Deep learning method and system based on station-city fusion space energy consumption model

Abstract

The invention provides a deep learning method and system based on a station-city fusion space energy consumption model, and the method comprises the steps: analyzing a core energy consumption system of a core space region in a station-city fusion three-dimensional network space through employing a space region analysis method, and constructing a station-city fusion three-dimensional network energy consumption model of the core energy consumption system; constructing a carbon emission mechanism model through a station-city fusion three-dimensional network energy consumption model, defining a loss function of the carbon emission mechanism model according to a deep learning neural network structure of an initial input variable and an output variable in the carbon emission mechanism model, and performing iterative training on the loss function by adopting an iterative optimization function to obtain a carbon emission mechanism model; and outputting a deep learning neural network agent model until the loss value is lower than an allowable value, and finally evaluating the green operation performance of the station-city fusion three-dimensional network space by adopting the deep learning neural network agent model according to specification requirements to obtain an evaluation result. Compared with the prior art, the method is higher in evaluation efficiency and higher in reliability.

Description

Technical Field

[0001] The present invention belongs to the technical field of station-city space energy consumption assessment, and specifically relates to a deep learning method and system based on a station-city integrated space energy consumption model. Background Art

[0002] Station-city integration is also called "station-city integration". At the macro level, it refers to the relationship of mutual connection and interaction between railways, urban rail transit stations and surrounding cities in various elements such as functional space and transportation mechanism. At the project level, it means that when planning and constructing hubs, combined with the city's own conditions, relying on the spillover effects of railway and rail development, measures are taken according to local conditions to achieve the coordinated development of hubs and their surrounding areas. Station-city integration advocates the close integration of important transportation stations with urban space, that is, to arrange daily office, residence and urban services within the walking distance of the station, so as to provide citizens with convenient lifestyles and economic activities. The three-dimensional network space of station-city integration covers the spatial structure of traditional comprehensive passenger transport hubs and various spatial areas connected with cities. As the development direction of future comprehensive transportation hubs, the performance evaluation requirements of its green operation are becoming increasingly prominent. Among them, the current relevant regulations have put forward higher requirements on how to quickly and accurately evaluate carbon emissions for green evaluation indicators of carbon emissions.

[0003] In the existing technology, the carbon emission assessment for station-city integration mainly adopts a data-driven deep learning network model, which requires manual preparation of a large amount of historical data to train the model, which is costly and time-consuming. In addition, the data-driven deep learning network model often directly adopts foreign black box models. When inputting data into the black box model, the input and data relationship of the data are not understood. The trained model has poor reliability and is difficult to meet actual needs. Summary of the invention

[0004] The present invention provides a deep learning method and system based on a station-city integrated spatial energy consumption model, which are used to solve the problems existing in the prior art.

[0005] In a first aspect, to achieve the above-mentioned purpose, the present invention provides a deep learning method based on a station-city integrated spatial energy consumption model, comprising: S10: Use the spatial regional analysis method to analyze the core energy consumption system of the core space area in the station-city integrated three-dimensional network space, and construct a station-city integrated three-dimensional network energy consumption model corresponding to the core energy consumption system; S20: Construct a carbon emission mechanism model based on the energy consumption type and characteristics of the station-city integrated three-dimensional network energy consumption model; Among them, the core energy consumption systems include HVAC systems, lighting systems and elevator systems; the energy consumption types include HVAC energy consumption, lighting energy consumption and elevator energy consumption; the energy consumption characteristics include HVAC energy consumption characteristics, lighting energy consumption characteristics and elevator energy consumption characteristics; S30: According to the number of input variables, the matrix size of the input variables and the number of output variables in the carbon emission mechanism model, define the deep learning neural network structure of the input variables and the output variables, and construct the loss function of the carbon emission mechanism model according to the deep learning neural network structure; S40: using an iterative optimization function to iteratively train the loss function until the loss value obtained by the loss function is lower than a preset allowable value, and outputting a deep learning neural network agent model based on the carbon emission mechanism; S50: Using a deep learning neural network agent model, the green operation performance of the station-city integrated three-dimensional network is evaluated in accordance with the green building carbon emission specifications and the evaluation results are obtained.

