A method of estimating artificial heat
By subdividing and modeling heat sources such as buildings, transportation, industry, and metabolism, and using a stacking integrated model, the problem of insufficient accuracy in human-made heat prediction is solved, and more accurate human-made heat estimation is achieved, which is suitable for urban thermal environment research.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- AEROSPACE INFORMATION RES INST CAS
- Filing Date
- 2023-12-06
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies ignore the characteristic differences of anthropogenic heat from different sources in anthropogenic heat estimation, resulting in insufficient prediction accuracy, especially in the application of multi-source data where efficiency and accuracy are not high.
By modeling heat sources such as buildings, transportation, industry, and metabolism separately, and using remote sensing data, meteorological data, and geographic information data, a Stacking integrated model is constructed. Combined with population heat maps and industrial heat change coefficients, the hourly heat flux of each heat source is calculated, thereby improving prediction accuracy and spatiotemporal resolution.
It enables more accurate estimation of the spatiotemporal characteristics of anthropogenic heat, improves the accuracy and resolution of anthropogenic heat prediction, and is applicable to the spatiotemporal characterization of anthropogenic heat over a wide range.
Smart Images
Figure CN117610427B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of urban environment and resource research, and in particular to an anthropogenic heat estimation method. Background Technology
[0002] Rapid urbanization worldwide over the past few decades has resulted in a significant concentration of population and economic activity. Currently, urbanized areas encompass more than half of the global population, approximately 70% of economic activity and energy consumption, and the accompanying substantial anthropogenic heat emissions have exacerbated environmental and demographic problems. Although its share in the global energy system is negligible, the impact of anthropogenic heat in major urban built-up areas is undeniable, almost equivalent to the average daily solar radiation, and its intensity is rapidly increasing along with the continuous growth of global energy consumption. Therefore, anthropogenic heat emissions have become an important and unavoidable component of urban surface energy balance, significantly affecting local urban climate and exacerbating the urban heat island effect. Given the importance of anthropogenic heat in climate simulation, heat-driven mechanisms, ecological environment assessment, and sustainable development research, clarifying its spatiotemporal patterns is of both theoretical and practical significance.
[0003] Energy consumption inventory methods are the most widely used anthropogenic heat estimation method. They typically assume that all anthropogenic heat emissions from energy consumption are dissipated as sensible heat without hysteresis. Based on scale, they can be divided into top-down and bottom-up methods. Compared to bottom-up methods, which heavily rely on detailed multi-source geographic information data, statistical data, and heating / cooling load parameters, top-down inventory methods based on large-scale energy consumption data have greater applicability and can be applied to global scales or regions with limited data quantity and quality, but the results are relatively coarser. Integrating multiple methods for anthropogenic heat estimation has become an important research direction to address the challenges of anthropogenic heat estimation in complex scenarios. However, further clarification of the relationships and differences between different methods is needed to achieve more scientific multi-method integration.
[0004] In recent years, due to the strong feasibility of top-down inventory methods, an increasing number of new and improved methods have been proposed based on them. Downscaling methods utilizing the positive relationship between nighttime light and human activity intensity to obtain anthropogenic heat have been widely used. Although they can obtain large-scale anthropogenic heat in a relatively convenient way, they cannot characterize the complex spatiotemporal features of multi-source anthropogenic heat. Thanks to the development of communication and network technologies, location semantic information, spatial interaction information, and real-time dynamic information have been applied to some energy consumption inventory methods; however, their high data requirements and cumbersome workflows limit them to small-scale studies.
[0005] The involvement of machine learning can greatly simplify the application of multi-source data and effectively improve the efficiency and accuracy of anthropogenic heat estimation, and has gradually become a hot topic in the field of urban thermal environment research. However, current research mostly focuses on overall modeling of anthropogenic heat, ignoring the characteristic differences of anthropogenic heat from different sources, resulting in insufficient prediction accuracy. Summary of the Invention
[0006] In view of this, the main objective of the present invention is to provide a method for estimating anthropogenic heat, which, by modeling anthropogenic heat from different sources, can better reflect the spatiotemporal characteristics of anthropogenic heat from different sources, thereby improving the prediction accuracy and spatiotemporal resolution of anthropogenic heat.
