A transformer area electric carbon coupling modeling method and carbon emission factor prediction system

Abstract

This invention belongs to the field of low-carbon management and data modeling technology for power distribution networks. It discloses a method for modeling the coupling relationship between power flow and carbon emissions in distribution areas, comprising the following main steps: acquisition and fusion of multi-source data, construction of dynamic carbon emission factors, analysis of power carbon response sensitivity, construction of causal networks, prediction of dynamic carbon emission factors, and platform implementation and visualization. This invention also discloses a carbon emission factor prediction system, including a central control platform connected to a data acquisition terminal and a data preprocessor, and integrated into the central control platform are a database construction module, a dynamic carbon emission factor calculation module, a sensitivity analysis module, a causal network construction module, and a dynamic carbon emission factor prediction module. The method of this invention achieves accurate quantification and identification of key factors in the coupling relationship between power flow and carbon emissions in distribution areas. Its system enables accurate short-term prediction of future carbon emissions in distribution areas and supports visualized management and real-time monitoring, improving the scientific rigor and feasibility of low-carbon control decisions.

Description

Technical Field

[0001] This invention belongs to the field of low-carbon management and data modeling technology of power distribution networks, and particularly relates to a method for coupled modeling of electrical and carbon emissions in distribution areas and a carbon emission factor prediction system. Background Technology

[0002] Distributed photovoltaic (PV) power generation and electric vehicle (EV) charging stations, among other new energy facilities, have been extensively integrated into urban distribution substations, promoting the transformation of the distribution network from a traditional power transmission platform to a green and low-carbon energy hub. However, during this transformation, PV power generation exhibits significant intermittency and randomness, while EV charging load shows a clear concentration at night, creating a "source-load time mismatch" problem. This mismatch leads to significant fluctuations in the temporal dimension and heterogeneity in the spatial dimension of power flow and carbon emissions in the distribution substations, making it difficult to accurately characterize the carbon footprint of the substations.

[0003] Existing technologies for assessing the carbon footprint of power distribution areas and modeling carbon emissions still have several shortcomings, making it difficult to adapt to the complex coupling characteristics of electricity and carbon emissions. First, traditional modeling generally uses static carbon emission factors, failing to consider the influence of dynamic factors such as the main grid power structure, unit operating efficiency, and equipment energy consumption status, resulting in models that cannot accurately reflect carbon emission levels under different time periods and power supply structures. Second, existing models lack an interactive modeling system for the synergistic relationship between multiple elements such as photovoltaics, charging piles, and energy storage, failing to reveal the mechanistic impact of multiple factors such as photovoltaic output, energy storage regulation, and charging load behavior on carbon emissions. Third, during multivariate data analysis, spurious correlation factors are not effectively eliminated, making the model susceptible to interference during the training and prediction phases, resulting in low modeling accuracy and robustness, making it difficult to support accurate assessment and prediction of carbon emissions in complex power distribution area scenarios.

[0004] Furthermore, while existing photovoltaic-storage-charging scheduling technologies can achieve coordinated energy operation among devices, they generally only focus on operational economics or load balance, failing to quantify the marginal impact of key variables on carbon emissions, thus making it difficult to provide a scientific basis for optimized carbon scheduling. At the same time, current research is mostly concentrated at the overall distribution network level, lacking a carbon emission decomposition analysis of the "equipment-path-time period" at the end of the distribution area, making it impossible to achieve refined tracking and management of carbon emissions at the micro-level.

[0005] In summary, existing technologies mainly suffer from the following problems: 1. Carbon emission factor modeling uses static average values ​​and does not consider the dynamic impacts of grid power structure, equipment efficiency, and time-series characteristics; 2. There is a lack of an electricity-carbon interaction modeling system oriented towards the collaborative operation mechanism of "photovoltaics-energy storage-charging piles," which cannot reveal the causal relationship between electricity and carbon; 3. There are spurious correlations between variables, resulting in insufficient modeling accuracy and robustness; 4. No analytical tools have been developed to quantify the marginal impact of key variables, making it difficult to support low-carbon prediction and regulation at the distribution station level; 5. Existing methods mostly focus on the macro level and lack the ability to decompose and visualize refined carbon emissions at the distribution station end.