[0006] Through the above technical scheme, before constructing the deep learning neural network proxy model, the deep learning method of the present invention first constructs a carbon emission mechanism model under the operation state of the station-city integrated three-dimensional network space, and defines the loss function and iteratively optimizes the carbon emission mechanism model to obtain the deep learning neural network proxy model. Compared with the existing technology, the deep learning method of the present invention can independently complete the model training through real-time monitored physical data, and the evaluation efficiency is higher.

[0007] Preferably, step S10 includes: S11: Determine the core spatial area of ​​the station-city integrated three-dimensional network; The core space area includes at least one or more of the waiting room, platform, transfer space and concentrated commercial space; S12: Use spatial regional analysis method to analyze the core energy consumption system in the core spatial area; Among them, the core energy consumption system includes at least one or more of the HVAC system, lighting system and elevator system; S13: Constructing a corresponding energy consumption model according to the core energy consumption system, including: Constructing an air conditioning energy consumption model corresponding to the HVAC system; and / or Constructing a lighting energy consumption model corresponding to the lighting system; and / or Construct an elevator energy consumption model corresponding to the elevator system; Among them, the station-city integrated three-dimensional network energy consumption model includes at least one or more of the air conditioning energy consumption model, lighting energy consumption model and elevator energy consumption model.

[0008] Preferably, the air conditioning energy consumption model is specifically: (1) in, is the annual energy consumption of air conditioning, i.e. HVAC energy consumption; is the cooling / heating power of the air conditioner; The operating time of the air conditioner; is the coefficient of performance of the air conditioner, specifically the ratio of the cooling / heating effect of the air conditioner to the energy consumed; , and It belongs to the HVAC energy consumption characteristics.

[0009] Preferably, the lighting energy consumption model is specifically: (2) in, is the annual energy consumption of the lighting system, i.e., lighting energy consumption; For the day Lighting power density value of each room; For the Lighting area of ​​each room; For the j day i Room lighting time; The lighting power density of emergency lighting; Evaluate the building's footprint for the target; , , , and Belongs to the lighting energy consumption characteristics.

[0010] Preferably, the elevator energy consumption model is specifically: (3) in, is the annual energy consumption of the elevator, i.e., the energy consumption of the elevator; for specific energy consumption; is the average annual operating hours of the elevator; is the elevator speed; is the rated load capacity of the elevator; Energy consumption when the elevator is in standby mode; is the average annual standby hours of the elevator; , , , , and This is the energy consumption characteristic of the elevator.

[0011] Preferably, the carbon emission mechanism model in step S20 is specifically:

[0012] in, For the purpose of evaluating the building use r Carbon emissions from each refrigerant; For HVAC systems r Refrigerant charge amount; For the life of the HVAC system; For the r The global warming potential of the refrigerants; Evaluate the carbon emissions per unit floor area during the building operation phase for the target; Evaluate the building for the target Annual consumption of energy; For the Carbon emission factors of energy sources; Evaluate the annual carbon reduction of green space carbon sink systems in buildings for the target; Assess the design life of the building for the purpose; Evaluate the building's footprint for the target; The data type of is parameter matrix; The matrix elements include , and Respectively in Energy consumption data of the energy consumption of the category, i.e. , and , No. This type of energy includes electric energy, gas and oil energy, and municipal thermal energy; For the The energy consumption system consumes the energy provided by the renewable energy system. Type of energy consumption.

[0013] Preferably, step S30 includes: S31: Initialize the number of input variables, the matrix size of input variables and the number of output variables in the carbon emission mechanism model, define the number of input and output neurons, the number of neural network layers and the number of neurons in each neural network layer of the deep learning neural network structure; S32: After data preparation is completed, the loss function of the carbon emission mechanism model is constructed based on the network structure of the deep learning neural network.

[0014] Preferably, the loss function of the carbon emission mechanism model in step S32 is specifically: (7)

[0015] in, is the data domain weight, is the mechanism domain weight; is the loss function In the data domain-based sub-items; N is the number of iterations, The target evaluation building is obtained by random training of the deep learning model. Forecast of annual energy consumption, is the loss function of the carbon emission mechanism model.

