Human-based sensing thermal comfort simulation method, system and equipment and storage medium

Abstract

The embodiment of the invention provides a human-based sensing thermal comfort simulation method, system and device and a storage medium, and belongs to the technical field of artificial intelligence. According to the method, ground landscape image data, satellite remote sensing image data and meteorological sequence data of a prediction area are obtained, and the ground landscape image data, the satellite remote sensing image data and the meteorological sequence data are input into a thermal comfort prediction model to obtain the thermal comfort. According to the scheme, thermal comfort analysis is carried out based on the artificial intelligence technology, the thermal comfort prediction model is fused with satellite remote sensing image data, meteorological data and landscape image data can be supplemented into large-scale space features, and the model can adapt to large-scale space comfort prediction; the thermal comfort prediction model also fuses ground landscape image data, considers the influence of surrounding environments such as buildings on comfort perception, and improves the reliability of thermal comfort analysis results.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to a human perception thermal comfort simulation method, system, device and storage medium. Background Art

[0002] The thermal environment of a city directly affects the livability of the city and the health of its residents. With the acceleration of urbanization and the intensification of climate change, extreme high-temperature weather occurs frequently, significantly reducing outdoor comfort and leading to an increasing number of heat-related diseases and deaths year by year, seriously threatening the quality of life and work efficiency of urban residents. Human perception thermal comfort refers to the actual perception of the thermal environment by humans, usually evaluated by thermal sensors or directly collecting subjective thermal perceptions of humans. Currently, through various sensors and advanced modeling techniques, efforts are being made to measure thermal comfort more accurately.

[0003] In related technologies, the method for measuring thermal comfort is a ground sensor-based measurement method, that is, using the measurement environment data of ground sensors to predict comfort. The data of ground sensors can refer to the measurement data of static weather stations, which can measure meteorological parameters at a certain point at low cost and for a long time, and continuous environmental indicators can be generated based on the measurement results of multiple weather stations using interpolation methods. The data of ground sensors can also refer to the measurement data of dynamic vehicle-mounted (such as buses, private cars) or human-mounted and bicycle-mounted devices at an approximate walking speed, which can calculate more accurate thermal comfort indicators with higher spatial resolution.

[0004] Although the ground sensor-based measurement method has advantages in terms of accuracy, its limitation is that the spatial range is limited. In particular, the area covered by weather stations is usually small, and continuous data within the city cannot be provided. Although vehicle-mounted and human-mounted sensors can improve spatial resolution, the deployment and use costs of these devices are high, and it is difficult to conduct long-term monitoring over a large area. Therefore, this method cannot achieve the prediction of human perception thermal comfort in large-scale spaces, and the human perception thermal comfort parameters only consider meteorological-related data, without considering the influence of spatial environments such as buildings on the human perception of thermal comfort, resulting in relatively low reliability of thermal comfort analysis results. Summary of the Invention

[0005] The main purpose of the embodiments of this application is to propose a human perception thermal comfort simulation method, system, device and storage medium, which can be applicable to thermal comfort analysis in large-scale spaces and improve the reliability of thermal comfort analysis results.

[0006] To achieve the above object, on the one hand, an embodiment of this application proposes a human perception thermal comfort simulation method, including:

[0007] Obtain ground landscape image data, satellite remote sensing image data, and meteorological sequence data of the prediction area;

[0008] Input the ground landscape image data, the satellite remote sensing image data, and the meteorological sequence data into a thermal comfort prediction model to obtain the thermal comfort level.

[0009] In some embodiments, the step of inputting the ground landscape image data, the satellite remote sensing image data, and the meteorological sequence data into a thermal comfort prediction model to obtain the thermal comfort level includes the following steps:

[0010] Use a convolutional neural network to extract features from the ground landscape image data and the satellite remote sensing image data respectively, and correspondingly obtain street view features and satellite remote sensing features;

[0011] Use a time recurrent neural network to extract features from the meteorological sequence data to obtain meteorological time features;

[0012] Perform feature fusion based on the street view features, the satellite remote sensing features, and the meteorological time features to obtain fusion features;

[0013] Map the fusion features to the thermal comfort level through a fully connected layer.

[0014] In some embodiments, the ground landscape image data is obtained through the following steps:

[0015] Obtain the captured image data of multiple perspectives of the prediction point, where the prediction point is located in the prediction area;

[0016] Stitch multiple pieces of the captured image data in sequence according to the perspective continuous order to obtain the ground landscape image data.

[0017] In some embodiments, the satellite remote sensing image data is obtained through the following steps:

[0018] Based on a preset regional specification, determine a prediction area with the prediction point as the center;

[0019] Crop the corresponding part of the prediction area from the satellite remote sensing image of the entire area to obtain the satellite remote sensing image data of the prediction area.