[0006] Therefore, there is an urgent need for a technical solution that can integrate multi-source data, dynamically correct carbon emission factors, construct an electricity-carbon coupling model, and predict carbon emission factors, so as to achieve accurate quantification of the relationship between power flow and carbon emissions in distribution substations and support low-carbon regulation and management of distribution substations. Summary of the Invention

[0007] This invention provides a method for modeling the coupling of electricity and carbon emissions in power distribution areas and a carbon emission factor prediction system to solve existing technical problems.

[0008] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0009] A method for modeling the coupling of electricity and carbon in a transformer substation includes the following main steps:

[0010] S1. Collection and fusion of multi-source data: Collect multi-source data from the distribution area through data acquisition terminals to build a fusion database;

[0011] S2. Construction of dynamic carbon emission factor: Import multi-source data into the dynamic carbon emission factor calculation module to construct the dynamic carbon emission factor, and after path loss correction and equipment conversion efficiency correction, finally synthesize the overall dynamic carbon emission factor of the transformer area.

[0012] S3. Analysis of the sensitivity of the electrocarbon response: The sensitivity analysis module generates a library of electrocarbon response sensitivity functions;

[0013] S4. Construction of Causal Network: Through the causal network construction module and the multi-method fusion strategy, a multi-node causal network is constructed to identify the real causal relationship between each node, which is used to reveal the inherent coupling mechanism between power flow and carbon emissions in the transformer area.

[0014] S5. Prediction of dynamic carbon emission factors: Input key features in the causal network into the prediction module of dynamic carbon emission factors to perform short-term feature prediction, and predict the dynamic carbon emission factors at future times accordingly.

[0015] S6. Platform Implementation and Visualization: Integrate the above modules into the central control platform.

[0016] As a further improvement to the above technical solution:

[0017] In S1, the database construction module includes multi-source data. The multi-source data is processed by a data preprocessor, and then the multi-source data is cleaned, interpolated, and spatiotemporally aligned by the database construction module. The data cleaning includes using interpolation rules to straighten non-equal interval data and marking and reasonably filling in missing data.

[0018] In S1, the multi-source data includes load curves, photovoltaic power generation, charging pile power behavior, energy storage operation status, meteorological data, and main grid power supply structure, and is collected by data acquisition terminals configured in the transformer substation area.

[0019] In S2, dynamic carbon emission factors are constructed for four types of power supply sources: main grid power supply, photovoltaic power generation, energy storage discharge, and external power transmission. The dynamic carbon emission factor calculation module for main grid power supply adopts a real-time marginal carbon emission factor calculation model and combines it with the real-time power structure of the main grid for dynamic updates. The dynamic carbon emission factor calculation module for photovoltaic power generation calculates the instantaneous carbon emission factor based on the carbon emission data of the component life cycle and real-time output data. The dynamic carbon emission factor calculation module for energy storage system establishes carbon tags and performs dynamic calculations. The dynamic carbon emission factor calculation module for external power transmission calculates the weighted average carbon emission factor when the distribution area is used as a power source. Subsequently, a method combining PCA principal component analysis and entropy weight method is used to determine the dynamic contribution weight of carbon emission factors of various power supply sources, and the dynamic carbon emission factor of the distribution area as a whole is calculated accordingly.

[0020] In S2, the path loss correction includes amplifying the carbon emission factor to reflect the effective carbon emission intensity on the user side, wherein the line loss rate is the line loss rate measured in real time.

[0021] In S3, the sensitivity analysis module uses deep learning to analyze the marginal impact of key variables on carbon emissions. The deep learning adopts the gradient boosting tree (GBDT) regression model, uses the collected and processed data indicators as features, uses deep learning to predict the overall dynamic carbon emission factor of the power station area, and uses SHAP to analyze the marginal impact of each feature on the model output.

[0022] In S4, the causal network construction module first determines the influencing factors with a correlation degree greater than 0.6 through grey relational analysis, then eliminates spurious correlation interference through time-series partial correlation analysis, and finally constructs a structured causal graph and quantifies the causal strength.

[0023] In S5, the prediction module for dynamic carbon emission factors includes a causal feature data prediction module and a dynamic carbon emission factor calculation module. Based on the key feature indicators obtained by the causal network construction module, the causal feature data prediction module is used to predict its future data, and then the data is fed into the neural network model in the dynamic carbon emission factor calculation module for prediction, thereby obtaining the future dynamic carbon emission factor number.

