Traffic infrastructure full life cycle carbon emission assessment method based on knowledge graph

Abstract

The invention discloses a traffic infrastructure full life cycle carbon emission assessment method based on a knowledge graph. The method comprises the steps that knowledge is acquired through multi-source data acquisition, entities, attributes and relations related to carbon emission are defined, and knowledge graph ontology design and graph construction are carried out; establishing a carbon emission evaluation model based on an artificial neural network on a framework of the knowledge graph through data preprocessing, multi-layer perceptron network structure design and model training; and in combination with the carbon emission evaluation result and the optimization knowledge in the knowledge graph, generating a specific optimization strategy, and performing combinatorial optimization of the optimization strategy to formulate a recommendable optimization scheme for execution of an optimization decision. According to the scheme provided by the invention, carbon emission can be accurately quantified, a scientific and reasonable optimization strategy is generated, dynamic adjustment is supported, and a systematic and intelligent solution is provided for green and low-carbon development of traffic infrastructures and other engineering projects.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and carbon emission assessment, and particularly to a method for assessing the carbon emissions of the entire life cycle of transportation infrastructure based on a knowledge graph. Background Art

[0002] With the increasingly severe global environmental problems, energy conservation and emission reduction have become an important task in various fields, and the construction and building fields are no exception. During the entire life cycle of a building or structure related to an engineering project, from planning and construction to operation and maintenance until final demolition, a large amount of carbon emissions will be generated.

[0003] Traditional carbon emission assessment methods often have many problems. On the one hand, data is scattered in different sources and is difficult to integrate, making it difficult to comprehensively obtain the carbon emission information of the entire life cycle of a building or structure at the engineering project level, resulting in the lack of accuracy and integrity of the assessment results. For example, the material usage and energy consumption data in the construction stage may be independent of the energy management and maintenance data in the operation stage and cannot be effectively correlated. On the other hand, existing assessment models are usually relatively single and difficult to comprehensively consider the complex relationships between various factors, such as the synergistic effects between different materials and equipment, and the mutual constraints between strategies in different stages. In addition, existing optimization strategies are usually relatively general and lack consideration of the characteristics and actual needs of different engineering projects, making it difficult to provide specific and operable optimization measures. Summary of the Invention

[0004] To solve the technical problems, embodiments of the present invention provide a method for assessing the carbon emissions of the entire life cycle of transportation infrastructure based on a knowledge graph. Through knowledge graph technology, key information related to carbon emissions in each stage of the entire life cycle of an engineering project can be more comprehensively integrated, an advanced assessment model is used to achieve accurate quantitative assessment of carbon emissions, and a scientific, reasonable, and feasible optimization strategy is formulated based on the assessment results, thereby significantly reducing the carbon emissions of buildings or structures and strongly promoting the development of the engineering construction industry towards the direction of green, low-carbon, and sustainable.

[0005] The technical solution of the embodiments of the present invention is implemented as follows: Embodiments of the present invention provide a method for assessing the carbon emissions of the entire life cycle of transportation infrastructure, and the method includes: Obtain knowledge through multi-source data collection and define entities, attributes, and relationships related to carbon emissions for knowledge graph ontology design and graph construction; On the framework of the knowledge graph, through data preprocessing, multi-layer perceptron network structure design, and model training, a carbon emission assessment model based on an artificial neural network is established; Combine the carbon emission assessment results and the optimization knowledge in the knowledge graph to generate specific optimization strategies and perform combined optimization of the optimization strategies to formulate a recommendable optimization plan for the implementation of optimization decisions.

[0006] In one embodiment, the acquisition of knowledge through multi-source data collection includes: Collect data on all stages involved in the entire production process of buildings or structures; among them, all stages include the planning and design stage, the construction stage, the operation and maintenance stage, and the demolition and scrapping stage; the sources of the data include engineering documents, monitoring data, material databases, statistical reports, industry standards, technical specifications, and literature; the forms of the data include text, images, and tables.

[0007] In one embodiment, the ontology design and graph construction of the knowledge graph by defining entities, attributes, and relationships related to carbon emissions include: Extract entities and attributes and determine the relationships between entities to complete the ontology modeling of the knowledge graph; Use nodes to represent entities, edges to represent the relationships between entities, and add clear attributes and semantics to each node and edge to connect the nodes of each stage; Store the nodes and edges in a suitable graph database and use a visualization tool to display the structure of the knowledge graph.

[0008] In one embodiment, the data preprocessing includes data cleaning and data standardization; among them, the data cleaning includes removing noise, errors, and duplicate data in the collected data; the data standardization includes standardizing various types of data from different sources so that they have a unified dimension and numerical range; among them, numerical data uses the normalization method to map it to a specific interval; categorical data uses the one-hot encoding method for normalization processing.

[0009] In one embodiment, the design of the multi-layer perceptron network structure includes: Use multiple fully connected layers to construct the architecture of the multi-layer perceptron neural network; use each neuron to perform weighted summation on the input and introduce non-linearity through the activation function.

[0010] In one embodiment, the use of multiple fully connected layers to construct the architecture of the multi-layer perceptron neural network includes: Set an input layer to receive the input data and pass it to the hidden layer; Set a hidden layer composed of one or more layers of neurons to perform calculations and transform the input data; According to the task type, set the number of output layer nodes for the final output of the network.