[0007] To achieve the above objectives, this application provides a method for artificial heat estimation, comprising:
[0008] Preprocess remote sensing data, meteorological data, building grid tiles, road network data, and POI data for the target city and target time period to obtain public feature data, building height data, weighted road density data, and weighted factory density data;
[0009] Input public characteristic data and building height data into the building heat prediction model to output monthly building heat flux; input public characteristic data and weighted road density data into the traffic heat prediction model to output monthly traffic heat flux; input public characteristic data and weighted factory density data into the industrial heat prediction model to output monthly industrial heat flux; input public characteristic data and building height data into the metabolic heat prediction model to output monthly metabolic heat flux.
[0010] Based on the relative values of the daily population heat map of the target city, the industrial heat change coefficient, and the intensity of human activity and dormant metabolism, the hourly heat flux corresponding to the monthly heat flux of each heat source is calculated.
[0011] The calculated hourly heat fluxes are summed to obtain the hourly anthropogenic heat flux of the target city within the target time period.
[0012] The public features include: nighttime lights, day and night surface temperature, vegetation index, digital elevation model, air temperature, humidity, and wind speed.
[0013] In one possible implementation, it also includes:
[0014] A first sample set is constructed using the city-level monthly building heat flux as the sample label and public features and building height as the sample features; a second sample set is constructed using the city-level monthly traffic heat flux as the sample label and public features and weighted road density as the sample features; a third sample set is constructed using the city-level monthly industrial heat flux as the sample label and public features and weighted factory density as the sample features; and a fourth sample set is constructed using the city-level monthly metabolic heat flux as the sample label and public features and building height as the sample features.
[0015] Build a stacking integration model;
[0016] The Stacking ensemble model is trained using the first sample set to obtain the building heat prediction model; the Stacking ensemble model is trained using the second sample set to obtain the traffic heat prediction model; the Stacking ensemble model is trained using the third sample set to obtain the industrial heat prediction model; and the Stacking ensemble model is trained using the fourth sample set to obtain the metabolic heat prediction model.
[0017] In another possible implementation, the construction of the Stacking integration model includes:
[0018] The Stacking ensemble model is constructed using extreme gradient boosting, random forest, Cubist, and support vector machine as base models and multiple linear regression as the meta-model.
[0019] In another possible implementation, the formula for obtaining the weighted road density data is expressed as:
[0020]
[0021]
[0022] Among them, w l V represents the weight of the level l road; l The design traffic volume is denoted by m; m is the road level number; s is the search radius; Density s L is the weighted road density within the search radius; n is the number of roads within the search radius; r A is the length of road r within the search radius; s This represents the area within the search radius.
[0023] In another possible implementation, the formula for obtaining the weighted plant density data is expressed as:
[0024]
[0025]
[0026] Among them, w c The weight of factory type c; E c Energy consumption for factory type c; m is the number of factory types; s is the search radius; Density s A is the weighted factory density within the search radius; s denoted as , where is the area within the search radius; and 'n' is the number of factories within the search radius.
[0027] In another possible implementation, the hourly anthropogenic heat flux of the target city is expressed as:
[0028]
[0029] in, and The monthly building heat flux, monthly transportation heat flux, and monthly industrial heat flux of the target city in month m are, in order. The metabolic heat flux per hour (h) is calculated from the monthly metabolic heat flux of the target city in month m based on the activity and dormant metabolic intensities. The relative values of the population heatmap for the target city; The daily industrial heat variation coefficient of the target city. Attached Figure Description
[0030] Figure 1 This is a flowchart illustrating a human-based thermal estimation method according to an embodiment of the present invention;
[0031] Figure 2 A schematic diagram showing the monthly building heat flux, monthly transportation heat flux, and monthly industrial heat flux in some cities;
[0032] Figure 3 A schematic diagram of hourly anthropogenic heat flux in Beijing in April;
[0033] Figure 4 This is a schematic diagram of the hourly anthropogenic heat flux in Shanghai in April. Detailed Implementation
[0034] Specifically, the process of an artificial heat estimation method according to an embodiment of the present invention is as follows: Figure 1 As shown, steps 101 to 104 are included.