[0024] In S6, the dynamic carbon emission factor calculation module, sensitivity analysis module, causal network construction module, and dynamic carbon emission factor prediction module are integrated into the central control platform. The central control platform is used to realize data management, modeling operation, result visualization, and model effectiveness monitoring.

[0025] A carbon emission factor prediction system includes a central control platform connected to a data acquisition terminal and a data preprocessor, and a database construction module, a dynamic carbon emission factor calculation module, a sensitivity analysis module, a causal network construction module, and a dynamic carbon emission factor prediction module integrated into the central control platform. The data acquisition terminal is used to collect multi-source data such as transformer load, photovoltaic power generation, energy storage discharge, charging piles, and meteorological data in real time, and upload the collected data to the central control platform. The data preprocessor is used to clean, interpolate, spatiotemporally align, and normalize the collected data to construct a high-quality fusion database. The dynamic carbon emission factor calculation module is used to calculate the dynamic carbon emission factor based on multi-source data and synthesize the overall dynamic carbon emission factor of the transformer area through path loss and equipment efficiency correction. The sensitivity analysis module is used to analyze the marginal impact of each key variable on the overall carbon emission factor of the transformer area and generate an electricity carbon response sensitivity function library. The causal network construction module is used to construct a multi-node causal network based on a multi-method fusion strategy to identify the inherent coupling mechanism between transformer power flow and carbon emissions. The dynamic carbon emission factor prediction module is used to make short-term predictions of the dynamic carbon emission factor at future times using key features selected from the causal network.

[0026] As a further improvement to the above technical solution:

[0027] The central control platform is connected to a monitoring terminal, which is used to monitor data fluctuations during the modeling process and feed the collected modeling data back to the central control platform for subsequent modeling accuracy verification and method optimization.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] By collecting and fusing multi-source data, a unified spatiotemporal characterization of power flow and carbon emission data for power distribution areas is achieved, providing technical support for accurate assessment of carbon footprint at the distribution area level. The construction of dynamic carbon emission factors introduces a dual correction mechanism of path loss and equipment efficiency, enhancing the model's adaptability and robustness to the dynamic characteristics of multi-source power supply. A sensitivity analysis module generates a library of electrical carbon response sensitivity functions, quantifying the marginal impact of key variables on carbon emissions and providing a basis for low-carbon regulation strategy formulation. Causal network construction employs a fusion of grey relational analysis and structured causal graphs to accurately reveal the inherent coupling relationships of multi-node causal networks. The prediction module for dynamic carbon emission factors enables the prediction of carbon emission factors at future moments, supporting forward-looking carbon management and strategy optimization. Furthermore, this method relies on a central control platform to achieve integrated modeling and visualization, significantly reducing dependence on additional hardware, lowering implementation costs, resolving the disconnect between modeling and engineering applications, and improving the model's practical applicability. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a schematic diagram of the connection structure of the carbon emission factor prediction system.

[0032] Figure 2 This is a schematic diagram of the carbon emission factor prediction system. Detailed Implementation

[0033] To facilitate understanding of the present invention, the present invention will be described more fully and in detail below with reference to the accompanying drawings and preferred embodiments, but the scope of protection of the present invention is not limited to the following specific embodiments.

[0034] Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by those skilled in the art. The technical terms used herein are for the purpose of describing particular embodiments only and are not intended to limit the scope of the invention.

[0035] Unless otherwise specified, all raw materials, reagents, instruments and equipment used in this invention can be purchased from the market or prepared by existing methods.

[0036] Example: Figure 1 and Figure 2 As shown, the transformer substation carbon coupling modeling method of this embodiment includes the following main steps:

[0037] S1. Acquisition and Fusion of Multi-Source Data: Multi-source data from the distribution area is collected through multi-source data acquisition terminals, then processed by a data preprocessor, and finally cleaned, interpolated and spatiotemporally aligned by a database construction module to build a fused database.

[0038] S2. Construction of dynamic carbon emission factors: Multi-source data is imported into the dynamic carbon emission factor calculation module. Dynamic carbon emission factors are constructed for four types of power supply sources: main grid power supply, photovoltaic power generation, energy storage discharge, and external power transmission. After path loss correction and equipment conversion efficiency correction, the overall dynamic carbon emission factor of the distribution area is finally synthesized.

[0039] S3. Analysis of the sensitivity of the carbon response: Through the sensitivity analysis module, deep learning is used to analyze the marginal impact of key variables on carbon emissions and generate a library of carbon response sensitivity functions.