[0016] Preferably, the green operation performance of the station-city integrated three-dimensional network space is evaluated, and the evaluation results are as follows: (9) in, According to the US Department of Energy standard, 15kg CO2 / m 2 ; is the value in formula (7) After iterative calculation, the target evaluates the carbon emissions per unit building area during the building operation phase; if , then the target building is a green building. , then the target evaluation building is a non-green building. In the second aspect, to achieve the above purpose, the present invention provides a deep learning system based on the station-city integrated space energy consumption model, the system adopts the above deep learning method based on the station-city integrated space energy consumption model, the system includes: The energy consumption model building module is used to analyze the core energy consumption system of the core space area in the station-city integration three-dimensional network space by using the spatial area analysis method, and to build a station-city integration three-dimensional network energy consumption model corresponding to the core energy consumption system; The carbon emission mechanism model construction module is used to construct a carbon emission mechanism model based on the energy consumption type and energy consumption characteristics of the station-city integrated three-dimensional network energy consumption model; A loss function definition module is used to define the deep learning neural network structure of input variables and output variables according to the number of input variables, the matrix size of input variables and the number of output variables in the carbon emission mechanism model, and to construct the loss function of the carbon emission mechanism model according to the deep learning neural network structure; An iterative optimization module is used to use an iterative optimization function to iteratively train the loss function until the loss value obtained by the loss function is lower than a preset allowable value, and output a deep learning neural network agent model based on the carbon emission mechanism; The evaluation module is used to evaluate the green operation performance of the station-city integrated three-dimensional network by adopting a deep learning neural network agent model in accordance with the requirements of the green building carbon emission specifications to obtain the evaluation results.

[0017] The deep learning method and system based on the station-city integrated space energy consumption model of the present invention have at least the following advantages: 1) Compared with the existing technology, the deep learning method of the present invention can independently complete the model training through real-time monitored physical data, and the evaluation efficiency is higher.

[0018] 2) Compared with the existing technology, the deep learning method of the present invention fully understands the data input and data relationship, and the trained model is more reliable.

[0019] 3) Compared with the prior art, the deep learning method of the present invention imposes physical information constraints during the PINN training process, so that a more generalizable model can be learned with fewer data samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 This is a flow chart of the deep learning method based on the station-city integrated spatial energy consumption model in Example 1 of the present invention.

[0022] Figure 2 This is a structural diagram of the deep learning system based on the station-city integrated spatial energy consumption model in Example 2 of the present invention.

[0023] Figure 3 This is a schematic diagram of the structure of a computing device in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0024] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] This embodiment discloses a deep learning method and system based on a station-city integrated spatial energy consumption model, which are used to solve the problems existing in the prior art.

[0026] Embodiment 1 like Figure 1 As shown, this embodiment discloses a deep learning method based on a station-city integrated space energy consumption model, including: S10: The spatial regional analysis method is used to analyze the core energy consumption system of the core spatial area within the station-city integrated three-dimensional network space, and a station-city integrated three-dimensional network energy consumption model corresponding to the core energy consumption system is constructed.

[0027] In this embodiment, the core space areas in the above-mentioned station-city integrated three-dimensional network space include waiting rooms, platforms, transfer spaces and concentrated commercial spaces. The core energy consumption systems in each core space area are generally HVAC systems, lighting systems and elevator systems. Spatial regional analysis is an existing technology in geographic information systems (GIS), which uses geographic spatial data for various analyses to reveal spatial relationships, patterns and trends. The station-city integrated three-dimensional network energy consumption model includes an air conditioning energy consumption model, a lighting energy consumption model and an elevator energy consumption model.

[0028] Specifically, the above step S10 includes: S11: Determine the core spatial area of ​​the station-city integrated three-dimensional network; S12: Use spatial regional analysis method to analyze the core energy consumption system in the core spatial area; S13: Constructing a corresponding energy consumption model according to the core energy consumption system, including: Constructing an air conditioning energy consumption model corresponding to the HVAC system; and / or Constructing a lighting energy consumption model corresponding to the lighting system; and / or Construct an elevator energy consumption model corresponding to the elevator system.

