Cold region university campus functional region carbon emission prediction method based on deep learning and image recognition

Abstract

The invention relates to the field of environment, in particular to a cold-region university campus functional area carbon emission prediction method based on deep learning and image recognition. The objective of the invention is to solve the problem of low carbon emission prediction accuracy of the existing cold-region university campus functional area. According to the cold region university campus functional area carbon emission prediction method based on deep learning and image recognition, a performance simulation method and a deep learning method based on image recognition are combined, and carbon emission data of campus functional areas are obtained through performance simulation. The spatial distribution information of the spatial form and the function of the functional area is translated through the functional layout feature map, and the mapping relation among the spatial form, the functional layout and the carbon emission of the campus functional area is established through a deep learning algorithm, so that the carbon emission performance of the complex campus functional area can be quickly predicted in a short time. The method solves the problem that the existing carbon emission prediction accuracy of the functional area of the college campus in the cold region is low.

Description

Technical Field

[0001] This invention relates to the field of environment, and more particularly to a method for predicting carbon emissions from functional areas of university campuses in cold regions based on deep learning and image recognition. Background Technology

[0002] Existing methods for predicting carbon emissions on campuses and related methods for predicting carbon emissions in urban areas are mostly divided into two categories.

[0003] One type is based on performance simulation (including building energy consumption simulation and photovoltaic performance simulation). These methods require a lot of time in the prediction and optimization process (e.g., Analysis and Quantitative Simulation Method of Carbon Energy Spatiotemporal Distribution in Urban Residential Building Systems CN201810940667.8; A Regional Integrated Energy Planning Method and System Considering Urban Morphology CN202311314105.X; A Regional Energy System Planning Method Considering Building Morphology and Layout CN202210870180.3).

[0004] Another type is based on machine learning prediction models. These methods rely heavily on historical carbon emission data of existing buildings. However, historical carbon emission data of buildings cannot be obtained during the building design phase. It is also quite difficult to obtain a sufficient number of historical carbon emission data of buildings of the same region and function type (Regional carbon emission prediction method and system based on carbon emission and big data analysis CN202410494380.2; A regional carbon emission prediction method and system based on deep learning model CN202411306274.3; Building performance prediction method, device, equipment and medium CN202411129003.5; Carbon emission prediction method, device and electronic equipment of buildings in university area CN202411502226.1;).

[0005] A few studies have combined performance simulation with machine learning algorithms to predict building carbon emissions in urban areas (a machine learning-based multi-objective optimization method for energy conservation and carbon reduction in urban blocks, CN202410060267.3). However, this method is limited to single-function urban areas (such as residential areas), and the geometric parameters cannot fully capture the spatial distribution information of buildings with different functions in the functional areas of a campus. Therefore, existing carbon emission prediction technologies for functional areas of university campuses in cold regions suffer from low prediction accuracy. Summary of the Invention

[0006] The purpose of this invention is to address the low accuracy of existing carbon emission prediction methods for functional areas of university campuses in cold regions. It provides a carbon emission prediction method for functional areas of university campuses in cold regions based on deep learning and image recognition, comprising:

[0007] Step 1: Construct a carbon emission prediction model for functional zones of university campuses in cold regions; obtain a trained carbon emission prediction model for functional zones of university campuses in cold regions.

[0008] The universities in cold regions are those located in the cold and frigid regions as defined in the "Standard for Building Climate Zoning" (GB50178-93).

[0009] The functional areas of the university campus in the cold region include: teaching buildings, library, laboratory buildings, dormitories, canteen, and activity center;

[0010] Step 2: Obtain the building base outline vector data, elevation data, city where the university is located, functional type, and electricity carbon emission factor of the university campus in the predicted cold region.

[0011] The method of obtaining the building base contour vector data and elevation data of the functional area of ​​the university campus in the cold region to be predicted is to obtain urban information through a geographic information system (GIS) and extract it from the urban information, which is known to those skilled in the art.

[0012] The building base contour vector data is a digital representation of the projected contour of the building base on the ground, belonging to the "vector data" type in Geographic Information System (GIS).

[0013] Precisely recording the shape, location, and spatial relationships of the outer boundary of a building's foundation using "point-line-plane" geometric objects provides fundamental geographic information for describing the spatial form of a building; this is a definition well-known to those in the field.

[0014] Elevation data is the relative height of the highest point of a building to its lowest point (ignoring the limited undulations of the campus surface), used to describe the spatial position of the building;

[0015] Based on the obtained building base contour vector data and elevation data of the functional areas of the university campus in the cold region to be predicted, the HSV layout feature map of the functional areas of the university campus in the cold region to be predicted is obtained.

[0016] Step 3: Input the HSV layout feature map of the functional area of ​​the university campus in cold regions to be predicted, the city where the university is located, the functional type of the functional area of ​​the university campus, and the electricity carbon emission factor of the city where the university is located into the trained carbon emission prediction model for the functional area of ​​the university campus in cold regions.

[0017] The trained carbon emission prediction model for functional zones of university campuses in cold regions outputs predicted carbon emission results based on the input.

[0018] Step one involves constructing a carbon emission prediction model for functional zones of university campuses in cold regions; obtaining a trained carbon emission prediction model for functional zones of university campuses in cold regions; the specific process is as follows:

[0019] Step 11: Obtain the training dataset; the training dataset is divided into training set, validation set, and test set according to the specified proportions;

[0020] The training dataset in step one-one includes:

[0021] First training dataset (carbon emission intensity dataset of functional area buildings in universities in cold regions)

[0022] The first training dataset (feature map dataset of functional area layout of university campus in cold regions) is used as the prediction label for the second training dataset to calculate the loss function.

[0023] The first and second training datasets are divided into training, validation, and test sets in proportion. The validation set is used to determine whether the model has converged. After the model training is completed and the hyperparameters are tuned, the test set is used to finally evaluate the model's generalization ability and ensure that its performance on completely unknown data meets expectations. In actual use, the pre-trained model is used.

[0024] Steps 1 and 2: Construct a carbon emission prediction model for functional areas of university campuses in cold regions; train the carbon emission prediction model for functional areas of university campuses in cold regions based on the training dataset to obtain a well-trained carbon emission prediction model for functional areas of university campuses in cold regions.

[0025] The specific process of obtaining the training dataset in step one includes:

[0026] A1: Obtain the vector data of the building base contours and the elevation data of N functional areas of a university campus in a cold region; N is a positive integer;

[0027] The vector data of the building base contours and the elevation data of N functional areas of university campuses in cold regions were obtained; this was obtained by extracting urban information from a Geographic Information System (GIS); this is known to those skilled in the art.

[0028] A2: Based on the building base contour vector data and elevation data of N functional areas of university campuses in cold regions, and by labeling the functional types of buildings in the functional areas using layers, N three-dimensional digital models are created using the visualization programming plugin platform in the 3D modeling software.

[0029] The method described in A2, which involves creating N 3D digital models using the building base contour vector data and building elevation data of the nth functional area of ​​the cold-region university campus, employs a visual programming plugin platform within 3D modeling software; is well-known to those skilled in the art. The visual programming plugin platform is a plugin for 3D modeling software that provides visual programming functionality.

