Residential building energy efficiency rating method and system based on graph neural network

Abstract

The invention discloses a residential building energy efficiency rating method based on a graph neural network. The method comprises the following steps: S1, generating a residential building energy efficiency parameter graph topology based on residential building parameters; s2, selecting a training sample, and establishing a residential building energy efficiency rating model based on graph neural network training by taking the residential building energy efficiency parameter graph as input and the residential building energy efficiency grade as output; and S3, selecting a test sample, inputting test sample data based on the residential building energy efficiency rating model, and obtaining a corresponding residential building energy efficiency rating evaluation result. The invention further discloses a residential building energy efficiency rating system based on the graph neural network. In this way, a new method and a new thought are provided for residential building energy efficiency evaluation.

Description

Technical Field

[0001] The present invention relates to the field of building energy efficiency rating, and particularly to a method and system for rating the energy efficiency of residential buildings based on graph neural networks. Background Art

[0002] Urban and rural construction is one of the main areas of carbon emissions. With the rapid advancement of urbanization and the in-depth adjustment of the industrial structure, the carbon emissions in the field of urban and rural construction and their proportion in the total social carbon emissions will further increase.

[0003] By statistically verifying and calibrating the building operation energy consumption and building service volume, the building energy consumption ratio is calculated, and the building energy efficiency level is given in the form of scores.

[0004] Therefore, there is an urgent need to provide a new method for rating the energy efficiency of residential buildings based on graph neural networks, providing new methods and ideas for the evaluation of the energy efficiency of residential buildings. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for rating the energy efficiency of residential buildings based on graph neural networks, which can provide a new method for the evaluation of the energy efficiency of residential buildings.

[0006] To solve the above technical problem, a technical solution adopted by the present invention is: to provide a method for rating the energy efficiency of residential buildings based on graph neural networks, including the following steps:

[0007] S1: Generate a topology of the energy efficiency parameter graph of the residential building based on the residential building parameters;

[0008] S2: Select training samples, use the energy efficiency parameter graph of the residential building as the input and the energy efficiency grade of the residential building as the output, and train based on the graph neural network to establish an energy efficiency rating model for the residential building;

[0009] S3: Select test samples, input the test sample data based on the energy efficiency rating model of the residential building, and obtain the corresponding evaluation results of the energy efficiency grade of the residential building.

[0010] In a preferred embodiment of the present invention, the residential building parameters include building area, annual energy consumption of the heating system, annual energy consumption of the air conditioning system, annual energy consumption of the domestic hot water system, annual energy consumption of the lighting system, annual energy consumption parameter of the elevator system, and renewable energy power generation.

[0011] In a preferred embodiment of the present invention, in step S2, the specific steps of establishing an energy efficiency rating model for the residential building based on the training of the graph neural network include:

[0012] S201: Based on the residential building parameter graph, through the method of multi-round message passing, embed the structural information and node features of the graph into the representation of each node. Through a series of graph convolutional layers, each node gradually aggregates the information of its neighbor nodes and updates its embedding;

[0013] S202: Aggregate the embeddings of all nodes into a graph embedding with a fixed dimension;

[0014] S203: Use the aggregated graph embedding for graph classification, and the output classification result is the energy efficiency grade of the residential building.

[0015] Furthermore, in step S201, the information transfer mechanism in the graph convolutional network is carried out using the following formula:

[0016]

[0017] where, H (l) is the node embedding matrix of the th layer, H (0) = X, X is the residential building energy efficiency parameter graph, is the regularized adjacency matrix, W (l) is the weight matrix of the th layer, and σ(·) is the activation function.

[0018] Furthermore, in step S202, global pooling is used to aggregate the embeddings of the nodes into a single graph embedding:

[0019]

[0020] where, is the embedding of each node in the Lth layer, N is the number of nodes in the graph, and h G is the global embedding of the graph.

[0021] Furthermore, in step S203, the graph embedding is used for graph classification through a fully connected layer and a classifier:

[0022] z = W f h G + b f (3)

[0023] y = softmax(z) (4)

[0024] where, W f and b f are the weight matrix and bias vector of the fully connected layer, z is the mapped classification output, and y is the class probability distribution normalized by softmax.