[0029] Furthermore, the above air conditioning energy consumption model is specifically as follows: (1) in, is the annual energy consumption of air conditioning, i.e. HVAC energy consumption; is the cooling / heating power of the air conditioner; The operating time of the air conditioner; is the coefficient of performance of the air conditioner, specifically the ratio of the cooling / heating effect of the air conditioner to the energy consumed; , and The above air conditioning energy consumption model can calculate the annual total energy consumption of the air conditioning system, distribution system, air handling system and terminal air handling system in the HVAC system.

[0030] Furthermore, the above lighting energy consumption model is specifically as follows: (2) in, is the annual energy consumption of the lighting system, i.e., lighting energy consumption, in kWh / year; For the day The lighting power density value of each room, in watts per square meter; For the Lighting area of ​​each room, in square meters; For the j day i Lighting time of each room, in hours; is the lighting power density of emergency lighting, in watts per square meter; A The floor space of the building being assessed for the target, in square meters. , , , and Belongs to the lighting energy consumption characteristics.

[0031] Furthermore, the elevator energy consumption model is specifically as follows: (3) in, is the annual energy consumption of the elevator, i.e., the energy consumption of the elevator, in kWh / year; is the specific energy consumption in mWh / kgm; is the average annual operating hours of the elevator; is the elevator speed in meters per second; is the rated load capacity of the elevator, in kilograms; Energy consumption of the elevator in standby mode, in watts; is the average annual standby hours of the elevator; , , , , and This is the energy consumption characteristic of the elevator.

[0032] After the above step S10, the deep learning method of the present invention further includes: S20: Construct a carbon emission mechanism model based on the energy consumption types and characteristics of the station-city integrated three-dimensional network energy consumption model.

[0033] Among them, the core energy consumption systems include HVAC systems, lighting systems and elevator systems; energy consumption types include HVAC energy consumption, lighting energy consumption and elevator energy consumption; energy consumption characteristics include HVAC energy consumption characteristics, lighting energy consumption characteristics and elevator energy consumption characteristics.

[0034] Specifically, the carbon emission mechanism model in step S20 is:

[0035] It should be noted that since the use of refrigerants in HVAC systems will generate greenhouse gas emissions, formula (4) should be used separately for calculation. r Indicates the refrigerant type serial number. Different types of refrigerants correspond to unique serial numbers. The relevant information of the corresponding refrigerant can be extracted based on the refrigerant serial number.

[0036] in, For the purpose of evaluating the building use r Carbon emissions from each refrigerant, in tCO2 / year; For HVAC systems r The amount of refrigerant charged, in kg / unit; is the service life of the HVAC system, in years; For the r The global warming potential of the refrigerants; The target is to assess the carbon emissions per unit building area during the building operation phase, in kgCO2 / m2; Evaluate the building for the target Annual consumption of energy of the type, in units of corresponding energy standard units (electricity kW, gas and oil m 3 , municipal heat KJ) / year; For the Carbon emission factors of energy sources; For the Energy consumption system Energy consumption of the type, in units of corresponding energy standard units (electricity kW, gas and oil m 3 , municipal heat KJ) / year; The data type of is parameter matrix; The matrix elements include , and Respectively in Energy consumption data of the energy consumption of the category, i.e. , and , No. This type of energy includes electric energy, gas and oil energy, and municipal thermal energy; For the The energy consumption system consumes the energy provided by the renewable energy system. Energy consumption of the type, in units of corresponding energy standard units (electricity kW, gas and oil m 3 , municipal heat KJ) / year; The annual carbon reduction of the green carbon sink system in the building is evaluated for the target, in kgCO2 / year; The design life of the building is evaluated for the purpose, in years; A The floor space of the building being assessed for the target, in square meters.

[0037] It should be noted that the above formula (1) represents the model of the HVAC system at the output end, and the corresponding approximate model of the energy supply end is:

[0038] The total energy supply of the above three energy sources is equal to the calculation result in formula (1), , , It is the preset energy supply coefficient. , , As The elements of the parameter matrix are calculated by using the multivariate linear regression method in combination with formula (1). If some of the energy is generated by photovoltaic or hydroelectric power, it also belongs to elements.