[0011] In one embodiment, each neuron in the hidden layer performs a weighted sum of the inputs and introduces non-linearity through an activation function. The hidden layer is set to have 3 layers, each layer containing 128 neurons, and the activation function is ReLU. The output layer outputs the total carbon emissions over the entire life cycle and the carbon emissions at each stage: The functional formula of the activation function ReLU is: ReLU(x) = max(0, x); where x is the input value. If the input value is positive, the input value is output; otherwise, zero is output. The mathematical representation from the input layer to the first hidden layer is: h 1 = ReLU(W 1 X + b 1 ); where W 1 is a weight matrix of 128×32, b 1 is a bias vector of 128×1; X is the output of the input layer, and h 1 is the output of the first hidden layer; The mathematical representation from the first hidden layer to the second hidden layer is: h 2 = ReLU(W 2 h 1 + b 2 ); where W 2 is a weight matrix of 128×128, b 2 is a bias vector of 128×1, and h 2 is the output of the second hidden layer; The mathematical representation from the second hidden layer to the third hidden layer is: h 3 = ReLU(W 3 h 2 + b 3 ); where W 3 is a weight matrix of 128×128, b 3 is a bias vector of 128×1, and h 3 is the output of the third hidden layer; The mathematical representation from the third hidden layer to the output layer is: y = W 4 h3 + b 4 ; where W 4 is a weight matrix of 1×128, b 4 is a bias scalar, and y is the output of the output layer.

[0012] In one embodiment, the model training includes: Use a loss function to measure the error between the predicted result of the computational model output and the true label: ; Among them, is the error, is the true value, is the predicted value, and n is the number of samples; Use the backpropagation algorithm to calculate the gradient of the loss function with respect to the model parameters through the chain rule in calculus; Use the gradient descent algorithm to update the weight and bias parameters, so that the value of the loss function error gradually decreases: ; ; Among them, is the number of layers; is the learning rate, which determines the step size of each weight adjustment; is the gradient of the loss function with respect to the weight; is the gradient of the loss function with respect to the bias; is the weight before update, is the weight after update, is the bias parameter before update, is the bias parameter after update.

[0013] In one embodiment, combining the carbon emission assessment results and the optimization knowledge in the knowledge graph, generating specific optimization strategies and performing combinatorial optimization of the optimization strategies to formulate a recommendable optimization plan for the execution of the optimization decision, including: Set optimization goals according to the carbon emission assessment results; Locate high-carbon emission links through the knowledge graph and analyze carbon emission hotspots; Match the optimization strategies in the knowledge graph according to the carbon emission hotspots; Combine the optimization strategies to formulate multiple combinations of optimization strategies.

[0014] The solution of this embodiment has the following beneficial effects: In the embodiments of the present application, ontology design and graph construction of the knowledge graph are carried out through multi-source data collection, and then a carbon emission assessment model is established, and the assessment accuracy of the model is further improved through model training. Finally, based on the assessment results and knowledge graph technology, carbon emission optimization strategies for specific engineering projects are proposed. That is, by adopting knowledge graph technology, the present application can effectively integrate the scattered carbon emission-related data in each stage of the whole life cycle of engineering projects from multi-source heterogeneous data, learn and extract useful carbon emission optimization strategies, so as to form a complete data chain and provide comprehensive data support for accurate carbon emission assessment. Secondly, by using the established optimized carbon emission assessment model, the mutual relationship and influence of various factors can be comprehensively considered, and the key factors and control points of carbon emission can be analyzed more deeply, so as to more accurately reflect the actual situation of carbon emission. Finally, specific optimization strategies are provided according to different carbon emission situations for decision-making. This process realizes the intelligence and automation of the carbon emission optimization decision-making process, and improves the accuracy and efficiency of decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a schematic flow chart of a method for assessing carbon emissions in the whole life cycle of transportation infrastructure based on a knowledge graph according to an embodiment of the present invention; Figure 2 is a schematic flow chart of knowledge graph construction according to an embodiment of the present invention; Figure 3 is a schematic diagram of data flow of a multi-layer perceptron neural network architecture according to an embodiment of the present invention; Figure 4 is an internal structure diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0017] Embodiments of the present invention provide a method for assessing carbon emissions in the whole life cycle of transportation infrastructure based on a knowledge graph, as Figure 1 shown, the method includes: Step 101: Obtain knowledge and define entities, attributes, and relationships related to carbon emissions through multi-source data collection for ontology design and graph construction of the knowledge graph; Step 102: On the framework of the knowledge graph, establish a carbon emission assessment model based on an artificial neural network through data preprocessing, multi-layer perceptron network structure design, and model training; Step 103: Combine the carbon emission assessment results and the optimization knowledge in the knowledge graph to generate specific optimization strategies and perform combined optimization of the optimization strategies to formulate a recommendable optimization plan for the execution of optimization decisions.

[0018] Specifically, refer to Figure 2, in step 101, the knowledge graph construction includes knowledge acquisition and knowledge graph ontology design. Knowledge acquisition is achieved through multi-source data collection. The knowledge graph ontology design specifically involves extracting and determining entities, attributes, and relationships related to carbon emissions and constructing a knowledge graph framework.

[0019] Taking the transportation infrastructure industry as an example, the sources of the multi-source data collection include, but are not limited to, engineering documents, monitoring data, material databases, statistical reports, industry standards, technical specifications, and literature materials, etc. The data forms cover various types such as text, images, and tables. Knowledge is accurately extracted from the multi-source data, and through strict data cleaning and fine data annotation processes, noise data and error information are removed to ensure the accuracy and reliability of the knowledge.

[0020] Here, engineering documents refer to the engineering data documents collected for transportation infrastructure construction projects, including road planning documents, bridge design drawings, tunnel construction plans, etc. For example, in road construction, information such as the width of the road, the number of lanes, and the type of road surface material is obtained; parameters such as the structural form, span, and foundation type of the bridge are recorded. At the same time, construction process documents are collected, such as construction progress plans, construction quality inspection reports, and material procurement records. These documents can provide data such as energy consumption and material usage during the construction stage, providing basic data for carbon emission assessment.