[0025] To solve the above technical problems, another technical solution adopted by the present invention is: to provide a residential building energy efficiency rating system based on a graph neural network, including:

[0026] The residential building energy efficiency parameter map generation module is used to generate the topology of the residential building energy efficiency parameter map based on the residential building parameters;

[0027] The residential building energy efficiency rating model establishment module is used to select training samples, use the residential building energy efficiency parameter map among them as input, and the residential building energy efficiency grade as output, and train based on the graph neural network to establish a residential building energy efficiency rating model;

[0028] The residential building energy efficiency grade evaluation result output module is used to select test samples, based on the residential building energy efficiency rating model, input the test sample data, and obtain the corresponding residential building energy efficiency grade evaluation result.

[0029] In a preferred embodiment of the present invention, the residential building parameters on which the residential building energy efficiency parameter map generation module is based include building area, annual energy consumption of the heating system, annual energy consumption of the air conditioning system, annual energy consumption of the domestic hot water system, annual energy consumption of the lighting system, annual energy consumption parameter of the elevator system, and renewable energy power generation.

[0030] In a preferred embodiment of the present invention, the specific steps for the residential building energy efficiency rating model establishment module to establish the residential building energy efficiency rating model include:

[0031] S201: Based on the residential building parameter map, through the method of multi-round message passing, embed the structure information and node features of the graph into the representation of each node. Through a series of graph convolutional layers, each node gradually aggregates the information of neighboring nodes and updates its embedding;

[0032] S202: Aggregate the embeddings of all nodes into a graph embedding with a fixed dimension;

[0033] S203: Use the aggregated graph embedding for graph classification, and the output classification result is the residential building energy efficiency grade.

[0034] The beneficial effects of the present invention are: The present invention provides a new method and new idea for the energy efficiency evaluation of residential buildings. Based on parameters such as the building area of residential buildings, annual energy consumption of the heating system, annual energy consumption of the air conditioning system, etc., a residential building energy efficiency parameter map is generated. Using the residential building energy efficiency parameter map as input and the residential building energy efficiency grade as output, a single-graph input and single-variable output regression model is trained and established through the graph neural network for the energy efficiency rating of residential buildings. Description of the Drawings

[0035] Figure 1 It is a schematic flow chart of the residential building energy efficiency rating method based on the graph neural network of the present invention;

[0036] Figure 2It is the topological structure diagram of the energy efficiency parameter diagram of the residential building;

[0037] Figure 3 It is the structural schematic diagram of the residential building energy efficiency rating model based on the graph neural network;

[0038] Figure 4 It is the structural schematic diagram of the residential building energy efficiency rating system based on the graph neural network. Specific implementation mode

[0039] The following elaborates on the preferred embodiments of the present invention in conjunction with the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0040] Please refer to Figure 1 , the embodiments of the present invention include:

[0041] A residential building energy efficiency rating method based on the graph neural network, comprising the following steps:

[0042] S1: Generate the topology of the residential building energy efficiency parameter diagram based on the residential building parameters;

[0043] Specifically, generate the topology of the residential building energy efficiency parameter diagram with the residential building parameters as nodes and the membership relationship as edges based on the residential building parameters and their RDF triples.

[0044] The input residential building parameters include: building area, annual energy consumption of the heating system, annual energy consumption of the air conditioning system, annual energy consumption of the domestic hot water system, annual energy consumption of the lighting system, annual energy consumption parameter of the elevator system, renewable energy power generation, etc., as shown in Table 1.

[0045] Table 1 List of residential building parameters

[0046]

[0047] Based on the membership relationship between the "indicators" of each parameter in the table, convert the table into a series of RDF triples. For example:

[0048] "Building area" is a lower-level indicator of "building", and the constructed RDF triple is

[0049] <’Building area’, ’is a lower-level indicator’, ’building’>

[0050] "Annual energy consumption of the heating system" is a lower-level indicator of "building energy consumption", and the constructed RDF triple is

[0051] <’Annual energy consumption of the heating system’, ’is a lower-level indicator’, ’building energy consumption’>

[0052] Based on the RDF triples, generate the topological structure of the energy efficiency parameter graph of residential buildings in the neo4j software environment, as Figure 2 shown.