[0030] A3: Translate N three-dimensional digital models into N HSV layout feature maps, and construct the N HSV layout feature maps into a second training dataset (layout feature map dataset of functional areas of university campuses in cold regions).

[0031] A4: Obtain meteorological data for the cities where N university campuses in cold regions are located;

[0032] Obtain thermal carbon emission factor data for N cities where university campuses in cold regions are located.

[0033] Obtain electricity carbon emission data for N cities where university campuses in cold regions are located;

[0034] Obtain simulated energy consumption data for functional areas of N university campuses in cold regions;

[0035] Obtain simulation data on the solar power generation potential of functional areas in N university campuses in cold regions;

[0036] Meteorological data for a typical meteorological year in a city can be downloaded from the EPW MAP website.

[0037] For urban thermal carbon emission factors, please refer to: Building Carbon Emission Calculation Standard GB / T 51366-2019.

[0038] For information on urban electricity carbon emission factors, please refer to: Study on Carbon Dioxide Emission Factors of Regional Power Grids in China (2023).

[0039] Energy consumption and solar power generation potential were obtained through simulations, as explained in the following section.

[0040] A5: Based on meteorological data of the cities where N university campuses are located, thermal carbon emission factor data of the cities where N university campuses are located, electricity carbon emission data of the cities where N university campuses are located, energy consumption simulation data of the functional areas of N university campuses in cold regions, and solar power generation potential simulation data of the functional areas of N university campuses in cold regions, the building carbon emission intensity data of the functional areas of N university campuses in cold regions are calculated.

[0041] The first training set was constructed based on the building carbon emission intensity data of functional areas of N university campuses in cold regions.

[0042] In A3, N three-dimensional digital models are translated into N HSV layout feature maps. The specific process is as follows:

[0043] The process of translating the nth 3D digital model into the nth HSV layout feature map is as follows:

[0044] A3.1: In the visual programming plugin platform, use specific RGB colors to attach markers for buildings of different functional types, as H channel feature information in the nth HSV layout feature map; the specific process is as follows:

[0045] The teaching building in the nth three-dimensional digital model is marked with a red (255,0,0) marker.

[0046] The library building in the nth 3D digital model is marked with an orange (255,123,0) tag.

[0047] The experimental building of the nth three-dimensional digital model is marked with a green (102,255,0) anchor.

[0048] The dormitory building in the nth 3D digital model is marked with blue (0,157,255);

[0049] The cafeteria building in the nth 3D digital model is marked with purple (251,0,255);

[0050] Use yellow (255,234,0) to attach a marker to the nth 3D digital model of the activity center building;

[0051] HSV (Hue-Saturation-Lightness) is a perceptually related color model that is closer to human subjective perception of color.

[0052] H (Hue): Represents the "basic category" of color (such as red, orange, yellow, green, blue, and purple), and its value range is usually 0° to 360° (corresponding to the angle on the color wheel).

[0053] S (Saturation): Indicates the "vividness" (purity) of a color, 0% is gray (no color), and 100% is the purest color;

[0054] V (Value, Brightness): Represents the "lightness" of a color, where 0% is black and 100% is the brightest color;

[0055] The information in the H channel is the color's "hue value".

[0056] A3.2: In the visual programming plugin platform, the elevation information data of the nth 3D digital model is mapped to a saturation value, and the saturation value is used as the S-channel feature information in the nth HSV layout feature map; the specific process is as follows:

[0057] In the process of mapping elevation data to saturation indices, the elevation range of 0-100m is mapped to a saturation value of 0-1.

[0058] In the buildings on the university campuses in the cold regions involved, their height relative to the ground is almost all within this range. If there are a very few that exceed 100m, they are approximated as 100m.

[0059] Elevation information is the relative height of the highest point of a building unit relative to its lowest point. This patent ignores the limited undulations of the campus surface.

[0060] A3.3: Combine the H channel feature information from the nth HSV layout feature map obtained in A3.1, the S channel feature information from the nth HSV layout feature map obtained in A3.2, and the default V channel feature information in the visual programming plugin platform to form the nth HSV layout feature map;

[0061] The output campus functional area HSV layout feature map is 450 pixels × 450 pixels in size, and the corresponding original 3D digital model has a planar size of 1500m × 1500m.

[0062] In section A5, based on meteorological data, thermal carbon emission factor data, electricity carbon emission data, energy consumption simulation data, and solar power generation potential simulation data of the N cities where the N university campuses are located, the building carbon emission intensity data of the N university campuses in cold regions is calculated. The first training dataset is then constructed based on this building carbon emission intensity data. The specific process is as follows:

[0063] A5.1: Establish energy consumption simulation models for N functional areas of a university campus in a cold region within a visual programming plugin platform; these plugins are commonly used in the visual programming plugin platform and are well-known to those in the field.

[0064] A5.2: Input the meteorological data of the cities where the N university campuses are located and the energy consumption simulation data of the N functional areas of the N university campuses into the energy consumption simulation models of the N university campus functional areas respectively.

[0065] Meteorological data is one of the data required for building energy consumption simulation, but it is not the focus of this invention. It is common knowledge among those skilled in the art when inputting energy consumption simulation models of functional areas of university campuses in cold regions.

[0066] The annual cooling energy consumption of N functional areas of a university campus in a cold region was calculated.

[0067] Heating energy consumption of functional areas in N university campuses in cold regions

[0068] Lighting energy consumption of functional areas in N university campuses in cold regions

[0069] Energy consumption of electrical equipment in functional areas of N university campuses in cold regions;

[0070] These four values ​​can be directly output after simulation using an energy consumption simulation plugin, and the method is relatively mature and well-established, requiring no additional formulas. They are common knowledge to those in the field.

[0071] Building energy consumption intensity of functional areas in N university campuses in cold regions;

[0072] The energy intensity calculated in this invention is the overall building energy intensity of the entire functional area. The formula for calculating building energy intensity is as follows:

[0073]

[0074] in, Represents the building's heating energy consumption (kWh / y). Represents cooling energy consumption (kWh / y). Represents lighting energy consumption (kWh / y). This represents the energy consumption of the equipment (kWh / y). Let m represent the building area of ​​the i-th building, and m represent the number of buildings.

[0075] A5.3: Establish N simulation models of the solar power generation potential of functional areas of a university campus based on spatial morphology in a visual programming plugin platform;

[0076] A5.4: Input the meteorological data of the cities where N university campuses in cold regions are located and the simulated data of the solar power generation potential of the functional areas of N university campuses in cold regions into N simulation models of the solar power generation potential of the functional areas of university campuses based on spatial morphology; calculate the annual solar power generation of the N university campus functional areas in cold regions;

[0077] Annual solar power generation is calculated by multiplying the power generation intensity by the total building area, and the unit is kWh / m². 2 / y.

[0078] The formula for calculating the solar power generation intensity of the nth functional area of ​​a university campus in a cold region is:

[0079]

[0080]

[0081] in, The solar irradiance received on the surface of a photovoltaic module (kWh / m² / y) is derived through simulation using meteorological data. This process is well-known.