[0021] Monitoring data refers to the sensor network installed on transportation infrastructure for real-time monitoring data collection. For example, traffic flow monitoring sensors installed on roads can obtain data such as the number of vehicle passages, vehicle speed, and vehicle types. These data are crucial for evaluating carbon emissions during the traffic operation stage because the driving state of vehicles directly affects their fuel consumption and carbon emissions. For structures such as bridges and tunnels, strain sensors, displacement sensors, etc. are installed to monitor the structural stress and deformation conditions during their operation. These data can be used to evaluate carbon emissions during the maintenance stage because the health status of the structure affects its maintenance frequency and maintenance measures.

[0022] The material database refers to the carbon emission coefficients of various building materials involved in transportation infrastructure construction. For example, carbon emission data of different grades of concrete, different types of steel, asphalt, etc. during production and transportation. These data can be obtained from product manuals provided by material suppliers, industry standards, and relevant academic literature. Particular attention is paid to information on new materials (such as high-performance concrete, carbon fiber reinforced composite materials, etc.), and these data provide a calculation basis for carbon emission optimization assessment.

[0023] Statistical reports cover statistical reports in the process of transportation infrastructure construction, operation, and maintenance. For example, statistical reports on material usage in the construction phase, including the usage amount and procurement cost of various materials; statistical reports on traffic flow in the operation phase, including traffic flow data at different time periods and different road sections; and repair record reports in the maintenance phase, including the number of repairs, repair content, and usage amount of repair materials.

[0024] Industry standards and technical specifications, such as road design specifications, bridge construction acceptance standards, etc. Some information such as calculation methods for carbon emissions and material performance requirements may be included in these standards, which can be used as a reference basis for carbon emission assessment.

[0025] Literature materials, that is, obtaining the latest research results on carbon emissions of transportation infrastructure. For example, research on the impact of different traffic flow control strategies on carbon emissions, research on the carbon emission characteristics of new transportation infrastructure materials, etc. These research results can provide theoretical support for the formulation of optimization strategies.

[0026] Defining entities, attributes, and relationships related to carbon emissions refers to a detailed description of the core entities, attributes, and relationships involved in the carbon emission knowledge graph of transportation infrastructure, covering the key elements and their logical associations in each stage of the entire life cycle (planning, construction, operation, maintenance, etc.). Specifically, Tables 1 and 2 are taken as examples: Table 1 Entity Classification and Attributes

[0027] Table 2 Core Relationships between Entities

[0028] Entities and attributes can be extracted through Tables 1 and 2, and the relationships between entities can be determined to achieve ontology modeling of the knowledge graph. This process can not only break data silos, associate scattered data such as building materials, construction techniques, and traffic flow into a unified network, but also closely connect each stage of the entire life cycle of transportation infrastructure to form a complete knowledge network. This knowledge graph framework provides a structured knowledge base for the accurate analysis and optimization of carbon emissions of transportation infrastructure and can be extended to specific engineering scenarios.

[0029] Knowledge graph construction refers to integrating the extracted entities, attributes, and relationships, and constructing a knowledge graph framework for the transportation infrastructure industry domain based on ontology representation methods. Entities are represented by nodes, relationships between entities are represented by edges, and clear attributes and semantics are added to each node and edge. Then, semantic relationships such as "phase association", "carbon emission contribution degree", and "energy consumption characteristics" are used to connect the nodes. Exemplarily, taking a highway project as the core node, connecting various factors related to carbon emissions in each phase (such as connecting "construction technology - dependent - construction machinery" and "carbon emission amount" through the relationship of "generated in the construction phase"), a knowledge network with clear levels is formed.

[0030] In the knowledge graph, each entity and relationship has clear definitions and identifications, forming a structured and semantic knowledge network to comprehensively and accurately describe the carbon emission factors related to transportation infrastructure and their mutual relationships. At the same time, the nodes and edges are stored in a suitable graph database, and a visualization tool is used to display the knowledge graph structure.

[0031] Specifically, this embodiment also provides typical reasoning examples based on graph reasoning as shown in Table 3: Table 3 Typical Reasoning Examples Based on Graph Reasoning

[0032] In step 102, step 102 may include: Step S121, data collection and preprocessing.

[0033] Data collection can collect the line planning scheme, structural design parameters, and material selection basis in the planning and design phase of transportation infrastructure; the construction process flow, mechanical models and energy consumption logs, material transportation routes, and carbon emission coefficients in the construction phase; the equipment operation power curve, traffic flow of vehicles, and maintenance operation frequency in the operation and maintenance phase; and the waste treatment process and recycling rate in the demolition and scrapping phase.

[0034] Data preprocessing includes data cleaning, that is, removing noise, errors, and duplicate data in the collected data. For example, in traffic flow data, correcting abnormal values caused by equipment failures; in energy consumption data, removing records that are clearly not in line with the actual situation. Data standardization, that is, standardizing various types of data from different sources so that they have a unified dimension and numerical range. For example, for numerical data, a normalization method can be used to map it to a specific interval (such as [0,1] or [-1,1]); for categorical data, one-hot encoding and other methods can be used for processing to facilitate subsequent use in the model.