[0035] Step 101: Preprocess remote sensing data, meteorological data, building grid tiles, road network data, and point of interest (POI) data for the target city within the target time period to obtain common feature data, building height data, weighted road density data, and weighted factory density data.
[0036] Step 102: Input the public feature data and building height data into the building heat prediction model and output the monthly building heat flux; input the public feature data and weighted road density data into the traffic heat prediction model and output the monthly traffic heat flux; input the public feature data and weighted factory density data into the industrial heat prediction model and output the monthly industrial heat flux; input the public feature data and building height data into the metabolic heat prediction model and output the monthly metabolic heat flux.
[0037] Step 103: Based on the relative values of the daily population heat map of the target city, the industrial heat change coefficient, and the intensity of human activity and dormant metabolism, calculate the hourly heat flux corresponding to the monthly heat flux of each heat source.
[0038] Step 104: Sum the calculated hourly heat fluxes to obtain the hourly anthropogenic heat flux of the target city within the target time period.
[0039] in,
[0040] In step 101, the common features include: nighttime lights, day and night surface temperature, vegetation index, digital elevation model, air temperature, humidity, and wind speed.
[0041] Remote sensing data for the target city within the target time period were preprocessed to obtain grid data for nighttime light, diurnal surface temperature, vegetation index, and digital elevation model. Meteorological data for the target city within the target time period were also preprocessed to obtain grid data for temperature, humidity, and wind speed. All grid data were processed to a 500-meter resolution.
[0042] Here, the remote sensing data may originate from MODIS, Landsat, ASTER, NPP-VIIRS, DMSP-OLS, Luojia-1, etc.
[0043] The building height data for the target city within the target time period is preprocessed. Here, the valid building grid tiles for the target city within the target time period are selected from the building grid tiles nationwide, and the grid average method is used to resample them to a resolution of 500 meters and then stitch them together to obtain the building height data.
[0044] There is a strong correlation between building height and building heat emissions. Incorporating building height data will greatly improve the accuracy of building thermal space characteristics, thereby improving the accuracy of building heat flux prediction.
[0045] The formula for preprocessing road network data for the target city within the target time period to obtain weighted road density data is as follows:
[0046]
[0047]
[0048] Among them, w l V represents the weight of the level l road; l The design traffic volume is denoted by m; m is the road level number; s is the search radius; Density s L is the weighted road density within the search radius; n is the number of roads within the search radius; r A is the length of road r within the search radius; s This represents the area within the search radius.
[0049] Here, road classifications and their weights conform to Chinese highway engineering technical standards. Weighted road density data reflects the differences in traffic activity intensity between different road classifications, which can improve the accuracy of traffic heat flux prediction.
[0050] The formula for preprocessing POI data for the target city within the target time period to obtain weighted factory density data is expressed as follows:
[0051]
[0052]
[0053] Among them, w c The weight of factory type c; E c Energy consumption for factory type c; m is the number of factory types; s is the search radius; Density s A is the weighted factory density within the search radius; s denoted as , where is the area within the search radius; and 'n' is the number of factories within the search radius.
[0054] Here, the factory POI data of the target city is divided into light industry and heavy industry. Light industry is further divided into other light industries and printing, garment and furniture factories, which are more common in the city center, to obtain the factory types and numbers in the target city. The energy consumption of factory types is derived from energy consumption data of various industries in China. Weighted factory density data can reflect the differences in energy consumption of different types of factories or industrial areas. Adding weighted factory density features can improve the accuracy of industrial heat flux estimation.
[0055] Here, the preprocessing of the data is implemented in Python and ArcGIS, and the resulting feature data is converted into list data to be used as input data for each prediction model in step 102.
[0056] Step 102 is implemented in R, and the output is converted into a geographic raster.
[0057] In step 103, the monthly heat flux of each heat source is: monthly building heat flux originating from building heat, monthly transportation heat flux originating from transportation heat, monthly industrial heat flux originating from industrial heat, and metabolic heat flux originating from metabolic heat.