[0040] S4. Construction of Causal Network: Through the causal network construction module and the multi-method fusion strategy, a multi-node causal network is constructed to identify the real causal relationship between each node, which is used to reveal the inherent coupling mechanism between power flow and carbon emissions in the transformer area.

[0041] S5. Prediction of dynamic carbon emission factors: Input key features in the causal network into the prediction module of dynamic carbon emission factors to perform short-term feature prediction, and predict the dynamic carbon emission factors at future times accordingly.

[0042] S6. Platform Implementation and Visualization: The dynamic carbon emission factor calculation module, sensitivity analysis module, causal network construction module, and dynamic carbon emission factor prediction module are integrated into the central control platform to realize data management, modeling operation, result visualization, and model effectiveness monitoring.

[0043] By collecting and fusing multi-source data (S1), a unified spatiotemporal characterization of power flow and carbon emission data for power distribution areas is achieved, providing technical support for accurate assessment of carbon footprint at the distribution area level. Through the construction of dynamic carbon emission factors (S2), a dual correction mechanism of path loss and equipment efficiency is introduced to improve the model's adaptability and robustness to the dynamic characteristics of multi-source power supply. Combined with a sensitivity analysis module (S3), an electricity-carbon response sensitivity function library is generated, which can quantify the marginal impact of key variables on carbon emissions, providing a basis for low-carbon regulation strategy formulation. Through causal network construction (S4), a method combining grey relational analysis and structured causal graphs is used to accurately reveal the inherent coupling relationship of the causal network of multiple nodes in the "source-load-storage-carbon" chain. With the help of a dynamic carbon emission factor prediction module (S5), the prediction of carbon emission factors at future moments is achieved, supporting forward-looking carbon management and strategy optimization. Furthermore, this method relies on a central control platform (S6) to achieve integrated modeling and visualization, significantly reducing dependence on additional hardware, reducing implementation costs, solving the problem of disconnect between modeling and engineering applications, and improving the model's practicality.

[0044] In this embodiment, in S1, the database construction module includes multivariate data. Data cleaning includes normalizing non-uniformly spaced data using interpolation methods and labeling and appropriately filling in missing data. The accuracy error of the spatiotemporal alignment processing does not exceed 5 minutes, providing a high-quality data foundation for modeling. The database construction module can achieve efficient fusion of multi-source heterogeneous data, improve the integrity and continuity of data by using interpolation methods and missing value filling; through normalization of non-uniformly spaced data and spatiotemporal alignment processing (accuracy error not exceeding 5 minutes), it significantly improves the consistency and timeliness of data, providing a high-quality and reliable data foundation for subsequent electro-carbon coupling modeling and dynamic carbon emission factor calculation, thereby effectively avoiding the problem of decreased model accuracy due to data asynchrony or missing data in existing technologies.

[0045] In this embodiment, in S1, the multi-source data includes load curves, photovoltaic power generation, charging pile power behavior, energy storage operation status, meteorological data, and the main grid power supply structure, and is collected through data acquisition terminals configured in the distribution area. The acquisition frequency is no less than once every 15 minutes. Through high-frequency acquisition and multi-source fusion, the power flow characteristics and carbon emission characteristics of the distribution area under different operating conditions can be comprehensively reflected, providing a foundation for the subsequent construction of dynamic carbon emission factors and electricity-carbon coupling modeling.

[0046] In this embodiment, in S2, the dynamic carbon emission factor calculation module uses a real-time marginal carbon emission factor calculation model for grid power supply, combined with the real-time dynamic update of the grid power structure. For photovoltaic power generation, the module calculates the instantaneous carbon emission factor based on the carbon emission data of the component life cycle and the real-time output data. For energy storage systems, the module establishes carbon tags and performs dynamic calculations. For external power transmission, the module calculates the weighted average carbon emission factor when the distribution area is used as a power source. Subsequently, the module uses a combination of PCA principal component analysis and entropy weight method to determine the dynamic contribution weight of carbon emission factors of various power supply sources, and calculates the overall dynamic carbon emission factor of the distribution area accordingly. By establishing dynamic carbon emission factor models for different power supply sources, the real-time changing characteristics of the power supply structure of the distribution area can be fully reflected. By dynamically updating the models in conjunction with the real-time power structure of the main grid, photovoltaic output, and energy storage operation status, the timeliness and accuracy of carbon emission factor calculation are significantly improved. At the same time, the introduction of a weight determination mechanism that combines PCA principal component analysis and entropy weighting method can achieve adaptive optimization of the carbon emission contribution of various power supply sources without the need for subjective human weighting, thereby improving the scientificity and robustness of the overall model and providing reliable support for accurate assessment of carbon emissions in the distribution area.

