Electricity-carbon coupling modeling method based on neural network

Abstract

The invention discloses an electricity-carbon coupling modeling method based on a neural network, which relates to the technical field of emission control and carbon reduction, and comprises the following steps of: acquiring data, performing cleaning, aligning and standardized data preprocessing, and learning and simulating a complex nonlinear coupling relationship among power consumption, energy use and carbon emission through a time sequence diagram neural network. A dynamic coupling relation model between power data and carbon emission data is established, deep learning and physical rules are combined, and multi-source heterogeneous data are fused and analyzed through multi-source data fusion, space-time coupling modeling and fusion, so that the analysis result is more comprehensive and accurate, the carbon emission data are accurately predicted, and the one-sidedness of single data is prevented.

Description

Technical Field

[0001] This invention relates to the field of emission control and carbon reduction technology, and in particular to a neural network-based electro-carbon coupling modeling method. Background Technology

[0002] Power system carbon emission monitoring is a technical means to collect, analyze and manage the emissions of greenhouse gases such as carbon dioxide generated during the production, transmission and consumption of electricity in real time. It is usually achieved by deploying carbon emission sensors, collecting equipment operation data and fuel consumption, and using big data analysis, the Internet of Things and artificial intelligence to assist in the analysis.

[0003] The current power system contains multi-source heterogeneous data. Traditional methods are difficult to integrate multi-source heterogeneous data on power, energy and production, and there is interference from abnormal data. They cannot effectively capture the spatial dependence and time lag effects between equipment, resulting in low accuracy of carbon emission prediction. Furthermore, traditional methods only consider single-point or linear relationships, so the prediction results are relatively one-sided. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a neural network-based electrocarbon coupling modeling method to solve the problem that there are multiple heterogeneous data sources in the power system, and traditional methods are difficult to integrate the multiple heterogeneous data sources of electricity, energy and production.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a neural network-based method for modeling electro-carbon coupling, comprising, Collect multi-source heterogeneous data on power production, preprocess the multi-source heterogeneous data, and obtain a cleaned and aligned time-series dataset; Correlation analysis was performed based on time-series datasets to construct an electrocarbon coupling graph structure that varies with time and device status, and dynamic graph data was obtained. Correlation analysis was performed based on dynamic graph data, and spatiotemporal graph neural network modeling was carried out by integrating physical information through neural differential equations to obtain an electrocarbon coupling relationship model. The parameters of the electro-carbon coupling model were optimized and verified, and a verified model file was generated. Deploy the model files to the production environment to dynamically monitor carbon emissions.

[0007] As a preferred embodiment of the neural network-based electro-carbon coupling modeling method of the present invention, the method comprises: collecting multi-source heterogeneous data based on power, energy, and carbon monitoring devices, wherein the multi-source heterogeneous data includes power data, energy data, production data, and direct monitoring data; the power data includes real-time power consumption, power, and load curves of the equipment; the energy data includes fossil fuel consumption, lower heating value, elemental carbon content, and carbon oxidation rate; the production data includes power generation, heat supply, equipment operating hours, and load rate; and the direct monitoring data includes real-time monitoring data of carbon dioxide concentration and flow rate. The multi-source heterogeneous data is cleaned using the local outlier factor method to identify and remove sensor fault data, transmission error data, and abnormal operating condition fluctuation data, obtaining cleaned data; the cleaned data is then aligned to the same time-series granularity using a time alignment method, obtaining aligned data; and finally, the aligned data is standardized using a standardization method to eliminate dimensional differences, obtaining a cleaned and aligned time-series dataset.