[0058] In step 104, the hourly anthropogenic heat flux AH of the target city within the target time period. h The formula is expressed as:
[0059]
[0060] in, and The monthly building heat flux, monthly transportation heat flux, and monthly industrial heat flux of the target city in month m are, in order. The hourly building heat flux, hourly traffic heat flux, and hourly industrial heat flux of the target city are listed in order. The metabolic heat flux per hour (h) is calculated from the monthly metabolic heat flux of the target city in month m based on the activity and dormant metabolic intensities. The relative values of the population heatmap for the target city; The daily industrial heat variation coefficient of the target city.
[0061] Relative values of the population heatmap of the target city The formula is expressed as:
[0062]
[0063] Among them, Ph h The population heat map value for the target city within hour h is used to describe the real-time distribution of the population in the city and is used to calculate the changes in the intensity of heat emissions from buildings and traffic in the city throughout the day. Here, the population heat map value is based on the geographical location of mobile phone users.
[0064] The daily industrial heat variation coefficient of the target city is obtained from the daily industrial heat variation curve of the target city.
[0065] In one possible implementation, it also includes:
[0066] A first sample set is constructed using the city-level monthly building heat flux as the sample label and public features and building height as the sample features; a second sample set is constructed using the city-level monthly traffic heat flux as the sample label and public features and weighted road density as the sample features; a third sample set is constructed using the city-level monthly industrial heat flux as the sample label and public features and weighted factory density as the sample features; and a fourth sample set is constructed using the city-level monthly metabolic heat flux as the sample label and public features and building height as the sample features.
[0067] Build a stacking integration model;
[0068] The Stacking ensemble model is trained using the first sample set to obtain the building heat prediction model; the Stacking ensemble model is trained using the second sample set to obtain the traffic heat prediction model; the Stacking ensemble model is trained using the third sample set to obtain the industrial heat prediction model; and the Stacking ensemble model is trained using the fourth sample set to obtain the metabolic heat prediction model.
[0069] in,
[0070] The sample labels for each sample set were obtained by estimating and downscaling based on socioeconomic and energy consumption data using a top-down energy consumption inventory method, as detailed below:
[0071] A: Based on energy consumption data and socioeconomic data from provincial and municipal statistical yearbooks, a top-down energy consumption inventory method is used to estimate the annual building heat flux (Q) within the urban area. B Annual traffic heat flux (Q)T Annual industrial heat flux (Q) I ) and annual metabolic heat flux (Q M The formula is expressed as:
[0072]
[0073]
[0074]
[0075]
[0076] Among them, C L C C C I C T These are provincial-level energy consumption for residential use, commercial use, industrial use, and transportation infrastructure. All energy consumption is converted to standard coal equivalent, in tce; ε is the standard coal calorific value, based on my country's energy consumption standard ε = 29.3 MJ / kg; γ v This figure represents the percentage (%) of fuel consumption in total household energy consumption across four representative regions of China, based on statistics from the 2015 Chinese General Social Survey (CGSS). p It is the proportion of urban population to the total population of the province; α c It is the proportion of electricity consumption or GDP of the city's tertiary industry in the province; α i It represents the proportion of urban industrial electricity consumption or gross industrial output value to the province's total; A represents the urban area, in m². 2 T represents the length of a year, measured in seconds; H A and H S Metabolic heat emission intensity during active and dormant periods, T A and T S Let P represent activity time and rest time, and P represent the total population of the city.
[0077] B. Downscaling the annual building heat flux, annual transportation heat flux, and annual industrial heat flux according to the time dimension yields the monthly building heat flux, monthly transportation heat flux, and monthly industrial heat flux, expressed by the following formula:
[0078] Q month =Q year ×β m
[0079]
[0080] Among them, Q month Represents monthly building heat flux, monthly transportation heat flux, or monthly industrial heat flux; Qyear To Q month The corresponding annual building heat flux, annual transportation heat flux, or annual industrial heat flux; β m Weighting for each month.
[0081] The monthly weights of different heat sources are assigned using the corresponding alternative data δ. m For building heat calculations, δ is calculated based on the monthly average temperature, using the energy consumption variation with temperature and the temperature equilibrium point representing the minimum energy consumption. m Regarding traffic congestion, δ m This represents monthly freight volume or electricity consumption in transportation; for industrial heat, δ m This represents monthly industrial electricity consumption or total industrial output. Specific data selection is determined based on the actual availability of statistical data, and all data and calculations related to monthly weightings are performed at the provincial level.