[0082] This indicates the conversion efficiency of the photovoltaic module. These represent the loss efficiency of the photovoltaic power generation system, and are set to 18% and 15%, respectively. This indicates the building surface area (m2) available for installing photovoltaic modules. This indicates the area of ​​the building's roof (m2). This represents the area of ​​the building facade (m2). It was calculated using the battery plugin within the visual programming plugin platform prior to the simulation. This battery component automatically picks up the building's roof and facade areas, then multiplies them by the appropriate coefficients to calculate the building surface area suitable for installing photovoltaic modules. This is well-known to those skilled in the art.

[0083] Roof installation coefficient and facade installation coefficient They were set to 0.7 and 0.5 respectively;

[0084] A5.5: Based on the thermal carbon emission factor data of the cities where N university campuses are located, the electricity carbon emission data of the cities where N university campuses are located, and the annual solar power generation of the functional areas of N university campuses, the building carbon emission intensity data of the functional areas of N university campuses in cold regions are calculated; the specific process is as follows:

[0085] The formula for calculating the building carbon emission intensity data of the nth functional area of ​​a university campus in a cold region is:

[0086]

[0087] in, This indicates the building's carbon emission factor related to electricity, while This represents the thermal carbon emission factor. The settings are shown in Table 3. Set to 0.340;

[0088] A5.6: Construct the first training dataset using building carbon emission intensity data from functional areas of N university campuses in cold regions;

[0089] The carbon emission prediction model for functional zones of university campuses in cold regions in steps one and two includes, in sequence:

[0090] First convolutional layer, first ReLU activation function layer, first max pooling layer

[0091] Second convolutional layer, second ReLU activation function layer, second max pooling layer

[0092] Third convolutional layer, third ReLU activation function layer, third max pooling layer

[0093] Fourth convolutional layer, fourth ReLU activation function layer, fourth max pooling layer

[0094] Fifth convolutional layer, fifth ReLU activation function layer, fifth max pooling layer

[0095] First fully connected layer, sixth ReLU activation function layer, first dropout layer;

[0096] Second fully connected layer, seventh ReLU activation function layer, second dropout layer;

[0097] The third fully connected layer;

[0098] The beneficial effects of this invention are as follows:

[0099] This study combines performance simulation methods with image recognition-based deep learning methods. It obtains carbon emission data of campus functional areas through performance simulation, translates the spatial distribution information of functional area spatial form and function through functional layout feature maps, and establishes the mapping relationship between campus functional area spatial form, functional layout and carbon emission through deep learning algorithms. This enables rapid prediction of carbon emission performance of complex campus functional areas in a short period of time, so as to assist architects in carrying out low-carbon design of university campuses within a limited design cycle.

[0100] This invention systematically considers the spatial morphology, power consumption distribution, and regional climatic characteristics of functional areas in university campuses in cold regions, combined with the advantages of performance simulation methods and deep learning methods. It achieves rapid prediction of building carbon emission intensity based on functional area layout feature images, effectively assisting architects in conducting low-carbon design for university campuses in cold regions within a limited design cycle. Compared to traditional performance simulation methods, the method provided by this patent reduces the carbon emission intensity prediction time for a campus functional area from 1-3 hours to 1.2 milliseconds, significantly improving prediction efficiency. Compared to existing machine learning methods, the method provided by this patent does not rely on a large amount of historical carbon emission measurement data and has a stronger ability to identify building functional distribution characteristics, providing an effective approach for carbon emission performance evaluation of complex campus functional area schemes. Attached Figure Description

[0101] Figure 1 Flowchart of a carbon emission prediction method for functional zones in university campuses in cold regions based on deep learning and image recognition;

[0102] Figure 2 Schematic diagram of functional areas on a university campus in a cold region;

[0103] Figure 3 A schematic diagram of a carbon emission prediction model for functional zones in a university campus in a cold region based on image recognition. Detailed Implementation

[0104] Specific implementation method one: Combining Figures 1-3 This invention is described.

[0105] Step 1: Construct a carbon emission prediction model for functional zones of university campuses in cold regions; obtain a trained carbon emission prediction model for functional zones of university campuses in cold regions.

[0106] The universities in cold regions are those located in the cold and frigid regions as defined in the "Standard for Building Climate Zoning" (GB50178-93).

[0107] The functional areas of the university campus in the cold region include: teaching buildings, library, laboratory buildings, dormitories, canteen, and activity center;

[0108] Step 2: Obtain the building base outline vector data, elevation data, city where the university is located, functional type, and electricity carbon emission factor of the university campus in the predicted cold region.

[0109] The method of obtaining the building base contour vector data and elevation data of the functional area of ​​the university campus in the cold region to be predicted is to obtain urban information through a geographic information system (GIS) and extract it from the urban information, which is known to those skilled in the art.

[0110] The building base contour vector data is a digital representation of the projected contour of the building base on the ground, belonging to the "vector data" type in Geographic Information System (GIS).

[0111] Precisely recording the shape, location, and spatial relationships of the outer boundary of a building's foundation using "point-line-plane" geometric objects provides fundamental geographic information for describing the spatial form of a building; this is a definition well-known to those in the field.

[0112] Elevation data is the relative height of the highest point of a building to its lowest point (ignoring the limited undulations of the campus surface), used to describe the spatial position of the building;

[0113] Based on the obtained building base contour vector data and elevation data of the functional areas of the university campus in the cold region to be predicted, the HSV layout feature map of the functional areas of the university campus in the cold region to be predicted is obtained.

[0114] Step 3: Input the HSV layout feature map of the functional area of ​​the university campus in cold regions to be predicted, the city where the university is located, the functional type of the functional area of ​​the university campus, and the electricity carbon emission factor of the city where the university is located into the trained carbon emission prediction model for the functional area of ​​the university campus in cold regions.

[0115] The trained carbon emission prediction model for functional zones of university campuses in cold regions outputs predicted carbon emission results based on the input.

[0116] Specific Implementation Method Two: The difference between this implementation method and Specific Implementation Method One is that...

[0117] Step one involves constructing a carbon emission prediction model for functional zones of university campuses in cold regions; obtaining a trained carbon emission prediction model for functional zones of university campuses in cold regions; the specific process is as follows:

[0118] Step 11: Obtain the training dataset; the training dataset is divided into training set, validation set, and test set according to the specified proportions;

[0119] The training dataset in step one-one includes:

[0120] First training dataset (carbon emission intensity dataset of functional area buildings in universities in cold regions)

[0121] The first training dataset (feature map dataset of functional area layout of university campus in cold regions) is used as the prediction label for the second training dataset to calculate the loss function.

[0122] The first and second training datasets are divided into training, validation, and test sets in proportion. The validation set is used to determine whether the model has converged. After the model training is completed and the hyperparameters are tuned, the test set is used to finally evaluate the model's generalization ability and ensure that its performance on completely unknown data meets expectations. In actual use, the pre-trained model is used.

[0123] Steps 1 and 2: Construct a carbon emission prediction model for functional areas of university campuses in cold regions; train the carbon emission prediction model for functional areas of university campuses in cold regions based on the training dataset to obtain a trained carbon emission prediction model for functional areas of university campuses in cold regions.