[0020] In some embodiments, the meteorological sequence data is obtained through the following steps:

[0021] Obtain the meteorological monitoring data of the prediction area at the prediction target time point and a certain period of time before the prediction target time point;

[0022] Sample the meteorological monitoring data at a preset time interval to obtain the meteorological sequence data.

[0023] In some embodiments, the step of obtaining the fusion feature by fusing the street view feature, the satellite remote sensing feature, and the meteorological time feature includes the following steps:

[0024] Stitch the street view feature, the satellite remote sensing feature, and the meteorological time feature to obtain a multi-source feature;

[0025] Input the multi-source feature into a squeeze-and-excitation network to predict the weight of each feature channel in the multi-source feature, and fuse the compressed multi-source feature according to the weight to obtain a fusion feature.

[0026] In some embodiments, the thermal comfort prediction model is trained through the following steps:

[0027] Obtain a training data set, where the training data set includes sample data of multiple regions, the sample data includes ground landscape image data, satellite remote sensing image data, meteorological sequence data, and sample labels, and the sample labels are obtained by calculating the thermal comfort index based on the actually measured temperature and humidity;

[0028] Input the training data set into the initialized thermal comfort prediction model for forward prediction to obtain a prediction result;

[0029] Determine the model loss according to the prediction result and the sample label, and update the parameters of the thermal comfort prediction model backward according to the model loss until the model loss is less than a preset value to obtain a trained thermal comfort prediction model.

[0030] To achieve the above object, another aspect of the embodiments of the present application provides a human perception thermal comfort simulation system, including:

[0031] A first module, configured to obtain ground landscape image data, satellite remote sensing image data, and meteorological sequence data of a prediction region;

[0032] A second module, configured to input the ground landscape image data, the satellite remote sensing image data, and the meteorological sequence data into a thermal comfort prediction model to obtain a thermal comfort level.

[0033] To achieve the above object, another aspect of the embodiments of the present application provides an electronic device, where the electronic device includes a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for implementing connection communication between the processor and the memory. When the program is executed by the processor, the method described in the above embodiments is implemented.

[0034] To achieve the above object, on the other hand, an embodiment of the present application provides a storage medium, which is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method described in the above embodiment.

[0035] The human perception thermal comfort simulation method, system, device and storage medium proposed by the present application obtain the ground landscape image data, satellite remote sensing image data and meteorological sequence data of the prediction area, and input the ground landscape image data, satellite remote sensing image data and meteorological sequence data into the thermal comfort prediction model to obtain the thermal comfort. The solution of the present application performs thermal comfort analysis based on artificial intelligence technology. The thermal comfort prediction model integrates satellite remote sensing image data, which can supplement meteorological data and landscape image data into large-scale spatial features, enabling the model to adapt to large-scale spatial comfort prediction; the thermal comfort prediction model also integrates ground landscape image data, considering the influence of the surrounding environment such as buildings on the comfort perception, and improving the reliability of the thermal comfort analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flowchart of the human perception thermal comfort simulation method provided by an embodiment of the present application;

[0037] Figure 2 is a visualization diagram of the sampling data of the measured data provided by an embodiment of the present application;

[0038] Figure 3 is a schematic diagram of the overall framework of the thermal comfort prediction model provided by an embodiment of the present application;

[0039] Figure 4 is a schematic diagram of the structure of the satellite remote sensing feature extraction module provided by an embodiment of the present application;

[0040] Figure 5 is a schematic diagram of the structure of the street view feature extraction module provided by an embodiment of the present application;

[0041] Figure 6 is a schematic diagram of the structure of the meteorological time feature extraction module provided by an embodiment of the present application;

[0042] Figure 7 is a schematic diagram of the structure of the fusion and prediction module provided by an embodiment of the present application;

[0043] Figure 8 is a scatter diagram of the prediction accuracy results and the optimal model under different scenarios provided by an embodiment of the present application;

[0044] Figure 9 is a schematic diagram of the simulation results of the optimal model of different thermal comfort indicators in a certain area provided by an embodiment of the present application;

[0045] Figure 10 It is a schematic diagram of the hardware structure of the electronic device provided by the embodiment of the present application. Specific embodiments

[0046] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0047] It should be noted that although the functional modules are divided in the system and the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order from the module division in the system or the flowchart. Terms such as "first" and "second" in the description, claims and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0049] Based on this, the embodiments of the present application provide a human perception thermal comfort simulation method, system, device and storage medium, which can be applied to thermal comfort analysis in large-scale spaces and improve the reliability of thermal comfort analysis results.

[0050] The human perception thermal comfort simulation method provided by the embodiments of the present application relates to the field of artificial intelligence technology. The human perception thermal comfort simulation method provided by the embodiments of the present application can be applied to a terminal, a server, or software running on a terminal or a server. In some embodiments, the terminal may be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server may be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software may be an application that implements the human perception thermal comfort simulation method, etc., but is not limited to the above forms.