[0008] As a preferred embodiment of the neural network-based electro-carbon coupling modeling method of the present invention, the following steps are taken: A node definition method is used to partition the time-series dataset into nodes. A feature vector is extracted from each node, comprising static features and dynamic time-series features. These feature vectors are then integrated to generate a node feature set. The static features include the rated power and design efficiency of the equipment, while the dynamic time-series features include real-time power, load rate, and energy consumption per unit output. An edge definition method is used to partition the energy flow, material flow, and carbon flow relationships between nodes into edges. An initial weight is defined for each edge, and these initial weights are then integrated to generate an initial edge weight set. A dynamic evolution method is used to perform correlation analysis on the initial edge weight set. Dynamic weights are obtained through an edge weight evolution formula, and these dynamic weights are then integrated to generate a dynamic edge weight set. Finally, a graph structure integration method is used to combine the node feature set and the dynamic edge weight set to generate a dynamic graph structure. This dynamic graph structure includes a node feature matrix and a dynamic adjacency matrix. The dynamic graph data is then generated by integrating these dynamic graph structures.

[0009] As a preferred embodiment of the neural network-based electro-carbon coupling modeling method of the present invention, the following steps are taken: Spatial information aggregation of dynamic graph data is performed using the graph convolution module of a temporal graph neural network; the features of each node and its neighboring nodes are processed by the graph convolution network to capture the spatial dependencies between devices, resulting in spatially aggregated node embeddings; the temporal modeling module of the temporal graph neural network performs time-series processing on the node embeddings; the continuous time changes of node states are modeled using neural differential equations to generate temporally enhanced node states; correlation analysis of node states is performed using the physical constraint integration method, with the energy conservation equation as a constraint condition, to generate node outputs; the node outputs are integrated, and the total carbon emissions are calculated using a fully connected network to obtain the electro-carbon coupling relationship model.

[0010] As a preferred embodiment of the neural network-based electro-carbon coupling modeling method of the present invention, the time series dataset and dynamic graph data are divided into a training set, a validation set and a test set. Based on the training set, validation set and test set, the electro-carbon coupling relationship model is trained. The parameters of the electro-carbon coupling relationship model are optimized by the root mean square error. The weight parameters of the electro-carbon coupling relationship model are iteratively updated by the backpropagation algorithm and the optimizer.

[0011] As a preferred embodiment of the neural network-based electro-carbon coupling modeling method of the present invention, the trained electro-carbon coupling relationship model is solidified into a model file, the model file is deployed to the production environment, carbon emission data is monitored in real time based on the model file of the electro-carbon coupling relationship model, and prediction results are output. The prediction results are displayed, and carbon emission data and equipment monitoring charts are generated.

[0012] In a preferred embodiment of the neural network-based electro-carbon coupling modeling method described in this invention, the weight of each edge is recalculated using an edge weight evolution formula, the formula being: ,in, This is a device health factor used to reflect the rate at which the connection strength between devices decays over time. Used to reflect the dynamic weight between node i and node j at time t. The weights between nodes i and j are used to reflect the initial weights, and t is used to reflect the time step.

[0013] As a preferred embodiment of the neural network-based electro-carbon coupling modeling method of the present invention, wherein: the dynamically weighted device health factor The formula used is: ,in, For initial equipment health factors, To adjust parameters that reflect the degree to which the equipment's health status affects the attenuation coefficient, It is used to reflect the health status of the equipment at time t, and is generated based on the number of hours the equipment has been running and maintenance records.

[0014] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the neural network-based electro-carbon coupling modeling method described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the neural network-based electro-carbon coupling modeling method described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: by using a time-series graph neural network to learn and simulate the complex nonlinear coupling relationship between electricity consumption, energy use and carbon emissions, a dynamic coupling relationship model between electricity data and carbon emission data is established. By combining deep learning with physical laws, and through multi-source data fusion and spatiotemporal coupling modeling, multi-source heterogeneous data are fused and analyzed, making the analysis results more comprehensive and accurate, accurately predicting carbon emission data, and preventing the one-sidedness of single data. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 The flowchart shows the neural network-based electro-carbon coupling modeling method of this invention. Figure 2 This is a flowchart illustrating the construction process of the dynamic graph structure of this invention. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Reference Figures 1-2 As one embodiment of the present invention, this embodiment provides a first aspect: the present invention provides a neural network-based method for modeling electro-carbon coupling, comprising, S1. Collect multi-source heterogeneous data of power production, preprocess the multi-source heterogeneous data, and obtain a cleaned and aligned time series dataset. Multi-source heterogeneous data refers to a data set from different devices and sensors with different data formats and properties. Data cleaning is to remove erroneous or abnormal data, and time alignment is to ensure that data from different time sources can correspond to the same time point, which is convenient for subsequent analysis.