[0124] The other steps and parameters are the same as in Specific Implementation Method 1.

[0125] Specific Implementation Method Three: The difference between this implementation method and Specific Implementation Method One is that...

[0126] The specific process of obtaining the training dataset in step one includes:

[0127] A1: Obtain the vector data of the building base contours and the elevation data of N functional areas of a university campus in a cold region; N is a positive integer;

[0128] The vector data of the building base contours and the elevation data of N functional areas of university campuses in cold regions were obtained; this was obtained by extracting urban information from a Geographic Information System (GIS); this is known to those skilled in the art.

[0129] A2: Based on the building base contour vector data and elevation data of N functional areas of university campuses in cold regions, and by labeling the functional types of buildings in the functional areas using layers, N three-dimensional digital models are created using the visualization programming plugin platform in the 3D modeling software.

[0130] The method described in A2, which involves creating N 3D digital models using the building base contour vector data and building elevation data of the nth functional area of ​​the cold-region university campus, employs a visual programming plugin platform within 3D modeling software; is well-known to those skilled in the art. The visual programming plugin platform is a plugin for 3D modeling software that provides visual programming functionality.

[0131] A3: Translate N three-dimensional digital models into N HSV layout feature maps, and construct the N HSV layout feature maps into a second training dataset (layout feature map dataset of functional areas of university campuses in cold regions).

[0132] A4: Obtain meteorological data for the cities where N university campuses in cold regions are located;

[0133] Obtain thermal carbon emission factor data for N cities where university campuses in cold regions are located.

[0134] Obtain electricity carbon emission data for N cities where university campuses in cold regions are located;

[0135] Obtain simulated energy consumption data for functional areas of N university campuses in cold regions;

[0136] Obtain simulation data on the solar power generation potential of functional areas in N university campuses in cold regions;

[0137] Meteorological data for a typical meteorological year in a city can be downloaded from the EPW MAP website.

[0138] For urban thermal carbon emission factors, please refer to: Building Carbon Emission Calculation Standard GB / T 51366-2019.

[0139] For information on urban electricity carbon emission factors, please refer to: Study on Carbon Dioxide Emission Factors of Regional Power Grids in China (2023).

[0140] Energy consumption and solar power generation potential were obtained through simulations, as explained in the following section.

[0141] A5: Based on meteorological data of the cities where N university campuses are located, thermal carbon emission factor data of the cities where N university campuses are located, electricity carbon emission data of the cities where N university campuses are located, energy consumption simulation data of the functional areas of N university campuses in cold regions, and solar power generation potential simulation data of the functional areas of N university campuses in cold regions, the building carbon emission intensity data of the functional areas of N university campuses in cold regions are calculated.

[0142] The first training set was constructed based on the building carbon emission intensity data of functional areas of N university campuses in cold regions.

[0143] The other steps and parameters are the same as in one of the specific implementation methods one or two.

[0144] Specific Implementation Method Four: This implementation method differs from Specific Implementation Methods One through Four in that...

[0145] In A3, N three-dimensional digital models are translated into N HSV layout feature maps. The specific process is as follows:

[0146] The process of translating the nth 3D digital model into the nth HSV layout feature map is as follows:

[0147] A3.1: In the visual programming plugin platform, use specific RGB colors to attach markers for buildings of different functional types, as H channel feature information in the nth HSV layout feature map; the specific process is as follows:

[0148] The teaching building in the nth three-dimensional digital model is marked with a red (255,0,0) marker.

[0149] The library building in the nth 3D digital model is marked with an orange (255,123,0) tag.

[0150] The experimental building of the nth three-dimensional digital model is marked with a green (102,255,0) anchor.

[0151] The dormitory building in the nth 3D digital model is marked with blue (0,157,255);

[0152] The cafeteria building in the nth 3D digital model is marked with purple (251,0,255);

[0153] Use yellow (255,234,0) to attach a marker to the nth 3D digital model of the activity center building;

[0154] HSV (Hue-Saturation-Lightness) is a perceptually related color model that is closer to human subjective perception of color.

[0155] H (Hue): Represents the "basic category" of color (such as red, orange, yellow, green, blue, and purple), and its value range is usually 0° to 360° (corresponding to the angle on the color wheel).

[0156] S (Saturation): Indicates the "vividness" (purity) of a color, 0% is gray (no color), and 100% is the purest color;

[0157] V (Value, Brightness): Represents the "lightness" of a color, where 0% is black and 100% is the brightest color;

[0158] The information in the H channel is the color's "hue value".

[0159] A3.2: In the visual programming plugin platform, the elevation information data of the nth 3D digital model is mapped to a saturation value, and the saturation value is used as the S-channel feature information in the nth HSV layout feature map; the specific process is as follows:

[0160] In the process of mapping elevation data to saturation indices, the elevation range of 0-100m is mapped to a saturation value of 0-1.

[0161] In the buildings on the university campuses in the cold regions involved, their height relative to the ground is almost all within this range. If there are a very few that exceed 100m, they are approximated as 100m.

[0162] Elevation information is the relative height of the highest point of a building unit relative to its lowest point. This patent ignores the limited undulations of the campus surface.

[0163] A3.3: Combine the H channel feature information from the nth HSV layout feature map obtained in A3.1, the S channel feature information from the nth HSV layout feature map obtained in A3.2, and the default V channel feature information in the visual programming plugin platform to form the nth HSV layout feature map;

[0164] The output campus functional area HSV layout feature map is 450 pixels × 450 pixels in size, and the corresponding original 3D digital model has a planar size of 1500m × 1500m.

[0165] The other steps and parameters are the same as those in one of the specific implementation methods one to three.

[0166] Specific Implementation Method Five: The difference between this implementation method and Specific Implementation Methods One to Four is that...

[0167] The A4 section obtains thermal carbon emission factor data for the cities where N university campuses in cold regions are located.

[0168] Obtain electricity carbon emission data for N cities where university campuses in cold regions are located; the specific process is as follows:

[0169] Obtain the thermal carbon emission factor data for the city where the nth university campus in the cold region is located; set it to 0.340;

[0170] Obtain the electricity carbon emission data of the city where the nth university campus in a cold region is located; the specific process is as follows: query Table 3;

[0171] Table 3. Electricity Carbon Emission Factors in Different Cold Region Cities

[0172]

[0173] The energy consumption simulation data for the nth functional area of ​​a university campus in a cold region includes:

[0174] Personnel utilization rate of the nth functional area of ​​a university campus in a cold region

[0175] The lighting power of the functional area of ​​the nth cold-region university campus

[0176] The electrical equipment power of the nth functional area of ​​the cold region university campus

[0177] Ventilation intensity of the functional areas of the nth university campus in a cold region

[0178] The heating season during the year for the functional area of ​​the nth university campus in a cold region;

[0179] The cooling period of the nth functional area of ​​a university campus in a cold region during the year;

[0180] The university holidays during the year in the functional area of ​​the nth cold region university campus;

[0181] Architectural envelope parameters of the nth functional area of ​​a university campus in a cold region;

[0182] A 3D digital model of the functional areas of the nth university campus in a cold region;