[0053] S2: Collect the real residential building parameters and input them into the topological structure of the residential building energy efficiency parameter graph established in step S1 to generate training samples and test samples. Select the training samples, use the energy efficiency parameter graph of the residential building as the input, and the energy efficiency grade of the residential building as the output, and train based on the graph neural network to establish a residential building energy efficiency rating model; the specific steps include:

[0054] S201: Based on the residential building parameter graph, through the method of multi-round message passing, embed the structure information and node features of the graph into the representation of each node. Through a series of graph convolutional layers, each node gradually aggregates the information of its neighbor nodes and updates its embedding;

[0055] Among them, the information transfer mechanism in the graph convolutional network is carried out using the following formula:

[0056]

[0057] Among them, H (l) is the node embedding matrix of the, -th layer, H (0) =X, the initial node feature matrix, X is the energy efficiency parameter graph (matrix) of the residential building, is the regularized adjacency matrix, W (l) is the weight matrix of the, -th layer, σ(·) is the activation function.

[0058] S202: Aggregate the embeddings of all nodes into a graph embedding of a fixed dimension;

[0059] Specifically, use global pooling to aggregate the embeddings of the nodes into a single graph embedding:

[0060]

[0061] Among them, is the embedding of each node in the L-th layer, N is the number of nodes in the graph, h G is the global embedding of the graph.

[0062] S203: Use the aggregated graph embedding for graph classification, and the output classification result is the energy efficiency grade of the residential building.

[0063] Specifically, the graph embedding is used for graph classification through a fully connected layer and a classifier:

[0064] z = W f h G + b f (3)

[0065] y = softmax(z) (4)

[0066] where W f and b f are the weight matrix and bias vector of the fully connected layer, z is the classified output after mapping, and y is the class probability distribution normalized by softmax.

[0067] S3: Select test samples, and based on the residential building energy efficiency rating model, input the test sample data to obtain the corresponding residential building energy efficiency level evaluation results.

[0068] Refer to Figure 2 , in the example of the present invention, a residential building energy efficiency rating system based on a graph neural network is also disclosed, including:

[0069] A residential building energy efficiency parameter graph generation module, which is used to generate the topology of the residential building energy efficiency parameter graph based on the residential building parameters;

[0070] A residential building energy efficiency rating model establishment module, which is used to select training samples, use the residential building energy efficiency parameter graph among them as the input, and the residential building energy efficiency level as the output, and train based on the graph neural network to establish a residential building energy efficiency rating model;

[0071] A residential building energy efficiency level evaluation result output module, which is used to select test samples, and based on the residential building energy efficiency rating model, input the test sample data to obtain the corresponding residential building energy efficiency level evaluation results.

[0072] A residential building energy efficiency rating system based on a graph neural network in this example can execute a residential building energy efficiency rating method provided by the present invention, can execute any combination of implementation steps of the method example, and has the corresponding functions and beneficial effects of the method.

[0073] In order to verify the accuracy of the residential building energy efficiency rating method based on a graph neural network of the present invention, taking 107 residential buildings in the hot summer and cold winter regions of China as an example, data such as the building area, annual energy consumption of the heating system, annual energy consumption of the air conditioning system, annual energy consumption of the domestic hot water system, annual energy consumption of the lighting system, and annual energy consumption of the elevator system of these 107 residential buildings in a certain year are collected, and a residential building energy efficiency rating model is trained and established using the graph neural network algorithm.

[0074] It is proved by experiments that when using the residential building energy efficiency rating method based on a graph neural network proposed by the present invention for residential building energy efficiency evaluation, as shown in Table 2, it is completely consistent with the residential building energy efficiency results calculated by the "Technical Standard for Building Energy Efficiency Labeling" of the Ministry of Housing and Urban-Rural Development (shown in Table 3).