[0023] Based on the collection of multi-source heterogeneous data from power, energy and carbon monitoring devices, the multi-source heterogeneous data includes power data, energy data, production data and direct monitoring data. Power data includes real-time power consumption, power and load curves of the equipment. Energy data includes fossil fuel consumption, lower heating value, elemental carbon content and carbon oxidation rate. Production data includes power generation, heat supply, equipment operating hours and load rate. Direct monitoring data includes real-time monitoring data of carbon dioxide concentration and flow rate.

[0024] The local outlier factor method is used to clean multi-source heterogeneous data. Sensor fault data, transmission error data, and abnormal operating condition data are identified and removed to obtain cleaned data. Specifically, the local density of each data point is obtained, outliers such as sensor fault data, data transmission error, or abnormal operating condition data are marked according to density differences, and the marked outliers are removed to obtain the cleaned dataset.

[0025] The time alignment method is used to align the cleaned data to the same time granularity, thereby obtaining the aligned data. Specifically, the unified time granularity is determined, and the time series data is interpolated or sampled so that all data points correspond to the same time scale, thus generating the time-aligned dataset.

[0026] The aligned data is standardized using numerical features to eliminate dimensional differences and obtain a cleaned and aligned time series dataset. Specifically, the mean and standard deviation of the data are calculated, and the Z-score standardization formula is applied to generate a cleaned and aligned time series dataset that eliminates the influence of dimensional differences.

[0027] The local outlier factor method can effectively clean up abnormal data and ensure data quality; time alignment is used to ensure data synchronization; standardization process unifies the scale of different physical quantity characteristics and avoids model training bias caused by different units of measurement.

[0028] S2. Based on the time series dataset, correlation analysis is performed to construct an electro-carbon coupling graph structure that changes with time and equipment status, and dynamic graph data is obtained. The constructed graph structure is used to reflect the relationship between energy, materials, and carbon flow between equipment. In this invention, nodes represent equipment, edges represent interactions, and it has dynamic evolution attributes.

[0029] A node definition method is used to partition the time-series dataset into nodes. A feature vector is extracted for each node. The feature vector includes static features and dynamic time-series features. The feature vectors are integrated to generate a node feature set. Static features include the rated power and design efficiency of the equipment. Dynamic time-series features include real-time power, load rate, and energy consumption per unit output. Each equipment is treated as a node. A feature vector is extracted for each node. The feature vector includes the rated power, design efficiency, real-time power, load rate, and energy consumption per unit output, i.e., the energy consumption corresponding to unit output. The feature vectors are integrated to form a node feature set.

[0030] An edge definition method is used to divide the relationships of energy flow, material flow, and carbon flow between nodes, and an initial weight is defined for each edge. The initial weights are integrated to generate an initial edge weight set. The existence of edges is determined according to the actual process flow or energy transmission path between equipment. An initial weight is defined for each edge, and the weight represents the strength of the relationship between nodes. The initial weight set contains the weights of all edges.

[0031] A dynamic evolution method is used to perform correlation analysis on the initial edge weight set, and dynamic weights are obtained through the edge weight evolution formula. The dynamic weights are then integrated to generate a dynamic edge weight set.

[0032] The weight of each edge is recalculated using the edge weight evolution formula, which is based on the following formula: ,in, This is a device health factor used to reflect the rate at which the connection strength between devices decays over time. Used to reflect the dynamic weight between node i and node j at time t. The weights are used to reflect the initial weights between nodes i and j, and t reflects the time step. An exponential function is used to simulate the impact of equipment health on edge weights. The weights gradually decrease over time, reflecting equipment wear or performance degradation. The exponential decay model is simple and conforms to the law of equipment performance degradation. The larger the value, the faster the edge weight decays, indicating a poorer device health.