[0183] Table 1. Power Consumption Parameter Settings for Functional Area Buildings in Universities in Cold Regions

[0184]

[0185] Table 2. Parameter Setting Table for Building Envelope Structure of Functional Zones in Universities in Cold Regions

[0186]

[0187] The simulated solar power generation potential data for the nth functional area of ​​a cold-region university campus includes:

[0188] The photovoltaic module conversion efficiency of the nth functional area of ​​the cold region university campus

[0189] System loss efficiency of the nth functional area of ​​a university campus in a cold region

[0190] Roofing installation coefficient of the nth functional area of ​​a university campus in a cold region;

[0191] The facade installation coefficient of the nth functional area of ​​a university campus in a cold region;

[0192] This indicates the conversion efficiency of the photovoltaic module. The loss efficiency of the photovoltaic power generation system is represented by 18% and 15%, respectively; the roof installation factor. and facade installation coefficient They were set to 0.7 and 0.5 respectively;

[0193] A 3D digital model of the functional areas of the nth university campus in a cold region;

[0194] The other steps and parameters are the same as those in one of the specific implementation methods one to four.

[0195] Specific Implementation Method Six: The difference between this implementation method and Specific Implementation Methods One to Five is that...

[0196] In section A5, based on meteorological data, thermal carbon emission factor data, electricity carbon emission data, energy consumption simulation data, and solar power generation potential simulation data of the N cities where the N university campuses are located, the building carbon emission intensity data of the N university campuses in cold regions is calculated. The first training dataset is then constructed based on this building carbon emission intensity data. The specific process is as follows:

[0197] A5.1: Establish energy consumption simulation models for N functional areas of a university campus in a cold region within a visual programming plugin platform; these plugins, commonly used in the visual programming plugin platform, are well-known to those in the field.

[0198] A5.2: Input the meteorological data of the cities where the N university campuses are located and the energy consumption simulation data of the N functional areas of the N university campuses into the energy consumption simulation models of the N university campus functional areas respectively.

[0199] Meteorological data is one of the data required for building energy consumption simulation, but it is not the focus of this invention. It is common knowledge among those skilled in the art when inputting energy consumption simulation models of functional areas of university campuses in cold regions.

[0200] The annual cooling energy consumption of N functional areas of a university campus in a cold region was calculated.

[0201] Heating energy consumption of functional areas in N university campuses in cold regions

[0202] Lighting energy consumption of functional areas in N university campuses in cold regions

[0203] Energy consumption of electrical equipment in functional areas of N university campuses in cold regions;

[0204] These four values ​​can be directly output after simulation using an energy consumption simulation plugin, and the method is relatively mature and well-established, requiring no additional formulas. They are common knowledge to those in the field.

[0205] Building energy consumption intensity of functional areas in N university campuses in cold regions;

[0206] The energy intensity calculated in this invention is the overall building energy intensity of the entire functional area. The formula for calculating building energy intensity is as follows:

[0207]

[0208] in, Represents the building's heating energy consumption (kWh / y). Represents cooling energy consumption (kWh / y). Represents lighting energy consumption (kWh / y). This represents the energy consumption of the equipment (kWh / y). The area of ​​the i-th building is represented by m, and the number of buildings is represented by m.

[0209] A5.3: Establish N simulation models of the solar power generation potential of functional areas of a university campus based on spatial morphology in a visual programming plugin platform;

[0210] A5.4: Input the meteorological data of the cities where N university campuses in cold regions are located and the simulated data of the solar power generation potential of the functional areas of N university campuses in cold regions into N simulation models of the solar power generation potential of the functional areas of university campuses based on spatial morphology; calculate the annual solar power generation of the N university campus functional areas in cold regions;

[0211] Annual solar power generation is calculated by multiplying the power generation intensity by the total building area, and the unit is kWh / m². 2 / y.

[0212] The formula for calculating the solar power generation intensity of the nth functional area of ​​a university campus in a cold region is:

[0213]

[0214]

[0215] in, The solar irradiance received on the surface of a photovoltaic module (kWh / m² / y) is derived through simulation using meteorological data. This process is well-known.

[0216] This indicates the conversion efficiency of the photovoltaic module. These represent the loss efficiency of the photovoltaic power generation system, and are set to 18% and 15% respectively.

[0217] . This indicates the building surface area (m2) available for installing photovoltaic modules. This indicates the area of ​​the building's roof (m2). This represents the area of ​​the building facade (m2). It was calculated using the battery plugin within the visual programming plugin platform prior to the simulation. This battery component automatically picks up the building's roof and facade areas, then multiplies them by the appropriate coefficients to calculate the building surface area suitable for installing photovoltaic modules. This is well-known to those skilled in the art.

[0218] Roof installation coefficient and facade installation coefficient They were set to 0.7 and 0.5 respectively;

[0219] A5.5: Based on the thermal carbon emission factor data of the cities where N university campuses are located, the electricity carbon emission data of the cities where N university campuses are located, and the annual solar power generation of the functional areas of N university campuses, the building carbon emission intensity data of the functional areas of N university campuses in cold regions are calculated; the specific process is as follows:

[0220] The formula for calculating the building carbon emission intensity data of the nth functional area of ​​a university campus in a cold region is:

[0221]

[0222] in, This indicates the building's carbon emission factor related to electricity, while This represents the thermal carbon emission factor. The settings are shown in Table 3. Set to 0.340;

[0223] A5.6: Construct the first training dataset using building carbon emission intensity data from functional areas of N university campuses in cold regions.

[0224] The other steps and parameters are the same as those in any of the specific implementation methods one to five.

[0225] Specific Implementation Method Seven: The difference between this implementation method and Specific Implementation Methods One through Six is ​​that...

[0226] The carbon emission prediction model for functional zones of university campuses in cold regions in steps one and two includes, in sequence:

[0227] First convolutional layer, first ReLU activation function layer, first max pooling layer

[0228] Second convolutional layer, second ReLU activation function layer, second max pooling layer

[0229] Third convolutional layer, third ReLU activation function layer, third max pooling layer

[0230] Fourth convolutional layer, fourth ReLU activation function layer, fourth max pooling layer

[0231] Fifth convolutional layer, fifth ReLU activation function layer, fifth max pooling layer

[0232] First fully connected layer, sixth ReLU activation function layer, first dropout layer;

[0233] Second fully connected layer, seventh ReLU activation function layer, second dropout layer;

[0234] The third fully connected layer;

[0235] The other steps and parameters are the same as those in any of the specific implementation methods one to six.

[0236] Specific Implementation Method Eight: The difference between this implementation method and Specific Implementation Methods One to Seven is that...

[0237] The convolutional kernels of the first, second, third, fourth, and fifth convolutional layers are all 3×3;

[0238] The convolutional kernels of the first, second, third, fourth, and fifth convolutional layers are all 3×3;

[0239] The stride of the first, second, third, fourth, and fifth convolutional layers is 1.

[0240] The first, second, third, fourth, and fifth convolutional layers are all padded with 1;

[0241] The step size of the first max pooling layer, the second max pooling layer, the third max pooling layer, and the fourth max pooling layer is 3;

[0242] The convolution kernels of the fifth max pooling layer are all 4×4; the stride of the fifth max pooling layer is 4.