[0051] This application can be used in numerous general - purpose or special - purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi - processor systems, microprocessor - based systems, set - top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. This application can be described in the general context of computer - executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0052] Figure 1 is an alternative flowchart of the human - perception thermal comfort simulation method provided by an embodiment of this application. Figure 1 The method in may include but is not limited to steps S101 to S102.

[0053] Step S101: Obtain the ground landscape image data, satellite remote - sensing image data, and meteorological sequence data of the prediction area;

[0054] Step S102: Input the ground landscape image data, satellite remote - sensing image data, and meteorological sequence data into the thermal comfort prediction model to obtain the thermal comfort level.

[0055] Steps S101 to S102 illustrated in the embodiment of this application obtain the ground landscape image data, satellite remote - sensing image data, and meteorological sequence data of the prediction area, and input the ground landscape image data, satellite remote - sensing image data, and meteorological sequence data into the thermal comfort prediction model to obtain the thermal comfort level. The solution of this application conducts thermal comfort analysis based on artificial intelligence technology. The thermal comfort prediction model integrates satellite remote - sensing image data, which can supplement meteorological data and landscape image data into large - scale spatial features, enabling the model to adapt to large - scale spatial comfort prediction; the thermal comfort prediction model also integrates ground landscape image data, considering the impact of the surrounding environment such as buildings on comfort perception, and improving the reliability of the thermal comfort analysis results.

[0056] In step S101 of some embodiments, the ground landscape image data refers to the image data captured at the thermal comfort analysis points in the prediction area. The ground landscape image data contains the environmental status content of the analysis points. For example, the building distribution situation, etc. The ground landscape image data can be spliced from the captured image data, or can be spliced from the street view image data provided on the map service software. In the thermal comfort prediction model training stage, the street view image data provided on the map service software can be directly used to reduce the training cost. The ground landscape image data of each point contains a total of four RGB images in the front, right, back, and left directions, which can truly reflect the urban street view environment.

[0057] Satellite remote sensing image data refers to the ground data captured by satellites over the ground. The remote sensing image used in this embodiment can be a 0.4m resolution RGB image from Google, which is suitable for urban-scale environmental analysis and can provide accurate ground information. The satellite remote sensing image data covers a wide area and provides multi-dimensional urban environmental information in a large-scale space.

[0058] Meteorological sequence data refers to meteorological data such as temperature and humidity detected at the ground meteorological station. The meteorological station updates a piece of meteorological data every 5 minutes. The meteorological sequence data can provide spatio-temporal climate change information and can help the model to predict thermal comfort under different times and climate conditions.

[0059] In some embodiments, the ground landscape image data in step S101 can be obtained through, but not limited to, the following steps:

[0060] Step S201, obtain the captured image data of multiple perspectives of the prediction point, where the prediction point is in the prediction area;

[0061] Step S202, sequentially splice multiple captured image data in the order of the perspectives to obtain the ground landscape image data.

[0062] In this embodiment, first, the prediction points that need to be analyzed for thermal comfort are determined. Then, the captured image data from multiple perspectives at these points are obtained. Next, multiple captured image data are sequentially stitched in the order of the perspectives to obtain the ground landscape image data containing the environmental information around the prediction points. Exemplarily, the ground landscape image data can be directly obtained on a map software. The ground landscape image data consists of four street view pictures in four directions of the prediction point, recording the 360° visual information of the shooting point. The visual information in adjacent directions is continuous. To enable the model to correctly capture the continuous visual information, the four street view images need to be stitched according to the street view direction and width. The image dimension changes from (4, 3, 224, 224) to (3, 224, 896). Further, to make the stitched ground landscape image data meet the model input requirements, the image size can be adjusted. For example, the ground landscape image can be adjusted to the standard size of 224×224.

[0063] In some embodiments, the satellite remote sensing image data in step S101 can be obtained through, but not limited to, the following steps:

[0064] Step S301, based on a preset regional specification, determine a prediction area centered on the prediction point;

[0065] Step S302, crop the part corresponding to the prediction area from the satellite remote sensing image of the entire area to obtain the satellite remote sensing image data of the prediction area.

[0066] In this embodiment, first, the prediction points that need to be analyzed for thermal comfort are determined, a prediction area of a certain size is determined centered on the prediction point, and then the part corresponding to the prediction area is cropped from the satellite remote sensing image of the entire area to obtain the satellite remote sensing image data of the prediction area. Exemplarily, centered on the street view point, a grid vector of 100m×100m is constructed. According to this grid vector, the corresponding area is cropped from the Google remote sensing image, so as to obtain the corresponding remote sensing image for each street view point. Each grid area provides accurate remote sensing data input for the model.

