermal comfort control device based on visual perception and global sensitivity analysis, characterized in that, Including: A collection module for obtaining environmental information and indoor human image information; An identification module for inputting the indoor human image information into a pre-trained attribute identification model to identify human attribute information; A first calculation module for calculating the metabolic rate information of the person in the indoor human image information based on the human attribute information; A second calculation module for inputting the environmental information, human attribute information, and metabolic rate information into a pre-trained thermal comfort model to calculate the human thermal comfort; An adjustment module for adjusting the operating parameters of the air conditioner or heat pump based on the calculation result of the human thermal comfort.

Images