[0243] The optimal structure of the convolutional neural network model of this invention is as follows: Figure 3 As shown, the system consists of 5 convolutional layers and 3 fully connected layers. The convolutional modules use 3×3 kernels with a stride and padding of 1 to ensure consistent feature map size. The number of channels increases progressively from 3 to 32 to 64 to 128 to 256 to 512. Each convolutional layer is followed by a ReLU activation function and a max-pooling layer. The first four pooling layers use 3×3 kernels with a stride of 3, while the last layer uses a 4×4 kernel with a stride of 4. Through the AdaptiveAvgPool2d operation, the feature map is finally compressed to 1×1, achieving maximum reduction in feature dimension.

[0244] In the fully connected layer design, the fc1 layer takes a 512×1×1 feature tensor and three auxiliary features as input, outputting a 256-dimensional feature. This output is then activated by ReLU and followed by a dropout layer to reduce the risk of overfitting. The fc2 layer maps the 256-dimensional input to a 128-dimensional output, also using ReLU activation and dropout. The fc3 layer performs a linear transformation between the 128-dimensional input and output, ultimately outputting a single predicted value for the regression task. This model achieves a prediction accuracy of 94.88% for carbon emission intensity, meeting the prediction requirements of the early design phase.

[0245] The other steps and parameters are the same as those in any of the specific implementation methods one to seven.

[0246] Specific Implementation Method Nine: The difference between this implementation method and Specific Implementation Methods One through Eight is that...

[0247] The specific process of training the carbon emission prediction model for functional zones of university campuses in cold regions based on the training dataset in steps one and two, and obtaining the trained carbon emission prediction model for functional zones of university campuses in cold regions, is as follows:

[0248] S1: Input the second training dataset into the carbon emission prediction model of functional areas of university campuses in cold regions, and the carbon emission prediction model of functional areas of university campuses in cold regions will output the prediction results.

[0249] S2: Calculate the loss R² based on the first training dataset and the prediction results;

[0250] S3: Update the carbon emission prediction model parameters of functional zones in cold-region university campuses using the Adam optimizer based on the loss R².

[0251] The Adam optimizer has an initial learning rate of 0.001 and is equipped with the ReduceLROnPlateau learning rate scheduler;

[0252] S4: When the parameters of the carbon emission prediction model for functional zones of universities in cold regions converge, stop training and obtain the trained carbon emission prediction model for functional zones of universities in cold regions.

[0253] Preferably, to prevent overfitting, this invention selects a condition where the R² value of the validation set increases by less than 1e-5 over 30 consecutive epochs, and then considers the parameters to have converged.

[0254] R 2 RMSE and MAE are used to evaluate the prediction accuracy of prediction models and are well-known evaluation parameters in the field.

[0255] Training time, prediction time, and memory usage are used to evaluate the computational cost of a prediction model.

[0256] The formulas for calculating R², MAE, and RMSE are as follows:

[0257]

[0258]

[0259]

[0260] in, and These represent the predicted value and the actual value, respectively. The mean of the true values ​​is denoted as , and m is the number of samples. The data in the first training set of this invention are used as the true values, and the output of the carbon emission prediction model for functional areas of university campuses in cold regions is used as the predicted value. When training the campus functional area carbon emission prediction model based on convolutional neural networks, the hyperparameters of the algorithm are adjusted to optimize the prediction performance. The search range of the hyperparameters is shown in Table 4.

[0261] Table 4. Hyperparameter search range settings for convolutional neural networks

[0262]

[0263] The other steps and parameters are the same as those in Specific Implementation Methods 1 to 8.

[0264] Specific Implementation Method Ten: The difference between this implementation method and Specific Implementation Methods One through Nine is that...

[0265] In step two, the layout feature map of the functional area of ​​the university campus in the cold region to be predicted is obtained by acquiring the building base contour vector data and elevation data of the functional area of ​​the university campus in the cold region to be predicted; the specific process is as follows:

[0266] Step 21: Based on the building base contour vector data and elevation data of the functional areas of the university campus in the cold region to be predicted, and by labeling the functional types of the buildings in the functional areas with layers, a 3D digital model of the functional areas of the university campus in the cold region to be predicted is established in the visualization programming plugin platform of the 3D modeling software.

[0267] Step 22: Translate the 3D digital model of the functional areas of the university campus in the cold region to be predicted into an HSV layout feature map of the functional areas of the university campus in the cold region to be predicted; the specific process is as follows:

[0268] Step 221: In the visual programming plugin platform, use specific RGB colors to attach markers to buildings of different functional types in the 3D digital model of the functional areas of the university campus in the cold region to be predicted, as H-channel feature information in the HSV layout feature map of the functional areas of the university campus in the cold region to be predicted; the specific process is as follows:

[0269] The teaching building in the 3D digital model of the functional area of ​​the university campus in the cold region is marked with red (255,0,0).

[0270] The library building of the 3D digital model of the functional areas of a university campus in a cold region is marked with orange (255,123,0).

[0271] The experimental building in the 3D digital model of the functional areas of a university campus in a cold region is marked with green (102,255,0).

[0272] Dormitory buildings in a 3D digital model of a university campus in a cold region, marked with blue (0,157,255) tags, are used to predict the functional areas of the campus.

[0273] The canteen building of the three-dimensional digital model of the functional areas of the university campus in the cold region is marked with purple (251,0,255).

[0274] The activity center building of the 3D digital model of the functional areas of the university campus in the cold region to be predicted is marked with yellow (255,234,0);

[0275] HSV (Hue-Saturation-Lightness) is a perceptually related color model that is closer to human subjective perception of color.

[0276] H (Hue): Represents the "basic category" of color (such as red, orange, yellow, green, blue, and purple), and its value range is usually 0° to 360° (corresponding to the angle on the color wheel).

[0277] S (Saturation): Indicates the "vividness" (purity) of a color, 0% is gray (no color), and 100% is the purest color;

[0278] V (Value, Brightness): Represents the "lightness" of a color, where 0% is black and 100% is the brightest color;

[0279] The information in the H channel is the color's "hue value".

[0280] Step 222: Identify the elevation information of the buildings in the 3D digital model in the visual programming plugin platform, map the elevation data to saturation values, and use the saturation values ​​as the S-channel feature information in the HSV layout feature map of the functional area of ​​the university campus in the cold region to be predicted.

[0281] The specific process is as follows:

[0282] In the process of mapping elevation data to a saturation index, the elevation range of 0-100m is mapped to a saturation value of 0-1. Elevation information represents the relative height of the highest point of a building unit relative to its lowest point; this patent ignores the limited undulations of the campus surface.

[0283] Step 223: Combine the H channel feature information obtained in Step 221, the S channel feature information obtained in Step 222, and the default V channel feature information in the visualization programming plugin platform to form the HSV layout feature map of the functional area of ​​the university campus in the cold region to be predicted;

[0284] The output campus functional area HSV layout feature map is 450 pixels × 450 pixels in size, and the corresponding original 3D digital model has a planar size of 1500m × 1500m.