[0067] Further, since there may be significant differences in the shooting conditions of remote sensing images in different regions, the brightness difference of the pictures will have a greater impact on the results. However, what the model learns from the remote sensing images should be the physical information of the city. Therefore, histogram equalization processing is applied to the remote sensing images. The specific operation is to divide the image into three RGB channels, and respectively apply the "equalizeHist" function in the cv2 library for histogram standardization. Finally, the channels are merged and the picture is returned.

[0068] Further, to make the satellite remote sensing image data meet the model input requirements, the image size can be adjusted. For example, the satellite remote sensing image can be adjusted to the standard size of 224×224.

[0069] In some embodiments, the meteorological sequence data in step S101 can be obtained through, but not limited to, the following steps:

[0070] Step S401, obtain the meteorological monitoring data of the prediction area at the predicted target time point and a certain period of time before the predicted target time point;

[0071] Step S402, sample the meteorological monitoring data at a preset time interval to obtain meteorological sequence data.

[0072] In this embodiment, the predicted target time point can be determined in the following way. The thermal comfort can be measured by the temperature, humidity and wet-bulb globe temperature index (WBGT) measured on-site. Figure 2 It is a visualization graph of the sampled data of the measured data (temperature, humidity and WBGT). Data sampling was carried out in the afternoon of August 30, 2023 and August 31, 2023. The afternoon sampling was mainly concentrated in the hottest and most humid period of the day. The temperature values of the two-day sampling were around 32°C, and the difference between the two days was within 1°C; the humidity values were around 58%-62%, and the humidity on the 31st was slightly higher by 1%; the WBGT values were around 27°C-29°C, and the WBGT on the 31st was slightly higher by 0.5°C; generally speaking, the climate change degree during the two-day sampling time was small, and the data consistency was high. Among them, 15:00 in the afternoon was the sampling coincidence time for the two days, so 15:00 can be used as the prediction reference time (i.e., the predicted target time point) in the afternoon.

[0073] In this embodiment, according to the prediction date and the predicted target time point (such as 15:00 in the afternoon), extract the meteorological data corresponding to the time point from the meteorological data of the meteorological station, and obtain the meteorological monitoring data of the prediction area for a certain period of time (such as 24 hours) before this time point, and then sample the meteorological monitoring data every 5 minutes to obtain meteorological time series data, so that the model can understand the past meteorological changes and provide time series information for the prediction of thermal comfort.

[0074] In step S102 of some embodiments, by combining the advantages of different data sources, the thermal comfort prediction model can comprehensively extract spatial features related to thermal comfort assessment (remote sensing), ground perspective features (street view), and meteorological sequence features (meteorology), overcoming the deficiency that a single data source cannot comprehensively characterize the thermal environment, and realizing large-scale and high-precision human-centered perception thermal comfort assessment in a low-cost manner. By inputting street view images and remote sensing images to construct a fixed urban physical environment, and then adding continuous meteorological sequence data to set the overall meteorological change value range, finally, human-centered perception thermal comfort parameters (such as WBGT, temperature, and humidity) are simulated. In this embodiment, the thermal comfort level output by the prediction model can be index data such as temperature, humidity, dry bulb humidity, or WBGT of the prediction point. These outputs can be used to measure the thermal comfort level. Which specific index data to choose to measure the thermal comfort level can be set according to actual assessment requirements.

[0075] In some embodiments, step S102 may further include, but is not limited to, steps S501 to S504:

[0076] Step S501, using a convolutional neural network to respectively extract features from ground landscape image data and satellite remote sensing image data, and correspondingly obtaining street view features and satellite remote sensing features;

[0077] Step S502, using a time recurrent neural network to extract features from meteorological sequence data to obtain meteorological time features;

[0078] Step S503, performing feature fusion according to street view features, satellite remote sensing features, and meteorological time features to obtain fusion features;

[0079] Step S504, mapping the fusion features to the thermal comfort level through a fully connected layer.

[0080] In this embodiment, please refer to Figure 3 , Figure 3Schematic diagram of the overall framework of the thermal comfort prediction model. This model combines remote sensing images, street view images, and meteorological data to construct a multi-modal architecture, and uses multi-modal deep learning methods for thermal comfort prediction. The ground landscape image data, satellite remote sensing image data, and meteorological sequence data are combined to predict the WBGT index. Through in-depth mining of multi-source data, the spatial and temporal characteristics of the thermal environment are fully extracted. The ground landscape image data is stitched to form continuous images. In the model, a convolutional neural network (such as ResNet18) is used to extract the spatial features of the image and the feature dimensionality reduction is performed through a multi-layer perceptron (MLP). The satellite remote sensing image data is enhanced in detail by landscape histogram equalization. In the model, a convolutional neural network is also used to extract the spatial features of the image and the feature dimensionality reduction is performed through a multi-layer perceptron. The meteorological sequence data is processed by a time recurrent neural network (such as LSTM) in the model to extract temporal features. The SENet module is used in the model to dynamically adjust the contribution ratio of the features extracted by each network module, and the features are weighted and integrated to obtain fused features, and then the predicted thermal comfort is obtained through mapping by a fully connected layer (FC).