[0035] The fact that numerical data can be mapped to a specific interval using a normalization method means that considering normalization can eliminate the influence of different feature dimensions, thereby accelerating the model convergence speed and improving the model performance. In the analysis of carbon emissions throughout the life cycle of transportation infrastructure, common numerical influencing factors include construction scale (unit: square meters), operation energy consumption (unit: kilowatt-hours), transport turnover (unit: ton-kilometers), etc. When performing normalization, the following formula can be used to linearly map the data to the interval [0,1]: (1) where is the original data, and are the minimum and maximum values in this feature data respectively, is the normalized data. If the data is to be mapped to the interval [-1,1], the following formula can be used: (2) The data after normalization can be used as the node value of the input layer of the model. For example, when constructing a neural network model, these normalized data can be directly input into the input layer to help the model learn the relationship between different features and carbon emissions.

[0036] Categorical data can be processed using one-hot encoding for subsequent use in the model. Among the influencing factors of carbon emissions throughout the life cycle of transportation infrastructure, there is a non-numerical data type used to represent different categories or attributes of things. In machine learning models, most algorithms require the input data to be numerical, so categorical data needs to be appropriately processed to convert it into a numerical form suitable for input into the machine learning model. For each category in the original categorical feature, the value in the corresponding one-hot encoded feature is 1, and the values of the remaining features are 0.

[0037] Exemplarily, the transportation mode is a typical categorical variable, and common ones include highway, railway, aviation, waterway, etc. Assuming only these four transportation modes are considered, when performing one-hot encoding, four new binary features will be created, as shown in Table 4. The type of construction material is also an important categorical factor affecting carbon emissions, such as concrete, steel, wood, etc. Assuming these three materials are considered, three new binary features will be created after one-hot encoding, as shown in Table 5.

[0038] Table 4 Example of influencing factors of transportation mode and one-hot encoding

[0039] Table 5 Example of influencing factors of construction material type and one-hot encoding

[0040] After one-hot encoding, these categorical data can be used as independent numerical features and input into the input layer of the machine learning model in step S122. For example, in a simple linear regression model, assuming that in addition to the above categorical factors, there are other numerical influencing factors (such as construction scale, operation time, etc.), all features can be combined into a feature matrix as the input of the model.

[0041] Step S122, network structure design.

[0042] The network structure design uses multiple fully connected layers to construct a multi-layer perceptron neural network architecture. Each neuron performs a weighted sum on the input and introduces non-linearity through an activation function.

[0043] Specifically, the multi-layer perceptron is at least divided into three layers. The first layer is the input layer, the last layer is the output layer, and the middle is the hidden layer. Multiple layers can be built as needed, and each layer can have multiple neurons or nodes. These neurons or nodes are arranged in a hierarchical structure, and each node in adjacent layers is connected to each other.

[0044] Among them, the input layer is used to receive input data and transfer it to the hidden layer. The number of neurons in the input layer is equal to the number of input features.

[0045] Exemplarily, the number of input layer nodes is designed according to the carbon emission influencing factors involved in the knowledge graph. For example, for the carbon emission assessment of the whole life cycle of transportation infrastructure, the input layer may include the influencing parameters in the planning and design stage of transportation infrastructure, the generation parameters in the construction stage, the involved parameters in the operation and maintenance stage, and the associated parameters in the demolition and scrapping stage. Further, other input feature data such as energy consumption parameters (such as the usage amounts of different energies) and environmental parameters (such as local carbon emission factors, average temperature, etc.) can be considered to be included in the input layer node design as needed. The number of these parameters determines the number of nodes in the input layer. That is to say, the number of nodes in the input layer depends on the number of carbon emission influencing factors involved in the knowledge graph. The following is an example of the input layer node number design based on the carbon emission influencing factors of the whole life cycle of transportation infrastructure projects. See Table 6 for details.

[0046] Table 6 Example of input layer node design

[0047] Add up the number of nodes in the above-mentioned stages, that is, 6 (planning and design) + 9 (construction) + 7 (operation and maintenance) + 5 (demolition and scrapping) + 5 (environment and external factors) = 32 nodes. Finally, the total number of input layer nodes is obtained and represented by an input feature vector: (3) Among them, x1 to x32 respectively correspond to the example values of the influencing factor characteristics corresponding to the node numbers in Table 6.

[0048] Through the above design, the number of input layer nodes covers the carbon emission influencing factors in each stage of the whole life cycle of transportation infrastructure, ensuring the comprehensiveness and accuracy of the model. This design provides a scientific data input basis for the carbon emission assessment model based on artificial neural network.

[0049] Furthermore, the hidden layer in the embodiment is composed of one or more layers of neurons, which is used to perform calculations and transform the input data, and a multi-layer perceptron structure is constructed by using multiple fully connected layers. This connection method can fully capture the complex relationships between input features, such as Figure 3 shown. The number of hidden layers and the number of neurons in each hidden layer are determined according to the complexity of the problem and the scale of the data. For example, 2 to 3 hidden layers can be used, and each hidden layer contains 32 to 128 neurons. The activation function of the hidden layer can be selected as ReLU, which can effectively solve the problem of gradient disappearance and accelerate the training speed of the network.

[0050] Exemplarily, after designing the 32 nodes of the input layer (corresponding to the carbon emission influencing factors in the whole life cycle of transportation infrastructure, that is, the input feature vector X) in the above embodiment, the hidden layer structure is further designed. Specifically, 3 hidden layers are used, and the selection basis is that it can better capture the complex non-linear relationships between input features and avoid overfitting at the same time; the number of neurons in each layer is 128, and the selection basis is that it can control the computational complexity while ensuring the performance of the model. Among them, the number of hidden layer neurons is usually proportional to the number of input layer nodes. For example, if the number of neurons is too small, the model may not be able to capture the complex relationships between input features, resulting in underfitting; if the number of neurons is too large, the model may overfit and the computational cost increases.

[0051] The function formula of the activation function ReLU is as follows: ReLU(x)=max(0,x) (4) Among them, if the input value is positive, the input value is output; otherwise, zero is output.