[0082] Here, since the metabolic heat flux is relatively small, it is considered to have no monthly variation, and the annual metabolic heat flux is used as the monthly metabolic heat flux.
[0083] The method for obtaining sample features of any of the sample sets can be the same as the method for preprocessing data in step 101, which involves obtaining remote sensing data, meteorological data, building grid tiles, road network data, and POI data from the cities and time periods corresponding to the sample labels of the sample set.
[0084] In another possible implementation, the construction of the Stacking integration model includes:
[0085] The Stacking ensemble model is constructed using extreme gradient boosting, random forest, Cubist, and support vector machine as base models and multiple linear regression as the meta-model.
[0086] Here, we use five-fold cross-validation to implement Stacking ensemble model training in R.
[0087] The use of a stacking ensemble model enables each prediction model to have strong generalization ability and high prediction accuracy.
[0088] Figure 2 The figures show the monthly building heat flux, monthly transportation heat flux, and monthly industrial heat flux of some cities predicted by various prediction models according to embodiments of the present invention.
[0089] Figure 3 , 4 The figures shown are schematic diagrams of the hourly anthropogenic heat flux in Beijing and Shanghai in April, estimated using embodiments of the present invention.
[0090] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention.
Claims
1. A method for artificially estimating heat, characterized in that, include: Preprocess remote sensing data, meteorological data, building grid tiles, road network data, and POI data for the target city and target time period to obtain public feature data, building height data, weighted road density data, and weighted factory density data; Input public characteristic data and building height data into the building heat prediction model to output monthly building heat flux; input public characteristic data and weighted road density data into the traffic heat prediction model to output monthly traffic heat flux; input public characteristic data and weighted factory density data into the industrial heat prediction model to output monthly industrial heat flux; input public characteristic data and building height data into the metabolic heat prediction model to output monthly metabolic heat flux. Based on the relative values of the daily population heat map of the target city, the industrial heat change coefficient, and the intensity of human activity and dormant metabolism, the hourly heat flux corresponding to the monthly heat flux of each heat source is calculated. The calculated hourly heat fluxes are summed to obtain the hourly anthropogenic heat flux of the target city within the target time period. The public features include: nighttime lights, day and night surface temperature, vegetation index, digital elevation model, air temperature, humidity, and wind speed; A first sample set is constructed using the city-level monthly building heat flux as the sample label and public features and building height as the sample features; a second sample set is constructed using the city-level monthly traffic heat flux as the sample label and public features and weighted road density as the sample features; a third sample set is constructed using the city-level monthly industrial heat flux as the sample label and public features and weighted factory density as the sample features; and a fourth sample set is constructed using the city-level monthly metabolic heat flux as the sample label and public features and building height as the sample features. Build a stacking integration model; The Stacking ensemble model is trained using the first sample set to obtain the building heat prediction model; the Stacking ensemble model is trained using the second sample set to obtain the traffic heat prediction model; the Stacking ensemble model is trained using the third sample set to obtain the industrial heat prediction model; and the Stacking ensemble model is trained using the fourth sample set to obtain the metabolic heat prediction model. The construction of the Stacking integration model includes: The Stacking ensemble model is constructed using extreme gradient boosting, random forest, Cubist and support vector machine as base models and multiple linear regression as meta-model. The formula for obtaining the weighted road density data is expressed as follows: in, The weight of the level l road; The design traffic volume is denoted as l; j is the road level; s is the search radius. The weighted road density within the search radius; k is the number of roads within the search radius; The length of road r within the search radius; The area within the search radius; The formula for obtaining the weighted plant density data is expressed as follows: in, The weight of factory type c; s represents the energy consumption of factory type c; C represents the number of factory types; s represents the search radius; The weighted factory density within the search radius; The area within the search radius is n; n is the number of factories within the search radius. The formula for the hourly anthropogenic heat flux of the target city is as follows: in, , and The monthly building heat flux, monthly transportation heat flux, and monthly industrial heat flux of the target city in month m are, in order. The metabolic heat flux per hour (h) is calculated from the monthly metabolic heat flux of the target city in month m based on the activity and dormant metabolic intensities. The relative values of the population heatmap for the target city; The daily industrial heat variation coefficient of the target city.