[0081] In some embodiments, please refer to Figure 4 , Figure 4 For a detailed introduction to the satellite remote sensing feature extraction module, this module extracts the features of satellite remote sensing image data through the ResNet18 network, and then reduces the dimensionality of the data through a multi-layer perceptron (MLP). ResNet18 is a classic convolutional neural network (CNN) architecture suitable for deep learning tasks of image data. By removing the last fully connected layer of ResNet18, a low-dimensional feature representation of the image is extracted. Then, the image features are reduced to 64 dimensions through the MLP for subsequent processing. In addition, to prevent overfitting, the Dropout technique is used in this module.

[0082] The processing framework of the satellite remote sensing feature extraction module based on the ResNet18 framework is: ResNet18→MLP[(512,256),ReLU,(256,128),ReLU,(128,64)]→Dropout(0.5). The descriptions of each part in the framework are as follows:

[0083] ResNet18: Remove the last fully connected layer in the ResNet18 model, and retain the original convolutional layer and pooling layer to extract image features.

[0084] Multi-layer perceptron layer (MLP): Reduce the output features (512 dimensions) of ResNet18 to 64 dimensions.

[0085] ReLU activation function: Add a ReLU activation function after the fully connected layer to increase the non - linear expression ability of the network and help the model learn more complex patterns.

[0086] Dropout: Apply 50% Dropout to avoid overfitting during model training and improve the generalization ability of the model.

[0087] In some embodiments, please refer to Figure 5 , Figure 5 For a detailed introduction of the street view feature extraction module, the input of this module is the image after multi - perspective landscape stitching. The ground landscape image data is feature - extracted through the ResNet18 network, and then the data is dimension - reduced through a multi - layer perceptron (MLP). ResNet18 is a classic convolutional neural network (CNN) architecture suitable for deep - learning tasks of image data. By removing the last fully connected layer of ResNet18, a low - dimensional feature representation of the image is extracted. Then, the image features are dimension - reduced to 64 dimensions through the MLP for subsequent processing. In addition, to prevent overfitting, the Dropout technique is used in this module.

[0088] The processing framework of the street view feature extraction module based on the ResNet18 framework is: ResNet18→MLP[(512,256),ReLU,(256,128),ReLU,(128,64)]→Dropout(0.5). The explanations of each part in the framework are as follows:

[0089] ResNet18: Remove the last fully connected layer in the ResNet18 model and retain the original convolutional layer and pooling layer to extract image features.

[0090] Multi - layer perceptron layer (MLP): Dimension - reduce the output features (512 dimensions) of ResNet18 to 64 dimensions.

[0091] ReLU activation function: Add a ReLU activation function after the fully connected layer to increase the non - linear expression ability of the network and help the model learn more complex patterns.

[0092] Dropout: Apply 50% Dropout to avoid overfitting during model training and improve the generalization ability of the model.

[0093] In some embodiments, please refer to Figure 6 , Figure 6For the detailed introduction of the meteorological time feature extraction module, the input meteorological data includes temperature and humidity, which has obvious time series characteristics. Therefore, in this embodiment, the LSTM network is used to extract the features of time series data. A meteorological parameter is extracted every 5 minutes, and a total of 24 hours of data before the target time point is obtained, a total of 2×288 pieces. The LSTM network can capture the time-dependent relationship in the meteorological data, so as to provide more accurate meteorological information for the prediction of thermal comfort. The output of the LSTM is a 64-dimensional vector, which can be directly input into the fusion module.

[0094] The processing framework of the meteorological time feature extraction module based on the LSTM model is: LSTM(2,64). The LSTM extracts the time series features of the meteorological data. Its input is 2-dimensional data (temperature and humidity). The hidden layer size of the LSTM is 64, and 64-dimensional features are output.

[0095] In some embodiments, step S503 includes but is not limited to steps S601 to S602:

[0096] Step S601, splice the street view features, satellite remote sensing features and meteorological time features to obtain multi-source features;

[0097] Step S602, input the multi-source features into the squeeze-and-excitation network to predict the weights of each feature channel in the multi-source features, and fuse the compressed multi-source features according to the weights to obtain the fused features.

[0098] In this embodiment, the squeeze-and-excitation network (SENet) can be used for multi-feature fusion. Please refer to Figure 7 , Figure 7 For the detailed introduction of the fusion and prediction module, this module aims to fuse the feature vectors of multi-modal data (including street view features, satellite remote sensing features and meteorological time features), and optimize different features by introducing the attention mechanism of SENet to highlight the key features. Finally, the fused features are reduced in dimension through the fully connected layer, and the output is the result of the thermal comfort prediction task (such as WBGT, temperature or humidity).