[0052] Specifically, the mathematical representation from the input layer to the hidden layer 1 (i.e., the first hidden layer) is as follows: h 1 =ReLU(W 1 X + b 1 ) (5) Among them, W 1 is the weight matrix (128×32), and b 1 is the bias vector (128×1).

[0053] The mathematical representation from hidden layer 1 to hidden layer 2 (i.e., the second hidden layer) is as follows: h 2 = ReLU(W 2 h 1 + b 2 ) (6) where W 2 is the weight matrix (128×128), and b 2 is the bias vector (128×1).

[0054] The mathematical representation from hidden layer 2 to hidden layer 3 (i.e., the third hidden layer) is as follows: h 3 = ReLU(W 3 h 2 + b 3 ) (7) where W 3 is the weight matrix (128×128), and b 3 is the bias vector (128×1).

[0055] The mathematical representation from hidden layer 3 to the output layer is as follows: y = W 4 h3 + b 4 (8) where W 4 is the weight matrix (1×128), and b 4 is the bias scalar. W 4 can be expressed as: (9) According to Figure 3 the data flow example diagram provided in 1 , each connection line between nodes in adjacent layers described by the weight matrix W contains different weights, indicating the importance of the corresponding nodes in the previous layer. Generally, it is saved in matrix form, and there is also a bias b between adjacent layers, which is generally saved in vector form. Taking the input layer to hidden layer 1 as an example, if the element in the i-th row and j-th column of the matrix W ij is w (10) where i = 1, 2, …, 128 represents the row index of the matrix; j = 1, 2, …, 32 represents the column index of the matrix. In the weight matrix W 1 , use to represent the weight of the connection between the first node in the first layer and the second node in the 0th layer (input layer). That is, each element w ij in W is a real number, representing the connection weight from the j-th dimension of the input feature to the i-th dimension of the output feature. The corresponding bias b1 It can be expressed as: (11) Similarly, for the above-mentioned W 2 , W 3 , b 2 and b 3 can also refer to the above mathematical representation.

[0056] The output layer is used for the final output of the network, such as the regression target or classification label. The number of neurons in the output layer depends on the specific data. For example, in the carbon emission assessment of the whole life cycle of transportation infrastructure, the number of nodes in the output layer can be set to 1 or more according to the regression task, depending on the number of regression targets. A single output is the total carbon emission of the whole life cycle; multiple outputs are the carbon emissions in each stage. In the classification task, the number of nodes in the output layer is equal to the number of classes. For binary classification, there are 2 nodes (such as low carbon / high carbon); for multi-class classification, there are multiple nodes (such as carbon emission levels: low, medium, high).

[0057] Specifically, in the regression task, if there is 1 node set in the output layer, the total carbon emission of the whole life cycle uses a linear activation function (no activation function) to perform matrix multiplication operations and activation processing layer by layer according to formulas (3) to (9), and finally a single real value is obtained. If there are 2 nodes set in the output layer, that is, the total carbon emission of the whole life cycle and the proportion of carbon emissions in each stage (such as the proportion of carbon emissions in the planning and design, construction, operation and maintenance, and demolition and scrapping stages). Then the weight matrix W 4 (2×128) and the bias vector b 4 (2×1) can be expressed as: (12) (13) Among them, node 1 (total carbon emission of the whole life cycle) also uses a linear activation function, while node 2 (proportion of carbon emissions in each stage) uses the Softmax activation function for calculation to ensure that the sum of the proportions of carbon emissions in each stage is 1. The relevant calculation processes of node 1 and node 2 are carried out according to formulas (14) and (15): (14) (15) Exemplarily, through formulas (3) to (7), formula (12), formula (13), and formula (14), the output layer can output a single real value for Node 1, that is, the total carbon emissions over the entire life cycle is 8,000 tons of CO2e; through formulas (3) to (7), formula (12), formula (13), and formula (15), the output layer can output the probability distribution of Node 2, that is, the proportion of carbon emissions in each stage is 10% in the planning and design stage, 60% in the construction stage, 25% in the operation and maintenance stage, and 5% in the demolition and scrapping stage.

[0058] Furthermore, if the specific requirement of the output layer is a classification task type, and the number of output layer nodes is 3 (assuming three carbon emission levels: low, medium, and high), and the activation function Softmax is used, the mathematical expression of the output layer is: (16) where, W 4 is the weight matrix (3×128), and b 4 is the bias vector (3×1). Combining formulas (3) to (7), the probability distribution result of the classification task can be obtained (exemplarily, the probability of the low carbon emission level is 0.2, the probability of the medium carbon emission level is 0.5, and the probability of the high carbon emission level is 0.3).

[0059] Through the above steps, the input data is passed to the input layer of the neural network. Each neuron in the hidden layer will receive the output of the previous layer, perform weighted summation on it, and process it through the activation function, thereby generating the output of the hidden layer. These outputs of the hidden layer are then passed to the next layer until the data reaches the output layer, and finally the predicted output is obtained. In step S122, this entire process is called forward propagation (that is, calculating the output of each layer). In the multi-layer perceptron model, the weights W and biases b contained in the connections between layers are variables. These parameters are learned during the trial calculation and training process using a large number of known samples. By repeatedly adjusting the weights and biases of each layer, the final determined values of the weights and biases of each layer are obtained to minimize the difference between the network prediction and the actual target value.

[0060] Step S123, model training.

[0061] Model training includes calculating the loss, backpropagating the error, and updating the parameters.

[0062] The calculation of the loss refers to calculating the loss of the model on the training data after obtaining the output result. Usually, a loss function is used to measure the error between the prediction result and the true label. In this embodiment, the mean squared error (MSE) is selected as the loss function, and the specific calculation formula is as follows: 2 (17) where, is the true value, is the predicted value, and n is the number of samples.