[0033] Dynamically weighted device health factors The formula used is: ,in, For initial equipment health factors, To adjust parameters that reflect the degree to which the equipment's health status affects the attenuation coefficient, This value reflects the equipment health status at time t, and is generated based on equipment operating hours and maintenance records. A higher value indicates a worse equipment health status. The specific calculation logic is as follows: Equipment operating hours are extracted from the time-series dataset as an indicator of equipment usage intensity. Combined with maintenance records, such as the date of the most recent repair and the frequency of parts replacement, a weighted summation is performed to obtain a health score between 0 and 1, where 1 represents optimal health and 0 represents severe degradation. When the equipment health status is good, H(t) is close to 1. The item increases, and the equipment health factor increases. An increase in the value indicates a faster decay of edge weights, reflecting that when equipment is operating at high efficiency, its connectivity may change more rapidly due to increased load. When the equipment health status is poor, and H(t) approaches 0, the equipment health factor... The value approaches The decay rate is kept at the baseline level to avoid excessive adjustment of weights due to equipment degradation, ensuring the stability of the electro-carbon coupling relationship model, and dynamically adjusting the graph structure to adapt to the actual operating conditions of the equipment.

[0034] A graph structure integration method is used to combine the node feature set and the dynamic edge weight set to generate a dynamic graph structure. The dynamic graph structure includes a node feature matrix and a dynamic adjacency matrix. Based on the dynamic graph structure, dynamic graph data is generated by integration. Specifically, a node feature matrix is ​​constructed, where the rows of the feature matrix correspond to the nodes and the columns correspond to the features. A dynamic adjacency matrix is ​​constructed, where the matrix elements correspond to the dynamic weights of the edges between nodes. The two are combined to form a dynamic graph structure and generate dynamic graph data.

[0035] The node feature set is used to reflect device attributes and operating status, and the dynamic edge weight set is used to reflect the changes in the relationship between devices over time. It is based on the exponential decay model to simulate the impact of device health on connection strength, and finally forms a dynamic graph that can reflect the spatiotemporal changes between devices.

[0036] S3. Based on dynamic graph data, correlation analysis is performed. Spatiotemporal graph neural network modeling is carried out by integrating physical information through neural differential equations to obtain the electro-carbon coupling relationship model. Neural differential equations are a model that combines neural networks and differential equations, which can capture the continuous time variation law of the system and are suitable for the dynamic characteristics of electro-carbon coupling.

[0037] The graph convolution module of the temporal graph neural network is used to aggregate spatial information of dynamic graph data. The graph convolution network processes the features of each node and its neighboring nodes to capture the spatial dependencies between devices and obtain the spatially aggregated node embedding. Specifically, the node feature matrix and the dynamic adjacency matrix are input, and the features of each node and its neighboring nodes are weighted and summed. The weights are determined by the dynamic edge weights in the adjacency matrix. The results are then processed by a nonlinear activation function to obtain the spatially aggregated node embedding vector.

[0038] The node embeddings are processed using the time series modeling module of the time series graph neural network. The continuous time changes of the node state are modeled using neural differential equations to generate time-enhanced node states. Specifically, the node embeddings are treated as continuous time state variables, differential equations are constructed to describe the law of node state change over time, and the time evolution of the node state is solved by numerical integration method to obtain the time-enhanced node states.

[0039] The correlation analysis of node states is performed using the physical constraint integration method. The energy conservation equation is used as a constraint condition to generate node output. Specifically, the expression of the energy conservation equation is defined to ensure the balance between input energy, output energy and loss. This equation is used as part of the loss function to constrain model training, enhance the physical consistency of the model, and generate node output from the constrained node states.