[0099] The processing framework of the fusion and prediction module based on the SENet model is: SENet(192,192)→FC(192,1). The description of each part in the framework is as follows:

[0100] SENet: SENet is used to learn the weights of each feature channel to enhance the features useful for the task and suppress irrelevant or redundant features.

[0101] Final fully connected layer (FC): Map the 192-dimensional features to 1 dimension for outputting the final prediction result.

[0102] In some embodiments, the thermal comfort prediction model in step S102 can be trained through but not limited to the following steps:

[0103] Step S701: Obtain a training data set, where the training data set includes sample data of multiple regions. The sample data includes ground landscape image data, satellite remote sensing image data, meteorological sequence data, and sample labels. The sample labels are obtained by calculating the thermal comfort index based on the actually measured temperature and humidity;

[0104] Step S702: Input the training data set into the initialized thermal comfort prediction model for forward prediction to obtain a prediction result;

[0105] Step S703: Determine the model loss based on the prediction result and the sample labels, and update the parameters of the thermal comfort prediction model backward according to the model loss until the model loss is less than a preset value to obtain a trained thermal comfort prediction model.

[0106] In this embodiment, the training data set includes sample data of multiple regions (i.e., the distribution of sampling points). The sample data includes ground landscape image data, satellite remote sensing image data, meteorological sequence data, and sample labels. Among them, the ground landscape image data, satellite remote sensing image data, and meteorological sequence data are processed in the above manner and will not be elaborated here. The sample labels are obtained by calculating the thermal comfort index based on the actually measured temperature and humidity. The thermal comfort index can be the wet bulb temperature and WBGT, that is, the model's predicted thermal comfort can be measured by the wet bulb temperature and WBGT. Specifically, data such as the dry bulb temperature, relative humidity, and black globe temperature can be collected through a meteorological device installed on a bicycle, and these data are used to calculate the wet bulb temperature and WBGT (wet bulb globe temperature index). The wet bulb temperature is calculated by the following formula:

[0107]

[0108] where ATAN is the arctangent function, ATAN(number) returns the arctangent value of number, and the arctangent value is the angle whose tangent value is number. The values of 0.151977, 8.313659, 1.676331, 0.00391838, and 0.023101 in the formula are fixed coefficients, all obtained by inverting the most accurate formula for Bolton's equivalent potential temperature θE to obtain the adiabatic wet bulb temperature Tw at a given pressure through an iterative technique for a given pseudo-adiabatic measurement.

[0109] WBGT is calculated by the following formula:

[0110] WBGT = 0.7·T w + 0.2·T a+0.1·T b ;

[0111] where T a is the dry-bulb temperature, T W is the wet-bulb temperature, T b is the globe temperature, RH is the relative humidity, WBGT is the wet-bulb globe temperature index, and 0.7, 0.2, and 0.1 are all weight values that can be obtained through statistical analysis.

[0112] During the model training process, the following details are used for optimization and training:

[0113] Dataset construction: A dataset is constructed based on the actual sampled data. A unique data point is determined according to the spatial location, and the corresponding input data (street view + remote sensing + meteorology) is obtained; there are 3,114 data points in the deduplicated dataset, including 1,760 data points in the afternoon and 1,354 data points at night. The dataset is divided into a training set and a test set according to a ratio of 7:3.

[0114] Batch size: 32 is used as the batch size for training. This batch size can be used for training in a relatively short time while avoiding out-of-memory errors in the video memory.

[0115] Optimizer: The AdamW optimizer is used. AdamW is an improved version of the Adam optimizer, which is suitable for large-scale datasets in deep learning and can effectively handle regularization problems. This optimizer can update the learning rate according to the gradient of each parameter, improving the training efficiency and model accuracy. The learning rate of the AdamW optimizer is set to 1e-3, and the weight decay is set to 1e-4.

[0116] Number of training times and epochs: Set num_epochs = 70, that is, 70 epochs of training are performed. At the same time, num_trials = 3 training sessions are carried out to select the best model. Each time during training, the weight parameters of the model are reset to increase the generalization ability of the model.

[0117] Loss function: The mean absolute error (MAE) is used as the loss function, and the specific definition is as follows: loss = criterion(outputs.squeeze(), WBGT)

[0118] The definition of MAE is as follows:

[0119]

[0120] y i is the true value (target value); is the predicted value; m is the number of samples.

[0121] Among them, outputs are the predicted values of the model, and WBGT is the actual WBGT data (which can be switched to the corresponding parameter according to the prediction target).