[0039] Similarly, the above formulas (2)-(3) also represent the model of the corresponding system at the output end level. The corresponding functional end model is similar to the above description, and the element calculation method is consistent with the above method, which will not be repeated here.

[0040] S30: According to the number of input variables, the matrix size of the input variables and the number of output variables required in the carbon emission mechanism model, define the deep learning neural network structure of the input variables and the output variables, and construct the loss function of the carbon emission mechanism model based on the deep learning neural network structure.

[0041] Specifically, step S30 includes: S31: Initialize the number of input variables, the matrix size of input variables and the number of output variables in the carbon emission mechanism model, define the number of input and output neurons, the number of neural network layers and the number of neurons in each neural network layer of the deep learning neural network structure; S32: After data preparation is completed, the loss function of the carbon emission mechanism model is constructed based on the network structure of the deep learning neural network.

[0042] It needs to be explained that the above-mentioned data preparation refers to the routine processing of processing, integrating, transforming and normalizing the original input data to finally obtain the data set.

[0043] Furthermore, the loss function of the carbon emission mechanism model is specifically: (7)

[0044] in, is the data domain weight, is the mechanism domain weight; is the loss function In the data domain-based sub-items; N is the number of iterations, The target evaluation building is obtained by random training of the deep learning model. Forecast of annual energy consumption, is the loss function of the carbon emission mechanism model.

[0045] S40: Using an iterative optimization function, iteratively train the loss function until the loss value obtained by the loss function is lower than a preset allowable value, and output a deep learning neural network agent model based on the carbon emission mechanism.

[0046] In this embodiment, the iterative optimization function adopts the quasi-Newton method in the second-order optimization method, and the allowable value is 0.01.

[0047] Specifically, the data set a and data set b prepared in step S32 are input into formula (7) and formula (8) respectively, and the quasi-Newton method is used to perform adaptive iterative training until the loss value is lower than the allowable value, and then the adaptive iterative training is stopped.

[0048] Preferably, after obtaining the deep learning neural network proxy model based on the carbon emission mechanism, the test set c is input into the deep learning neural network proxy model based on the carbon emission mechanism, and the model accuracy is evaluated. If the evaluation accuracy reaches 99%, it means that the model training is successful.

[0049] S50: Using a deep learning neural network agent model, the green operation performance of the station-city integrated three-dimensional network is evaluated in accordance with the green building carbon emission specifications and the evaluation results are obtained.

[0050] Among them, the above-mentioned evaluation of the green operation performance of the station-city integrated three-dimensional network space has the following specific evaluation results: (9) in, According to the US Department of Energy standard, 15kg CO2 / m 2 , for After iterative calculation, the target evaluates the carbon emissions per unit building area during the building operation phase; if , then the target building is a green building. , then the target assessment building is a non-green building.

[0051] Through the above steps S10-S50, the carbon emission mechanism model under the operation state of the station-city integrated three-dimensional network space is constructed by the deep learning method of the present invention, and the accurate evaluation of energy consumption under the operation performance of the station-city integrated three-dimensional network space is realized, and the efficiency of real-time evaluation is improved. At the same time, the deep learning method of the present invention specifically constructs the deep learning neural network agent model, which is not a black box model. In addition, the deep learning method of the present invention fully understands the input and data relationship of the data, and the trained model is more reliable. Physical information constraints are imposed during the PINN training process, so that a more generalized model can be learned with fewer data samples.

[0052] Embodiment 2 like Figure 2 As shown, this embodiment discloses a deep learning system based on a station-city integrated space energy consumption model. The system adopts the deep learning method based on a station-city integrated space energy consumption model in Embodiment 1. The system of the present invention includes: The energy consumption model building module is used to analyze the core energy consumption system of the core space area in the station-city integration three-dimensional network space by using the spatial area analysis method, and to build a station-city integration three-dimensional network energy consumption model corresponding to the core energy consumption system; The carbon emission mechanism model construction module is used to construct a carbon emission mechanism model based on the energy consumption type and energy consumption characteristics of the station-city integrated three-dimensional network energy consumption model; A loss function definition module is used to define the deep learning neural network structure of input variables and output variables according to the number of input variables, the matrix size of input variables and the number of output variables in the carbon emission mechanism model, and to construct the loss function of the carbon emission mechanism model according to the deep learning neural network structure; An iterative optimization module is used to use an iterative optimization function to iteratively train the loss function until the loss value obtained by the loss function is lower than a preset allowable value, and output a deep learning neural network agent model based on the carbon emission mechanism; The evaluation module is used to evaluate the green operation performance of the station-city integrated three-dimensional network by adopting a deep learning neural network agent model in accordance with the requirements of the green building carbon emission specifications to obtain the evaluation results.