[0285] The other steps and parameters are the same as those in Specific Implementation Methods 1 to 9.

[0286] Specific implementation method eleven: This implementation method is a storage medium that stores at least one instruction. The at least one instruction is loaded and executed by a processor to implement the carbon emission prediction method for functional areas of university campuses in cold regions based on deep learning and image recognition.

[0287] It should be understood that the instructions include any computer program product, software, or computerized method corresponding to any method described in this invention; the instructions can be used to program a computer system or other electronic device. Computer storage media may include readable media on which the instructions are stored, and may include, but are not limited to, magnetic storage media and optical storage media.

[0288] Specific Implementation Method Twelve: This implementation method is an underwater acoustic target recognition device based on direct push small sample semi-supervised learning. The device includes a processor and a memory. It should be understood that this includes any device described in this invention that includes a processor and a memory. The device may also include other units and modules that perform display, interaction, processing, control and other functions through signals or instructions.

[0289] The memory stores at least one instruction, which is loaded and executed by the processor to implement the aforementioned method for predicting carbon emissions from functional areas of university campuses in cold regions based on deep learning and image recognition.

[0290] Those skilled in the art will understand that at least one stored instruction constitutes a computer program product corresponding to a method or system. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application can be implemented using various computer languages.

[0291] This application is described with reference to flowchart illustrations and / or block diagrams of methods, systems, and computer program products according to embodiments of this application, and can also be used with corresponding devices. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0292] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0293] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0294] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0295] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

[0296] The above description is merely of preferred embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent substitutions, and improvements made to the above embodiments without departing from the scope of the present invention, based on the technical essence of the present invention, and within the spirit and principles of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. A cold region university campus functional area carbon emission prediction method based on deep learning and image recognition, characterized in that, The application comprises the following steps: Step one, constructing a carbon emission prediction model for the functional area of a university campus in a cold region; obtaining a trained carbon emission prediction model for the functional area of a university campus in a cold region; The functional area of the university campus in a cold region comprises a teaching building, a library, a laboratory building, a dormitory, a dining hall, and an activity center; Step two, obtaining the architectural bottom contour vector data of the functional area of the university campus in a cold region to be predicted, the elevation data of the functional area of the university campus in a cold region to be predicted, the city where the university in a cold region is located, the functional type of the functional area of the university campus in a cold region to be predicted, and the electric power carbon emission factor of the city where the university in a cold region is located; According to the architectural bottom contour vector data of the functional area of the university campus in a cold region to be predicted and the elevation data of the functional area of the university campus in a cold region to be predicted, the HSV layout feature map of the functional area of the university campus in a cold region to be predicted is obtained; Step three, inputting the HSV layout feature map of the functional area of the university campus in a cold region to be predicted, the city where the university in a cold region is located, the functional type of the functional area of the university campus in a cold region to be predicted, and the electric power carbon emission factor of the city where the university in a cold region is located into the trained carbon emission prediction model for the functional area of a university campus in a cold region; The trained carbon emission prediction model for the functional area of a university campus in a cold region outputs a predicted carbon emission result according to the input.

2. The carbon emission prediction method for functional areas of a university campus in a cold region based on deep learning and image recognition according to claim 1, characterized in that, The step one of constructing a carbon emission prediction model for the functional area of a university campus in a cold region; obtaining a trained carbon emission prediction model for the functional area of a university campus in a cold region; The specific process is as follows: Step one: obtaining a training data set; The training data set in step one comprises: A first training data set and a second training data set Step two: constructing a carbon emission prediction model for the functional area of a university campus in a cold region; According to the training data set, the carbon emission prediction model for the functional area of a university campus in a cold region is trained to obtain a trained carbon emission prediction model for the functional area of a university campus in a cold region.

3. The carbon emission prediction method for the functional area of a university campus in a cold region based on deep learning and image recognition according to claim 2, characterized in that The specific process of obtaining the training data set in step one comprises: A1: obtaining the architectural bottom contour vector data of N functional areas of university campuses in cold regions and the elevation data of N functional areas of university campuses in cold regions; N is a positive integer; A2: according to the architectural bottom contour vector data of N functional areas of university campuses in cold regions and the elevation data of N functional areas of university campuses in cold regions, using a visual programming plug-in platform in a three-dimensional modeling software to establish N three-dimensional digital models; A3: translating the N three-dimensional digital models into N HSV layout feature maps, and constructing the N HSV layout feature maps into a second training data set; A4: obtaining thermal carbon emission factor data of N cities where universities in cold regions are located, obtaining electric power carbon emission data of N cities where universities in cold regions are located; obtaining energy consumption simulation data of N functional areas of university campuses in cold regions; obtaining solar power generation potential simulation data of N functional areas of university campuses in cold regions A5: According to the meteorological data of N cold region university campus cities, the thermal carbon emission factor data of N cold region university campus cities, the electric power carbon emission data of N cold region university campus cities, the energy consumption simulation data of N cold region university campus functional areas, and the solar power generation potential simulation data of N cold region university campus functional areas, the building carbon emission intensity data of N cold region university campus functional areas is calculated; According to the building carbon emission intensity data of N cold region university campus functional areas, a first training set is constructed.

4. The cold region university campus functional area carbon emission prediction method based on deep learning and image recognition according to claim 3, characterized in that, In A3, the N three-dimensional digital models are translated into N HSV layout feature maps, and the specific process is: Translate the nth three-dimensional digital model into the nth HSV layout feature map, and the specific process is: A3.1: In the visual programming plug-in platform, different functional types of buildings are marked with specific RGB colors as H channel feature information in the nth HSV layout feature map; The specific process is: Use red (255, 0, 0) to mark the teaching building of the nth three-dimensional digital model; Use orange (255, 123, 0) to mark the library building of the nth three-dimensional digital model; Use green (102, 255, 0) to mark the experimental building of the nth three-dimensional digital model; Use blue (0, 157, 255) to mark the dormitory building of the nth three-dimensional digital model; Use purple (251, 0, 255) to mark the canteen building of the nth three-dimensional digital model; Use yellow (255, 234, 0) to mark the activity center building of the nth three-dimensional digital model; A3.2: In the visual programming plug-in platform, the elevation information data of the nth three-dimensional digital model is mapped into the saturation value, and the saturation value is taken as the S channel feature information in the nth HSV layout feature map; A3.3: The H channel feature information of the nth HSV layout feature map obtained in A3.1, the S channel feature information of the nth HSV layout feature map obtained in A3.2, and the default V channel feature information in the visual programming plug-in platform are combined into the nth HSV layout feature map.

5. The carbon emission prediction method for functional areas in a cold region university campus based on deep learning and image recognition according to claim 4, characterized in that, In A4, the energy consumption simulation data of the nth cold region university campus functional area in the energy consumption simulation data of N cold region university campus functional areas includes: The personnel utilization rate of the nth cold region university campus functional area, The lighting power of the nth cold region university campus functional area, The electrical equipment power of the nth cold region university campus functional area, The ventilation intensity of the nth cold region university campus functional area, The heating period of the nth cold region university campus functional area in a year; The cooling period of the nth cold region university campus functional area in a year; The university holiday of the nth cold region university campus functional area in a year; The building envelope parameters of the nth cold region university campus functional area; The three-dimensional digital model of the nth cold region university campus functional area; In the solar power generation potential simulation data of N cold region university campus functional areas, the solar power generation potential simulation data of the nth cold region university campus functional area includes: The conversion efficiency of the photovoltaic module of the nth cold region university campus functional area, The system loss efficiency of the nth cold region university campus functional area, The roof installation coefficient of the nth cold region university campus functional area; The facade installation coefficient of the nth cold region university campus functional area; The three-dimensional digital model of the nth cold region university campus functional area.