[0122] To examine the contributions of different modules to the overall model, in this embodiment, the outputs of different modules are disabled to test their impacts on the prediction accuracy. Figure 8 It is a scatter plot of the prediction accuracy results for different scenarios and the optimal model. Since meteorological parameters are the core variables for predicting thermal comfort, the meteorological module is used as the base module, and other modules are gradually disabled to examine their contributions. A total of four scenarios are set. Among them, Scenario 1 is to use only the meteorological module; Scenario 2 is to use only the meteorological module and the street view module; Scenario 3 is to use only the meteorological module and the remote sensing module; Scenario 4 is to use the meteorological module, the street view module, and the remote sensing module.

[0123] In this embodiment, R2 and MAPE are used to examine the accuracy results of different measurement data. Through Figure 8 the results, it can be seen that for all environmental indicators, the prediction accuracy gradually increases from Scenario 1 to Scenario 4. The contribution of only using meteorological parameters to the model prediction is relatively low; the contribution of the remote sensing module is greater than that of the street view module; using all modules can achieve the optimal accuracy; all modules contribute to the model. The optimal R2 for the temperature indicator (T) is 0.7312, and the MAPE is 0.7174%; the optimal R2 for the humidity indicator (RH) is 0.7457, and the MAPE is 0.8344%; the optimal R2 for the WBGT indicator is 0.8228, and the MAPE is 0.8390%; for each environmental indicator, the model can generally achieve a relatively high prediction performance.

[0124] Figure 9 It is the simulation results of the optimal model for different thermal comfort indicators in a certain area. The model is only simulated at street view points. By cropping the street view point remote sensing images and obtaining meteorological data to construct a dataset, the prediction results are finally averaged and displayed within a 100m grid. The prediction results show that all three thermal comfort indicators exhibit good spatial differentiation. The high values of WBGT are mainly concentrated in the left area; the high values of temperature (T) are mainly concentrated in the upper left area; the high values of humidity (RH) are mainly concentrated in the lower left area.

[0125] According to some embodiments of the present application, the embodiments of the present application have the following beneficial effects:

[0126] First, by fusing multi-modal data, a framework for accurately evaluating human-centered perception of thermal comfort at low cost and on a large scale is proposed. Considering the influencing factors of human-centered perception of thermal sensation, fine human-centered perception of hot and humid indices (such as WBGT and dry bulb humidity, etc.) are constructed by combining the temperature and humidity data collected on site.

[0127] Second, based on the characteristics of different modal data, a deep learning model architecture combining ResNet18-LSTM-SENet is adopted to extract the deep features of various types of data, improving the prediction ability of the human-based perceived thermal comfort index.

[0128] Third, referring to the theoretical framework of urban heat, street view and remote sensing images are used to construct the urban geometric structure, and the spatial heterogeneity of human-based perceived thermal comfort is predicted by inputting the meteorological parameters of the target time.

[0129] The embodiment of the present application also provides a human-based perceived thermal comfort simulation system, including:

[0130] The first module is used to obtain the ground landscape image data, satellite remote sensing image data and meteorological sequence data of the prediction area;

[0131] The second module is used to input the ground landscape image data, satellite remote sensing image data and meteorological sequence data into the thermal comfort prediction model to obtain the thermal comfort.

[0132] It can be understood that the content in the above embodiments of the human-based perceived thermal comfort simulation method is applicable to the embodiments of this system. The functions specifically implemented by the embodiments of this system are the same as those of the above embodiments of the human-based perceived thermal comfort simulation method, and the beneficial effects achieved are also the same as those of the above embodiments of the human-based perceived thermal comfort simulation method.

[0133] The embodiment of the present application also provides an electronic device. The electronic device includes: a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing the connection and communication between the processor and the memory. When the program is executed by the processor, it implements the above-mentioned human-based perceived thermal comfort simulation method. The electronic device can be any intelligent terminal including a tablet computer, an in-vehicle computer, etc.

[0134] Please refer to Figure 10 , Figure 10 which shows the hardware structure of an electronic device in another embodiment. The electronic device includes:

[0135] The processor 1001 can be implemented by using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided by the embodiments of the present application;

[0136] The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), etc. The memory 1002 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1002 and are called by the processor 1001 to execute the human perception thermal comfort simulation method in the embodiments of this application;

[0137] The input / output interface 1003 is used to implement information input and output;

[0138] The communication interface 1004 is used to implement communication and interaction between this device and other devices. It can communicate through wired means (such as USB, network cable, etc.) or through wireless means (such as mobile network, WIFI, Bluetooth, etc.);

[0139] The bus 1005 transmits information between the various components of the device (such as the processor 1001, the memory 1002, the input / output interface 1003, and the communication interface 1004);

[0140] Among them, the processor 1001, the memory 1002, the input / output interface 1003, and the communication interface 1004 are communicatively connected to each other inside the device through the bus 1005.