[0053] Embodiment 3 like Figure 3 As shown, this embodiment discloses a computer device, which includes an internal memory, a processor, a non-volatile storage medium, and a computer program stored in the non-volatile storage medium and executable on the processor. When the processor executes the computer program, the information scheduling method based on the large language model in Embodiment 2 is implemented.

[0054] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc. Various media that can store program codes.

[0055] For those skilled in the art, various other corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all of these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A deep learning method based on the station-city integrated space energy consumption model is characterized by: include: S10: Analyze the core energy consumption system of the core space area in the station-city integrated three-dimensional network space by using the space area analysis method, and construct a station-city integrated three-dimensional network energy consumption model corresponding to the core energy consumption system; S20: constructing a carbon emission mechanism model according to the energy consumption type and energy consumption characteristics of the station-city integrated three-dimensional network energy consumption model; The core energy consumption systems include HVAC systems, lighting systems and elevator systems; the energy consumption types include HVAC energy consumption, lighting energy consumption and elevator energy consumption; the energy consumption characteristics include HVAC energy consumption characteristics, lighting energy consumption characteristics and elevator energy consumption characteristics; S30: defining a deep learning neural network structure of the input variables and the output variables according to the number of input variables, the matrix size of the input variables and the number of output variables in the carbon emission mechanism model, and constructing a loss function of the carbon emission mechanism model according to the deep learning neural network structure; S40: using an iterative optimization function to iteratively train the loss function until the loss value obtained by the loss function is lower than a preset allowable value, and outputting a deep learning neural network proxy model based on the carbon emission mechanism; S50: Using the deep learning neural network agent model, the green operation performance of the station-city integrated three-dimensional network is evaluated in accordance with the green building carbon emission specification requirements to obtain an evaluation result.

2. The deep learning method based on the station-city integrated space energy consumption model according to claim 1 is characterized in that: The step S10 comprises: S11: Determine the core space area of ​​the station-city integrated three-dimensional network; The core space area includes at least one or more of the waiting room, platform, transfer space and concentrated commercial space; S12: Analyze the core energy consumption system of the core space area by using a space area analysis method; Wherein, the core energy consumption system includes at least one or more of the HVAC system, the lighting system and the elevator system; S13: Constructing a corresponding energy consumption model according to the core energy consumption system, including: Constructing an air conditioning energy consumption model corresponding to the HVAC system; and / or Constructing a lighting energy consumption model corresponding to the lighting system; and / or Constructing an elevator energy consumption model corresponding to the elevator system; Among them, the station-city integrated three-dimensional network energy consumption model includes at least one or more of the air-conditioning energy consumption model, the lighting energy consumption model and the elevator energy consumption model.

3. The deep learning method based on the station-city integrated space energy consumption model according to claim 2 is characterized in that: The air conditioning energy consumption model is specifically: （1） in, is the annual energy consumption of air conditioning, i.e. the HVAC energy consumption; is the cooling / heating power of the air conditioner; The operating time of the air conditioner; is the coefficient of performance of the air conditioner, specifically the ratio of the cooling / heating effect of the air conditioner to the energy consumed; , and Belongs to the HVAC energy consumption characteristics.

4. The deep learning method based on the station-city integrated space energy consumption model according to claim 2 is characterized in that: The lighting energy consumption model is specifically: （2） in, is the annual energy consumption of the lighting system, i.e. the lighting energy consumption; For the day Lighting power density value of each room; For the Lighting area of ​​each room; For the j day i Room lighting time; The lighting power density of emergency lighting; Evaluate the building's footprint for the target; , , , and Belongs to the lighting energy consumption characteristics.