[0063] The backpropagation error refers to calculating the gradient of the loss function with respect to the model parameters using the chain rule in calculus through the backpropagation algorithm. Specifically, starting from the output layer, the gradient of each parameter is calculated layer by layer, and then the gradient is propagated backward along the network until the gradient of the input layer parameters is calculated.

[0064] Specifically, taking the output layer as having 1 node, the weight matrix W 4 , and the bias scalar b 4 , and the linear activation function as an example, matrix multiplication operations and activation processing are performed layer by layer according to formulas (3) to (9), and finally a single real value is obtained, that is, the predicted value of the total carbon emissions over the entire life cycle = 7900 tons of CO2e. Assume the true value = 8000 tons of CO2e, and backpropagation is performed according to formula (17) to calculate the mean squared error (MSE = 10000). The direction in which the total error of 10000 decreases fastest should be the negative gradient.

[0065] Therefore, further, the output layer gradients include the gradient of the loss function with respect to the output of the output layer: (18) Output layer weight gradient: (19) Output layer bias gradient: (20) Further, the hidden layer 3 gradients include the gradient of the loss function with respect to the output of hidden layer 3: (21) Hidden layer 3 weight gradient: (22) Hidden layer 3 bias gradient: (23) Further, the hidden layer 2 gradients include the gradient of the loss function with respect to the output of hidden layer 2: (24) Hidden layer 2 weight gradient: (25) Hidden layer 2 bias gradient: (26) Further, the hidden layer 1 gradient includes the gradient of the loss function with respect to the output of hidden layer 1: (27) Hidden layer 1 weight gradient: (28) Hidden layer 1 bias gradient: (29) All the data in the above equations have been calculated during the forward propagation and are known numbers. Thus, these partial derivatives can be used to adjust the weights and biases. During the backpropagation process, the gradient descent algorithm is used to update the parameters, making the value of the loss function gradually decrease. The update formulas for the weights and biases of each layer can be expressed as: (30) (31) Where, is the number of layers; is the learning rate, which determines the step size of each weight adjustment; is the gradient of the loss function with respect to the weight; is the gradient of the loss function with respect to the bias.

[0066] Through the above specific implementation, it shows how to calculate the gradient of each parameter layer by layer starting from the output layer and backpropagate the gradient along the network until the gradient of the input layer parameters is calculated. By the gradient descent method, the weights W and biases b are adjusted layer by layer and node by node, finally achieving the accurate assessment of the carbon emissions throughout the life cycle of transportation infrastructure. That is, using the training set data to train the constructed artificial neural network model. When the value of MSE is smaller, it indicates that the prediction result of the model is closer to the real observed data, thus indicating that the model has higher prediction accuracy. During the training process, the input data is fed into the network, the predicted value is obtained through forward propagation, the loss is calculated based on the predicted value and the actual value, and then the weights and biases of the network are adjusted using the optimization algorithm to reduce the loss value. Repeat this process until the performance of the model on the validation set no longer improves or reaches the preset number of training epochs.

[0067] However, as described above, although the implementation method of backpropagation has been explained. But in actual operation, these are far from enough, and there are many in-depth details that need to be mastered, such as the selection of activation functions, the parameter initialization method, the determination of the learning rate, and how to iterate using batch samples, etc.

[0068] After completing the data collection and preprocessing in step S121, the network structure design in step S122, and the model training in step S123, the establishment of the carbon emission assessment model based on the artificial neural network in step S102 of this embodiment is completed.

[0069] In step 103, combining the carbon emission assessment results and the optimization knowledge in the knowledge graph, specific optimization strategies are generated and the combined optimization of the optimization strategies is carried out to formulate a recommendable optimization plan for the implementation of the optimization decision: First, according to the carbon emission assessment results, optimization goals are set. For example, for the carbon emission goal: reducing the total carbon emissions in the whole life cycle; for the economic goal: controlling the optimization cost within an acceptable range; for the engineering goal: ensuring the technical feasibility and operability of the optimization plan, etc. Further, according to the types of optimization goals, the corresponding target values are determined. For example, for the carbon emission target value: reducing the total carbon emissions in the whole life cycle by 15%; for the economic target value: the increase in the optimization cost does not exceed 10%; for the engineering target value: the implementation period of the optimization plan does not exceed 3 months, etc.

[0070] Second, high carbon emission links (such as construction machinery energy consumption, material transportation distance, etc.) are located through the knowledge graph, and carbon emission hotspots are analyzed; according to the carbon emission hotspots, optimization strategies in the knowledge graph are matched, such as material substitution (using low-carbon materials, such as recycled concrete, low-carbon steel), process optimization (improving construction processes, such as using energy-saving equipment, optimizing construction processes), energy structure adjustment (using clean energy, such as solar energy, wind energy), and operation management optimization (optimizing traffic flow management, such as intelligent transportation systems), etc. The exemplary generated optimization strategies are shown in Table 7; further, multiple combinations of optimization strategies are formulated (such as material substitution + process optimization, process optimization + energy structure adjustment, etc. combination strategies), and the exemplary optimization strategy combinations and strategy evaluations can be seen in Table 8.

[0071] Table 7 Specific Optimization Strategies

[0072] Table 8 Strategy Combination Types and Optimization Plans

[0073] Specifically, the whole life cycle assessment and optimization effects of the traffic infrastructure project in this embodiment are shown in Table 9: Table 9 Embodiment Assessment and Optimization Effects

[0074] Through the above specific implementation manners, in combination with the carbon emission assessment results and the optimization knowledge in the knowledge graph, an optimization strategy is formulated and used for decision-making execution. This method can comprehensively consider carbon emissions, economic costs, and engineering feasibility, providing a scientific basis for the green and low-carbon construction of transportation infrastructure.