[0047] In this embodiment, in step S2, path loss correction is used to amplify the carbon emission factor to reflect the effective carbon emission intensity on the user side, where the line loss rate is obtained from real-time monitoring data. The path loss correction can be calculated using the following formula to determine the load point carbon emission factor: Among them, CI 负荷点 The carbon emission factor, CI, represents the point of load. 台区 The overall carbon emission factor for the transformer area is represented by λ, and the line loss rate is represented by λ, which can be dynamically updated based on real-time monitoring data. Equipment conversion efficiency correction mainly targets the power supply source for "energy storage"; other power supply types can be ignored. This is used when calculating the carbon emission factor CI for energy storage discharge. 储能放电 At time (t), the denominator is the discharge capacity of the energy storage system, and the numerator is the total carbon emission cost during the charging phase. The total carbon emission cost already includes the energy conversion efficiency loss of the energy storage system during the charge-discharge cycle.

[0048] In this embodiment, in S3, deep learning adopts the gradient boosting tree (GBDT) regression model, uses the collected and processed data indicators as features, uses deep learning to predict the overall dynamic carbon emission factor of the transformer area, and analyzes the marginal impact of each feature on the model output through SHAP.

[0049] In this embodiment, in S4, the causal network construction module first identifies influencing factors with a correlation degree greater than 0.6 through grey relational analysis, then eliminates spurious correlation interference through time-series partial correlation analysis, and finally constructs a structured causal graph and quantifies the causal strength. Specifically, the module selects data indicators from the sensitivity analysis module that play a key role in the overall carbon emission factor of the distribution area. Then, it analyzes the correlation between these data indicators and the overall carbon emission factor of the distribution area through grey relational analysis, eliminating data indicators with a correlation degree less than 0.6. Next, it uses time-series partial correlation analysis to clarify whether the overall carbon emission factor of the distribution area has a time dependency on the selected data indicators. Finally, it constructs a structured causal graph based on the analysis results to further extract key factors. By introducing a dynamic correction mechanism for path loss and equipment conversion efficiency, the dynamic carbon emission factor can be updated over time, significantly improving the model's adaptability to actual operating conditions and computational accuracy. Simultaneously, considering energy storage cycle efficiency loss helps to more accurately reflect the end-point carbon emission intensity of the distribution area, providing a solid foundation for accurate carbon footprint assessment and low-carbon optimized scheduling.

[0050] In this embodiment, in step S5, the dynamic carbon emission factor prediction module includes a causal feature data prediction module and a dynamic carbon emission factor calculation module. Based on the key feature indicators obtained from the causal network construction module, the causal feature data prediction module predicts future data, which is then input into the neural network model in the dynamic carbon emission factor calculation module for prediction, thereby obtaining the future dynamic carbon emission factor number. For key feature indicators, missing values ​​are adjusted using interpolation, timestamps are standardized, and data normalization is performed. Reasonable values ​​predicted by regression or time series models are used to replace obviously abnormal values. By determining key feature indicators through causal networks, the prediction model focuses on the factors with the greatest impact on carbon emission factors, improving the relevance and accuracy of the prediction. Using the causal feature data prediction module combined with neural networks to predict dynamic carbon emission factors can capture the nonlinear relationship between power flow and carbon emissions in the distribution area, achieving accurate prediction of short-term carbon emission trends. Simultaneously, interpolation of missing values, standardization of timestamps, and reasonable correction of abnormal values ​​improve data integrity and consistency, thereby enhancing the robustness and feasibility of the model and providing reliable support for distribution area-level carbon emission management and low-carbon regulation.