[0141] The embodiments of this application also provide a storage medium. The storage medium is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned human perception thermal comfort simulation method.

[0142] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes a memory remotely located relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0143] The embodiments described in the embodiments of the present application are to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

[0144] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than those shown, or combine certain steps, or different steps.

[0145] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0146] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations.

[0147] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0148] It should be understood that in this application, "at least one (item)" means one or more, and "a plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. "At least one (one)" or similar expressions below refer to any combination of these items, including any combination of single item (one) or plural items (ones). For example, at least one (one) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0149] In several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative. For example, the above division of units is only a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of systems or units can be in electrical, mechanical or other forms.

[0150] The units described above as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0151] In addition, the functional units in each embodiment of this application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0152] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of this application. The foregoing storage medium includes: various media that can store programs such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0153] The preferred embodiments of the embodiments of this application have been described above with reference to the accompanying drawings, and thus do not limit the scope of the rights of the embodiments of this application. Any modifications, equivalent replacements, and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of this application shall be within the scope of the rights of the embodiments of this application.

Claims

1. A method for simulating human-perceived thermal comfort, characterized in that: The following steps are involved: Obtain ground landscape image data, satellite remote sensing image data and meteorological series data of the forecast area; The ground landscape image data, the satellite remote sensing image data and the meteorological sequence data are input into a thermal comfort prediction model to obtain thermal comfort.

2. The method for simulating human-perceived thermal comfort according to claim 1, characterized in that: The step of inputting the ground landscape image data, the satellite remote sensing image data and the meteorological sequence data into a thermal comfort prediction model to obtain thermal comfort comprises the following steps: Using a convolutional neural network to extract features from the ground landscape image data and the satellite remote sensing image data, respectively, to obtain street view features and satellite remote sensing features; Using a time recurrent neural network to extract features from the meteorological sequence data to obtain meteorological time features; Performing feature fusion according to the street view feature, the satellite remote sensing feature and the meteorological time feature to obtain a fusion feature; The fused features are mapped to thermal comfort through a fully connected layer.

3. The method for simulating human-perceived thermal comfort according to claim 1, characterized in that: The ground landscape image data is obtained by the following steps: Acquire captured image data of a plurality of viewing angles of a predicted point, wherein the predicted point is in the predicted area; The plurality of captured image data are sequentially spliced ​​in a continuous order of viewing angles to obtain the ground landscape image data.

4. The method for simulating human-perceived thermal comfort according to claim 3, characterized in that: The satellite remote sensing image data is obtained by the following steps: Based on the preset area specifications, determining the prediction area with the prediction point as the center; The portion corresponding to the predicted area is cut out from the satellite remote sensing image of the entire area to obtain satellite remote sensing image data of the predicted area.

5. The method for simulating human-perceived thermal comfort according to claim 1, characterized in that: The meteorological sequence data is obtained by the following steps: Obtaining meteorological monitoring data of the forecast area at a forecast target time point and a certain period of time before the forecast target time point; The meteorological monitoring data is sampled at preset time intervals to obtain meteorological sequence data.

6. The method for simulating human-perceived thermal comfort according to claim 2, characterized in that: The step of fusing the street view feature, the satellite remote sensing feature and the meteorological time feature to obtain the fused feature comprises the following steps: splicing the street view features, the satellite remote sensing features and the meteorological time features to obtain multi-source features; The multi-source features are input into a compression and excitation network to predict the weight of each feature channel in the multi-source features, and the compressed multi-source features are fused according to the weights to obtain fused features.

7. The method for simulating human-perceived thermal comfort according to claim 1, characterized in that: The thermal comfort prediction model is trained by the following steps: Acquire a training data set, wherein the training data set includes sample data of multiple regions, the sample data includes ground landscape image data, satellite remote sensing image data, meteorological sequence data and sample labels, and the sample labels are obtained by calculating thermal comfort indexes through actually measured temperature and humidity; Inputting the training data set into the initialized thermal comfort prediction model for forward prediction to obtain a prediction result; The model loss is determined according to the prediction result and the sample label, and the parameters of the thermal comfort prediction model are reversely updated according to the model loss until the model loss is less than a preset value, thereby obtaining a trained thermal comfort prediction model.

8. A human-perceived thermal comfort simulation system, characterized in that: include: The first module is used to obtain ground landscape image data, satellite remote sensing image data and meteorological sequence data of the prediction area; The second module is used to input the ground landscape image data, the satellite remote sensing image data and the meteorological sequence data into a thermal comfort prediction model to obtain thermal comfort.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the steps of the method described in any one of claims 1 to 7 are realized.

10. A storage medium, the storage medium being a computer-readable storage medium, used for computer-readable storage, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of claims 1 to 7.

Images