[0075] Table 2 List of residential building parameters

[0076]

[0077]

[0078] Table 4 List of Grading for Building Energy Efficiency Labels

[0079]

[0080] The above are only the embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A residential building energy efficiency rating method based on graph neural network, characterized in that: The following steps are involved: S1: Generate residential building energy efficiency parameter map topology based on residential building parameters; S2: Select training samples, take the residential building energy efficiency parameter map as input, and the residential building energy efficiency grade as output, and establish a residential building energy efficiency rating model based on graph neural network training; S3: Select test samples, input test sample data based on the residential building energy efficiency rating model, and obtain corresponding residential building energy efficiency grade evaluation results.

2. The residential building energy efficiency rating method based on graph neural network according to claim 1 is characterized in that: The residential building parameters include building area, annual energy consumption of heating system, annual energy consumption of air conditioning system, annual energy consumption of domestic hot water system, annual energy consumption of lighting system, annual energy consumption parameters of elevator system, and renewable energy power generation.

3. The residential building energy efficiency rating method based on graph neural network according to claim 1 is characterized in that: In step S2, the specific steps of establishing a residential building energy efficiency rating model based on graph neural network training include: S201: Based on the residential building parameter graph, the structural information and node features of the graph are embedded into the representation of each node through a multi-round message passing method. Through a series of graph convolutional layers, each node gradually aggregates the information of neighboring nodes and updates its embedding; S202: Aggregate the embeddings of all nodes into a graph embedding of fixed dimension; S203: Use the aggregated graph embedding to perform graph classification, and the output classification result is the energy efficiency grade of the residential building.

4. The residential building energy efficiency rating method based on graph neural network according to claim 3 is characterized in that: In step S201, the information transmission mechanism in the graph convolutional network is performed using the following formula: Among them, H (l) is the node embedding matrix of the ,th layer, H (0) =X, where X is the energy efficiency parameter diagram of residential buildings, is the regularized adjacency matrix, W (l) is the weight matrix of the,th layer, and σ(·) is the activation function.

5. The residential building energy efficiency rating method based on graph neural network according to claim 3 is characterized in that: In step S202, global pooling is used to aggregate the embeddings of the nodes into a single graph embedding: in, is the embedding of each node in the Lth layer, N is the number of nodes in the graph, and h G is the global embedding of the graph.

6. The residential building energy efficiency rating method based on graph neural network according to claim 3 is characterized in that: In step S203, the graph embedding is classified through the fully connected layer and the classifier: z=W f h G +b f (3) y=softmax(z) (4) Among them, W f and b f is the weight matrix and bias vector of the fully connected layer, z is the classified output after mapping, and y is the category probability distribution after normalization by softmax.

7. A residential building energy efficiency rating system based on graph neural network, characterized in that: include: A residential building energy efficiency parameter diagram generation module, used to generate a residential building energy efficiency parameter diagram topology based on residential building parameters; The residential building energy efficiency rating model establishment module is used to select training samples, take the residential building energy efficiency parameter map as input, and the residential building energy efficiency grade as output, and establish the residential building energy efficiency rating model based on graph neural network training; The residential building energy efficiency rating evaluation result output module is used to select test samples, input test sample data based on the residential building energy efficiency rating model, and obtain the corresponding residential building energy efficiency rating evaluation results.

8. The residential building energy efficiency rating system based on graph neural network according to claim 7 is characterized in that: The residential building parameters based on which the residential building energy efficiency parameter diagram generation module is generated include building area, annual energy consumption of the heating system, annual energy consumption of the air-conditioning system, annual energy consumption of the domestic hot water system, annual energy consumption of the lighting system, annual energy consumption parameters of the elevator system, and renewable energy power generation.

9. The residential building energy efficiency rating system based on graph neural network according to claim 7, characterized in that: The residential building energy efficiency rating model establishment module establishes the residential building energy efficiency rating model in the following specific steps: S201: Based on the residential building parameter graph, the structural information and node features of the graph are embedded into the representation of each node through a multi-round message passing method. Through a series of graph convolutional layers, each node gradually aggregates the information of neighboring nodes and updates its embedding; S202: Aggregate the embeddings of all nodes into a graph embedding of fixed dimension; S203: Use the aggregated graph embedding to perform graph classification, and the output classification result is the energy efficiency grade of the residential building.

Images