[0040] The node outputs are integrated, and the total carbon emissions are calculated using a fully connected network to obtain an electro-carbon coupling model. Specifically, all node outputs are concatenated into a vector, input into the fully connected network, and the predicted total carbon emissions are calculated, ultimately yielding the electro-carbon coupling model. The formula used in the fully connected network calculation is as follows: , The matrix output by the node. , These are the weights and bias parameters of the hidden layer. It is the ReLU activation function. , The weights and bias parameters of the output layer, Let be the linear activation function of the publishing layer. This represents total carbon emissions.

[0041] S4. Optimize and verify the parameters of the electrocarbon coupling relationship model, and generate a verified model file.

[0042] S5. Deploy the model file to the production environment to dynamically monitor carbon emissions.

[0043] The time series dataset and dynamic graph data are split into training, validation and test sets. The electro-carbon coupling relationship model is trained based on the training, validation and test sets. The parameters of the electro-carbon coupling relationship model are optimized by the root mean square error. The weight parameters of the electro-carbon coupling relationship model are iteratively updated by the backpropagation algorithm and optimizer.

[0044] All data is divided into training, validation, and test sets in chronological order, with the proportions determined according to actual needs. The weight parameters of the electricity-carbon coupling relationship model are initialized, and the training data is input into the model to predict carbon emissions. The root mean square error between the predicted and actual values ​​is generated, and the parameter direction is adjusted according to the root mean square error gradient. The parameter weights are updated using an optimizer. During training, the model performance is evaluated periodically using the validation set. The learning rate is adjusted or training is stopped based on the validation error to prevent overfitting. After the test set reaches the prediction performance, parameter optimization is completed, and model training ends.

[0045] The trained electro-carbon coupling model is solidified into a model file, which is then deployed to the production environment. Based on the model file, carbon emission data is monitored in real time, and prediction results are output. The prediction results are then displayed, and carbon emission data and equipment monitoring charts are generated.

[0046] The standard format file for saving the weights, structure, and related parameters of the electro-carbon coupling relationship model to a PyTorch model file is used to import the model file into servers, edge computing devices, and cloud platforms to ensure that the model can be called in real time. Based on the deployed electro-carbon coupling relationship model file, dynamic monitoring is performed on the real-time collected carbon emission data, and the latest time series data is input into the electro-carbon coupling relationship model in real time to generate carbon emission prediction results.

[0047] Multi-source heterogeneous data acquisition and preprocessing ensure data accuracy and consistency; the construction of dynamic graph structures enables the electro-carbon coupling relationship model to reflect complex dynamic relationships between devices; neural differential equations ensure its temporal continuity and the accuracy of physical laws; the electro-carbon coupling relationship model is optimized and deployed to monitor carbon emission data in real time.

[0048] This embodiment also provides a computer device applicable to the neural network-based electro-carbon coupling modeling method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the neural network-based electro-carbon coupling modeling method proposed in the above embodiment.

[0049] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0050] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the neural network-based electro-carbon coupling modeling method as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0051] In summary, this invention achieves a closed-loop dynamic modeling and monitoring of electric carbon coupling by: collecting multi-source heterogeneous data, preprocessing the data through cleaning, alignment, and standardization, defining and dynamically evolving a dynamic graph structure based on nodes and edges, using a spatiotemporal graph neural network, modeling with physical constraints through graph convolution and neural differential equations, performing time series partitioning and parameter optimization, and finally deploying and predicting the electric carbon coupling relationship model in real time.

[0052] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A neural network-based method for modeling electro-carbon coupling, characterized in that: include, Collect multi-source heterogeneous data on power production, preprocess the multi-source heterogeneous data, and obtain a cleaned and aligned time-series dataset; Correlation analysis was performed based on time-series datasets to construct an electrocarbon coupling graph structure that varies with time and device status, and dynamic graph data was obtained. Correlation analysis was performed based on dynamic graph data, and spatiotemporal graph neural network modeling was carried out by integrating physical information through neural differential equations to obtain an electrocarbon coupling relationship model. The parameters of the electro-carbon coupling model were optimized and verified, and a verified model file was generated. Deploy the model files to the production environment to dynamically monitor carbon emissions.