6. The cold region university campus functional area carbon emission prediction method based on deep learning and image recognition according to claim 5, wherein, The building carbon emission intensity data of the N cold region university campus functional areas is calculated according to the meteorological data of the N cold region university campuses, the thermal carbon emission factor data of the N cold region university campuses, the electric power carbon emission data of the N cold region university campuses, the energy consumption simulation data of the N cold region university campus functional areas, and the solar power generation potential simulation data of the N cold region university campus functional areas; and the first training data set is constructed according to the building carbon emission intensity data of the N cold region university campus functional areas. The specific process is: A5.1: Establishing an energy consumption simulation model of the N cold region university campus functional areas in the visual programming plug-in platform; A5.2: Inputting the energy consumption simulation data of the N cold region university campus functional areas into the energy consumption simulation model of the N cold region university campus functional areas, calculating the annual refrigeration energy consumption of the N cold region university campus functional areas, the heating energy consumption of the N cold region university campus functional areas, the lighting energy consumption of the N cold region university campus functional areas, the electrical equipment energy consumption of the N cold region university campus functional areas; the building energy consumption intensity of the N cold region university campus functional areas; A5.3: Establishing a solar power generation potential simulation model of the N university campus functional areas based on spatial form in the visual programming plug-in platform; A5.4: Inputting the solar power generation potential simulation data of the N cold region university campus functional areas into the solar power generation potential simulation model of the N university campus functional areas based on spatial form; and calculating the annual solar power generation of the N cold region university campus functional areas; A5.5: Calculating the building carbon emission intensity data of the N cold region university campus functional areas according to the thermal carbon emission factor data of the N cold region university campuses, the electric power carbon emission data of the N cold region university campuses, and the annual solar power generation of the N cold region university campus functional areas; A5.6: Constructing the first training data set from the building carbon emission intensity data of the N cold region university campus functional areas.

7. The cold region university campus functional area carbon emission prediction method based on deep learning and image recognition according to claim 6, wherein, The cold region university campus functional area carbon emission prediction model in steps one and two successively includes: a first convolutional layer, a first ReLU activation function layer, and a first max-pooling layer, a second convolutional layer, a second ReLU activation function layer, and a second max-pooling layer, a third convolutional layer, a third ReLU activation function layer, and a third max-pooling layer, a fourth convolutional layer, a fourth ReLU activation function layer, and a fourth max-pooling layer, a fifth convolutional layer, a fifth ReLU activation function layer, and a fifth max-pooling layer, a first fully connected layer, a sixth ReLU activation function layer, and a first dropout layer; a second fully connected layer, a seventh ReLU activation function layer, and a second dropout layer; a third fully connected layer.

8. The carbon emission prediction method for functional areas in a cold region university campus based on deep learning and image recognition according to claim 7, characterized in that, The convolution kernel of the first convolution layer, the second convolution layer, the third convolution layer, the fourth convolution layer, and the fifth convolution layer is 3×3; The convolution kernel of the first convolution layer, the second convolution layer, the third convolution layer, the fourth convolution layer, and the fifth convolution layer is 3×3; The stride of the first convolution layer, the second convolution layer, the third convolution layer, the fourth convolution layer, and the fifth convolution layer is 1; The padding of the first convolution layer, the second convolution layer, the third convolution layer, the fourth convolution layer, and the fifth convolution layer is 1; The stride of the first max pooling layer, the second max pooling layer, the third max pooling layer, and the fourth max pooling layer is 3; The convolution kernel of the fifth max pooling layer is 4×4, and the stride of the fifth max pooling layer is 4.

9. The method according to claim 8, wherein, The specific process of training the cold region university campus functional area carbon emission prediction model according to the training data set in step one is as follows: S1: inputting the second training data set into the cold region university campus functional area carbon emission prediction model, and outputting a prediction result by the cold region university campus functional area carbon emission prediction model; S2: calculating a loss R² according to the first training data set and the prediction result; S3: updating the parameters of the cold region university campus functional area carbon emission prediction model using an Adam optimizer according to the loss R²; S4: when the parameters of the cold region university campus functional area carbon emission prediction model converge, stopping the training, and obtaining the trained cold region university campus functional area carbon emission prediction model.

10. The method according to claim 9, wherein, The specific process of obtaining the layout feature map of the to-be-predicted cold region university campus functional area according to the building bottom contour vector data of the to-be-predicted cold region university campus functional area and the elevation data of the to-be-predicted cold region university campus functional area in step two is as follows: Step two one: establishing a three-dimensional digital model of the to-be-predicted cold region university campus functional area in a visual programming plug-in platform of a three-dimensional modeling software according to the building bottom contour vector data of the to-be-predicted cold region university campus functional area and the elevation data of the to-be-predicted cold region university campus functional area; Step two two: translating the three-dimensional digital model of the to-be-predicted cold region university campus functional area into an HSV layout feature map of the to-be-predicted cold region university campus functional area; the specific process is as follows: Step two two one: using a specific RGB color to attach a marker to different functional types of buildings in the three-dimensional digital model of the to-be-predicted cold region university campus functional area in the visual programming plug-in platform, as H channel feature information in the H channel of the HSV layout feature map of the to-be-predicted cold region university campus functional area; The specific process is as follows: using red (255, 0, 0) to attach a marker to the teaching building of the three-dimensional digital model of the to-be-predicted cold region university campus functional area; using orange (255, 123, 0) to attach a marker to the library building of the three-dimensional digital model of the to-be-predicted cold region university campus functional area; using green (102, 255, 0) to attach a marker to the experimental building of the three-dimensional digital model of the to-be-predicted cold region university campus functional area; The dormitory building of the three-dimensional digital model of the functional area of the university campus in the cold region to be predicted is marked using blue (0, 157, 255); The canteen building of the three-dimensional digital model of the functional area of the university campus in the cold region to be predicted is marked using purple (251, 0, 255); The activity center building of the three-dimensional digital model of the functional area of the university campus in the cold region to be predicted is marked using yellow (255, 234, 0); Step two hundred and twenty-two: identify the elevation information of the building of the three-dimensional digital model in the visual programming plug-in platform, map the elevation data to the saturation value, and take the saturation value as the S channel feature information in the HSV layout feature map of the functional area of the university campus in the cold region to be predicted; Step two hundred and twenty-three: combine the H channel feature information obtained in step two hundred and twenty-one, the S channel feature information obtained in step two hundred and twenty-two, and the default V channel feature information in the visual programming plug-in platform into the HSV layout feature map of the functional area of the university campus in the cold region to be predicted.

Images