[0075] In summary, based on the above steps, the embodiment of the present application designs and constructs the knowledge graph ontology by collecting multi-source data and defining entities, attributes, and relationships related to carbon emissions; based on the framework of the knowledge graph, through data preprocessing, multi-layer perceptron network structure design, and model training, a carbon emission assessment model based on an artificial neural network is established; in combination with the carbon emission assessment results and the optimization knowledge in the knowledge graph, specific optimization strategies are generated and the combination optimization of the optimization strategies is carried out to formulate a recommended optimization plan for optimizing decision-making execution. That is, the present application integrates carbon emission-related information in all stages of the whole life cycle by constructing a knowledge graph and uses a multi-layer perceptron (MLP) neural network model to achieve accurate quantitative assessment of carbon emissions. This method can not only comprehensively analyze the main sources and key influencing factors of carbon emissions, but also generate scientific and reasonable optimization strategies based on the assessment results, covering multiple aspects such as material selection, construction process improvement, operation management optimization, and energy structure adjustment. Through personalized recommendation, users can select the most suitable optimization plan according to their own needs, thereby significantly reducing the carbon emissions of transportation infrastructure and promoting the transportation industry to develop in a green, low-carbon, and sustainable direction. The present invention provides a systematic and intelligent solution for the carbon emission management and optimization of transportation infrastructure and has important practical application value.

[0076] For the engineering project of building construction, the data collection, model construction, carbon emission calculation, optimization strategy formulation, and final optimization plan recommendation of the whole life cycle knowledge graph can also be carried out according to the above steps. This method can not only comprehensively quantify the carbon emissions of the whole life cycle of building construction, identify high-carbon emission links, but also generate low-carbon optimization strategies covering multiple aspects such as materials, processes, energy, and management, and recommend the optimal plan according to user needs to balance carbon emissions, economic costs, and engineering feasibility. Through accurate assessment and scientific optimization, the carbon emissions of the whole life cycle of building construction are significantly reduced.

[0077] It can be seen from this that the method provided by this embodiment is not only applicable to structures such as transportation infrastructure, but can also be widely applied to building construction products, providing strong technical support for the green and low-carbon transformation of the construction industry.

[0078] To implement the method of the embodiments of the present invention, the embodiments of the present invention further provide a traffic infrastructure full-life-cycle carbon emission assessment system based on a knowledge graph, including: a processor and a memory for storing a computer program capable of running on the processor; wherein, when the processor is used to run the computer program, it executes the steps of the above-mentioned method.

[0079] The above system provided in this embodiment and the above method embodiment belong to the same concept. For the specific implementation process, please refer to the method embodiment and will not be elaborated here.

[0080] To implement the method of the embodiments of the present invention, the embodiments of the present invention further provide a computer program product. The computer program product includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps of the above-mentioned method.

[0081] Based on the hardware implementation of the above program module, and to implement the method of the embodiments of the present invention, the embodiments of the present invention further provide an electronic device (computer device). Specifically, in one embodiment, the computer device may be a terminal, and its internal structure diagram may be as Figure 4 shown. The computer device includes a processor A01, a network interface A02, a display screen A04, an input device A05 and a memory (not shown in the figure) connected through a system bus. Among them, the processor A01 of the computer device is used to provide computing and control capabilities. The memory of the computer device includes an internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A06. The network interface A02 of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor A01, it implements the method of any one of the above embodiments. The display screen A04 of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device A05 of the computer device may be a touch layer covered on the display screen, or a button, a trackball or a touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0082] Those skilled in the art can understand that Figure 4 the structure shown in

[0083] The device provided by the embodiment of the present invention includes a processor, a memory, and a program stored on the memory and executable on the processor. When the processor executes the program, the method of any one of the above embodiments is implemented.

[0084] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0085] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0086] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0087] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0088] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.

[0089] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0090] Computer-readable media includes both permanent and non-permanent, removable and non-removable media implemented by any method or technology for storing information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile discs (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0091] It can be understood that the memory in the embodiments of the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM, Read-Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read-Only Memory), an electrically erasable programmable read-only memory (EEPROM, Electrically Erasable Programmable Read-Only Memory), a ferromagnetic random access memory (FRAM, ferromagnetic random access memory), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM, Compact Disc Read-Only Memory); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM, Random Access Memory), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as a static random access memory (SRAM, Static Random Access Memory), a synchronous static random access memory (SSRAM, Synchronous Static Random Access Memory), a dynamic random access memory (DRAM, Dynamic Random Access Memory), a synchronous dynamic random access memory (SDRAM, Synchronous Dynamic Random Access Memory), a double data rate synchronous dynamic random access memory (DDR SDRAM, Double Data Rate Synchronous Dynamic Random Access Memory), an enhanced synchronous dynamic random access memory (ESDRAM, Enhanced Synchronous Dynamic Random Access Memory), a sync link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), and a direct rambus random access memory (DRRAM, Direct Rambus Random Access Memory).The memories described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memories.

[0092] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or apparatus. Without further limitation, an element defined by the phrase "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising the element.

[0093] The above are only the embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A method for evaluating the carbon emissions of the entire life cycle of transportation infrastructure based on a knowledge graph, characterized in that, The method includes: Obtaining knowledge through multi-source data collection, defining entities, attributes, and relationships related to carbon emissions, and conducting knowledge graph ontology design and graph construction; On the framework of the knowledge graph, establishing a carbon emission assessment model based on an artificial neural network through data preprocessing, multi-layer perceptron network structure design, and model training; Combining the carbon emission assessment results and the optimized knowledge in the knowledge graph, generating specific optimization strategies, and conducting combinatorial optimization of the optimization strategies to formulate a recommendable optimization plan for the implementation of optimization decisions.