2. The neural network-based electrocarbon coupling modeling method as described in claim 1, characterized in that: Based on multi-source heterogeneous data collected from power, energy, and carbon monitoring devices, the local outlier factor method is used to clean the multi-source heterogeneous data. Sensor fault data, transmission error data, and abnormal operating condition fluctuation data are identified and removed to obtain cleaned data. The time alignment method is used to align the cleaned data to the same time series granularity to obtain aligned data. The numerical features of the aligned data are standardized using the standardization method to eliminate dimensional differences and obtain a cleaned and aligned time series dataset.

3. The neural network-based electrocarbon coupling modeling method as described in claim 2, characterized in that: A node definition method is used to partition the time-series dataset into nodes. A feature vector is extracted from each node, comprising static and dynamic time-series features. These feature vectors are then integrated to generate a node feature set. An edge definition method is used to partition the relationships between nodes in terms of energy flow, material flow, and carbon flow, and an initial weight is defined for each edge. These initial weights are then integrated to generate an initial edge weight set. A dynamic evolution method is used to perform correlation analysis on the initial edge weight set, and dynamic weights are obtained through an edge weight evolution formula. These dynamic weights are then integrated to generate a dynamic edge weight set. Finally, a graph structure integration method is used to combine the node feature set and the dynamic edge weight set to generate a dynamic graph structure. Based on this dynamic graph structure, dynamic graph data is generated.

4. The neural network-based electrocarbon coupling modeling method as described in claim 3, characterized in that: The graph convolution module of the temporal graph neural network aggregates spatial information from dynamic graph data. The graph convolution network processes the features of each node and its neighboring nodes to capture the spatial dependencies between devices, resulting in spatially aggregated node embeddings. The temporal modeling module of the temporal graph neural network performs time-series processing on the node embeddings. The neural differential equation models the continuous time changes of node states, generating temporally enhanced node states. The physical constraint integration method performs correlation analysis on the node states, using the energy conservation equation as a constraint condition to generate node outputs. The node outputs are integrated, and the total carbon emissions are calculated through a fully connected network to obtain an electro-carbon coupling relationship model.

5. The neural network-based electrocarbon coupling modeling method as described in claim 4, characterized in that: The time series dataset and dynamic graph data are split into training, validation and test sets. The electro-carbon coupling relationship model is trained based on the training, validation and test sets. The parameters of the electro-carbon coupling relationship model are optimized by the root mean square error. The weight parameters of the electro-carbon coupling relationship model are iteratively updated by the backpropagation algorithm and optimizer.

6. The neural network-based electrocarbon coupling modeling method as described in claim 5, characterized in that: The trained electro-carbon coupling relationship model is solidified into a model file, which is then deployed to the production environment. Based on the deployed electro-carbon coupling relationship model file, real-time carbon emission data is dynamically monitored and input into the electro-carbon coupling relationship model in real time to generate carbon emission prediction results. The prediction results are then displayed, and carbon emission data and equipment monitoring charts are generated.

7. The neural network-based electrocarbon coupling modeling method as described in claim 3, characterized in that: The weight of each edge is recalculated using the edge weight evolution formula: The formula used is: ； in, For equipment health factors, Used to reflect the dynamic weight between node i and node j at time t. The weights between nodes i and j are used to reflect the initial weights, t reflects the time step, and both i and j are used to index the nodes.

8. The neural network-based electrocarbon coupling modeling method as described in claim 7, characterized in that: The dynamically weighted device health factor The formula used is: ； in, For initial equipment health factors, To adjust parameters that reflect the degree to which the equipment's health status affects the attenuation coefficient, It is used to reflect the health status of the equipment at time t, and is generated based on the number of hours the equipment has been running and maintenance records.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the neural network-based electro-carbon coupling modeling method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the neural network-based electro-carbon coupling modeling method according to any one of claims 1 to 8.

Images