5. The deep learning method based on the station-city integrated space energy consumption model according to claim 2 is characterized in that: The elevator energy consumption model is specifically: （3） in, is the annual energy consumption of the elevator, i.e., the energy consumption of the elevator; for specific energy consumption; is the average annual operating hours of the elevator; is the elevator speed; is the rated load capacity of the elevator; Energy consumption when the elevator is in standby mode; is the average annual standby hours of the elevator; , , , , and This belongs to the energy consumption characteristics of the elevator.

6. The deep learning method based on the station-city integrated space energy consumption model according to claim 2 is characterized in that: The carbon emission mechanism model in step S20 is specifically: in, For the purpose of evaluating the building use r Carbon emissions from each refrigerant; The HVAC system r Refrigerant charge amount; is the useful life of the HVAC system; For the r The global warming potential of the refrigerants; Evaluate the carbon emissions per unit floor area during the building operation phase for the target; Evaluate the building for the target Annual consumption of energy; For the Carbon emission factors of energy sources; Evaluate the annual carbon reduction of green space carbon sink systems in buildings for the target; Assess the design life of the building for the purpose; Evaluate the building's footprint for the target; The data type of is parameter matrix; k include a、l and e , The matrix elements include , and Respectively in Energy consumption data in the energy consumption category, i.e. , and , No. This type of energy includes electric energy, gas and oil energy, and municipal thermal energy; For the The energy consumption system consumes the energy provided by the renewable energy system. Type of energy consumption.

7. The deep learning method based on the station-city integrated space energy consumption model according to claim 6 is characterized in that: The step S30 comprises: S31: Initialize the number of input variables, the matrix size of the input variables and the number of output variables in the carbon emission mechanism model, and define the number of input and output neurons, the number of neural network layers and the number of neurons in each neural network layer of the deep learning neural network structure; S32: After the data preparation is completed, a loss function of the carbon emission mechanism model is constructed according to the network structure of the deep learning neural network.

8. The deep learning method based on the station-city integrated space energy consumption model according to claim 7 is characterized in that: The loss function of the carbon emission mechanism model in step S32 is specifically: （7） in, is the data domain weight, is the mechanism domain weight; is the loss function In the data domain-based sub-items; N is the number of iterations, The target evaluation building is obtained by random training of the deep learning model. Forecast of annual energy consumption, is the loss function of the carbon emission mechanism model.

9. The deep learning method based on the station-city integrated space energy consumption model according to claim 1 is characterized in that: The green operation performance of the station-city integrated three-dimensional network is evaluated, and the evaluation results are as follows: （9） in, is the benchmark parameter, for The target after iterative calculation is to evaluate the carbon emissions per unit building area during the building operation phase; like , then the target assessment building is a green building; like , then the target assessment building is a non-green building.

10. A deep learning system based on a station-city integrated space energy consumption model, using a deep learning method based on a station-city integrated space energy consumption model as described in any one of claims 1 to 9, characterized in that: The deep learning system comprises: An energy consumption model construction module is used to analyze the core energy consumption system of the core space area in the station-city integration three-dimensional network space by using the space area analysis method, and to construct a station-city integration three-dimensional network energy consumption model corresponding to the core energy consumption system; A carbon emission mechanism model construction module is used to construct a carbon emission mechanism model according to the energy consumption type and energy consumption characteristics of the station-city integrated three-dimensional network energy consumption model; A loss function definition module, used to define the deep learning neural network structure of the input variables and the output variables according to the number of input variables, the matrix size of the input variables and the number of output variables in the carbon emission mechanism model, and construct the loss function of the carbon emission mechanism model according to the deep learning neural network structure; An iterative optimization module, used to use an iterative optimization function to iteratively train the loss function until the loss value obtained by the loss function is lower than a preset allowable value, and output a deep learning neural network proxy model based on the carbon emission mechanism; The evaluation module is used to use the deep learning neural network agent model to evaluate the green operation performance of the station-city integrated three-dimensional network in accordance with the green building carbon emission specification requirements to obtain an evaluation result.

Images