2. The method for evaluating the carbon emissions of the whole life cycle of transportation infrastructure based on the knowledge graph according to claim 1, characterized in that The obtaining knowledge through multi-source data collection includes: Collecting data for all stages involved in the entire production process of buildings or structures; among them, all stages include the planning and design stage, the construction stage, the operation and maintenance stage, and the demolition and scrapping stage; the sources of the data include engineering documents, monitoring data, material databases, statistical reports, industry standards, technical specifications, and literature materials; the forms of the data include text, images, and tables.

3. The method for evaluating the carbon emissions of the whole life cycle of transportation infrastructure based on the knowledge graph according to claim 1, characterized in that, The defining entities, attributes, and relationships related to carbon emissions and conducting knowledge graph ontology design and graph construction includes: Extracting entities and attributes and determining the relationships between entities to complete the knowledge graph ontology modeling; Using nodes to represent entities, edges to represent the relationships between entities, adding clear attributes and semantics to each node and edge, and connecting the nodes of each stage; Storing the nodes and edges in a suitable graph database and using a visualization tool to display the knowledge graph structure.

4. The method for evaluating the carbon emissions throughout the life cycle of transportation infrastructure based on a knowledge graph according to claim 1, characterized in that, The data preprocessing includes data cleaning and data standardization; among them, the data cleaning includes removing noise, errors, and duplicate data in the collected data; the data standardization includes standardizing various types of data from different sources to make them have a unified dimension and numerical range; among them, numerical data uses the normalization method to map it to a specific interval; categorical data uses the one-hot encoding method for normalization processing.

5. The method for evaluating the carbon emissions throughout the life cycle of transportation infrastructure based on a knowledge graph according to claim 1, wherein The multi-layer perceptron network structure design includes: Constructing a multi-layer perceptron neural network architecture using multiple fully connected layers; using each neuron to perform a weighted sum of the input and introducing non-linearity through an activation function.

6. The method for evaluating the carbon emissions throughout the life cycle of transportation infrastructure based on a knowledge graph according to claim 5, wherein The constructing a multi-layer perceptron neural network architecture using multiple fully connected layers includes: Setting an input layer to receive input data and transfer it to the hidden layer; Setting a hidden layer composed of one or more layers of neurons to perform calculations and transform the input data; According to the task type, setting the number of output layer nodes for the final output of the network.

7. The method for evaluating the carbon emissions of the whole life cycle of transportation infrastructure based on the knowledge graph according to claim 6, characterized in that, Each neuron in the hidden layer performs a weighted sum of the input and introduces non-linearity through an activation function, including setting 3 layers in the hidden layer, each layer containing 128 neurons, the activation function being ReLU, and the output layer outputting the total carbon emissions over the entire life cycle and the carbon emissions of each stage: The function formula of the activation function ReLU is: ReLU(x)=max(0,x); where x is the input value, if the input value is positive, the input value is output; otherwise, zero is output; The mathematical representation from the input layer to the first hidden layer is: h 1 = ReLU(W 1 X + b 1 ); Among them, W 1 is a weight matrix of 128×32, and b 1 is a bias vector of 128×1; X is the output of the input layer, and h 1 is the output of the first hidden layer; The mathematical representation from the first hidden layer to the second hidden layer is: h 2 = ReLU(W 2 h 1 + b 2 ); Among them, W 2 is a weight matrix of 128×128, b 2 is a bias vector of 128×1, h 2 is the output of the second hidden layer; The mathematical representation from the second hidden layer to the third hidden layer is: h 3 = ReLU(W 3 h 2 + b 3 ); Among them, W 3 is a weight matrix of 128×128, b 3 is a bias vector of 128×1, and h 3 is the output of the third hidden layer; The mathematical representation from the third hidden layer to the output layer is as follows: y = W 4 h3 + b 4 ; Among them, W 4 is a 1×128 weight matrix, b 4 is a bias scalar, and y is the output of the output layer.

8. The method for evaluating the carbon emissions of the whole life cycle of transportation infrastructure based on the knowledge graph according to claim 1, characterized in that, The model training includes: Using a loss function to measure the error between the predicted result of the model output and the true label: ； wherein, is the error, is the true value, is the predicted value, and n is the number of samples; Calculating the gradient of the loss function with respect to the model parameters by using the chain rule in calculus through the backpropagation algorithm; Using the gradient descent algorithm to update the weight and bias parameters so that the value of the loss function error gradually decreases: ； ； Among them, is the number of layers; is the learning rate, which determines the step size of each weight adjustment; is the gradient of the loss function with respect to the weights; is the gradient of the loss function with respect to the bias; is the weight before update, is the weight after update, is the bias parameter before update, is the bias parameter after update.

9. The method for evaluating the carbon emissions throughout the life cycle of transportation infrastructure based on a knowledge graph according to claim 1, wherein Combining the carbon emission assessment results and the optimization knowledge in the knowledge graph to generate specific optimization strategies and perform combinatorial optimization of the optimization strategies to formulate a recommendable optimization plan for the execution of the optimization decision, including: Setting optimization goals according to the carbon emission assessment results; Locating high-carbon emission links through the knowledge graph and analyzing carbon emission hotspots; Matching the optimization strategies in the knowledge graph according to the carbon emission hotspots; Formulating multiple combinations of optimization strategies in combination with the optimization strategies.

Images