[0051] In this embodiment, a carbon emission factor prediction system includes a central control platform connected to a data acquisition terminal and a data preprocessor, and a database construction module, a dynamic carbon emission factor calculation module, a sensitivity analysis module, a causal network construction module, and a dynamic carbon emission factor prediction module integrated in the central control platform. The data acquisition terminal is used to collect multi-source data such as transformer load, photovoltaic power generation, energy storage discharge, charging piles, and meteorology in real time, and upload the collected data to the central control platform. The data preprocessor is used to clean, interpolate, spatiotemporally align, and normalize the collected data to build a high-quality fusion database. The dynamic carbon emission factor calculation module is used to calculate the dynamic carbon emission factor based on multi-source data, and synthesize the overall dynamic carbon emission factor of the transformer area through path loss and equipment efficiency correction. The sensitivity analysis module is used to analyze the marginal impact of each key variable on the overall carbon emission factor of the transformer area and generate an electric carbon response sensitivity function library. The causal network construction module is used to construct a multi-node causal network based on a multi-method fusion strategy to identify the inherent coupling mechanism between transformer power flow and carbon emissions. The dynamic carbon emission factor prediction module is used to make short-term predictions of the dynamic carbon emission factor at future times using key features selected from the causal network. The central control platform performs a normality test on the prediction results to check whether the data conforms to the model assumptions, and presents the results to the administrator in the form of a box plot, allowing the administrator to judge the effectiveness of the prediction model. Each module is preferably developed based on the Python platform, supporting multi-source data interface adaptation. No complex programming is required; the modeling process can be started through simple parameter configuration. Modeling results can be synchronized to the central control platform with one click, supporting visualization and data export, and enabling multi-user collaborative viewing and analysis. The data acquisition terminal collects multi-source data (data from transformers, photovoltaic inverters, energy storage systems, charging piles, etc.) and uploads it to the central control platform. The central control platform then processes the data using a data preprocessor, re-uploads the processed data, and sends the fused database to the dynamic factor calculation module to calculate the overall dynamic carbon emission factor for the transformer area under the corresponding time series. The central control platform then sends the newly formed fused database to the sensitivity analysis module for analysis, and sends the analysis results back to the central control platform. Based on the analysis results, the central control platform sends the corresponding data to the causal network construction module to construct a causal graph and sends the results back. Finally, the platform sends the results to the dynamic carbon emission factor prediction module to predict the future overall carbon emission factor for the transformer area. Administrators monitor the effectiveness of the model prediction and record relevant information through the central control platform. If the model prediction is effective, users can view the future carbon emission factor prediction analysis for the transformer area through the central control platform.

[0052] The system achieves high-frequency, multi-dimensional real-time data acquisition and high-quality fusion through multi-source data acquisition terminals and data preprocessors, ensuring a comprehensive reflection of the operating status and environmental conditions of the power distribution area. The dynamic carbon emission factor calculation module, combined with path loss and equipment efficiency correction, accurately characterizes the overall carbon emission factor of the power distribution area. The sensitivity analysis module and causal network construction module can reveal the marginal impact of key variables and the inherent coupling mechanism between power flow and carbon emissions in the power distribution area, improving the scientific nature and interpretability of the modeling. The dynamic carbon emission factor prediction module achieves short-term prediction based on causal characteristics, enabling the system to grasp carbon emission trends in advance. The central control platform provides visualization, normality testing, and multi-user collaboration functions, enhancing the system's ease of operation and decision support capabilities. The entire system is developed based on the Python platform, allowing the modeling process to be started without complex programming, significantly improving the system's robustness, feasibility, and implementation efficiency, providing reliable technical support for refined carbon management and low-carbon regulation at the power distribution area level.

[0053] In this embodiment, the central control platform is connected to a monitoring terminal. The monitoring terminal is used to monitor data fluctuations during the modeling process and feeds back the collected modeling data to the central control platform for subsequent modeling accuracy verification and method optimization. By monitoring data fluctuations in real time during the modeling process through the monitoring terminal, anomalies or deviations can be detected in a timely manner, ensuring data quality and model operation stability. Feeding the monitoring data back to the central control platform can be used for modeling accuracy verification and method optimization, thereby continuously improving the accuracy and robustness of the model and ensuring the reliability and feasibility of power grid carbon coupling modeling and dynamic carbon emission factor prediction.

Claims

1. A method for modeling electrical-carbon coupling in transformer substations, characterized in that, The main steps include: S1. Collection and fusion of multi-source data: Collect multi-source data from the distribution area through data acquisition terminals to build a fusion database; S2. Construction of dynamic carbon emission factor: Import multi-source data into the dynamic carbon emission factor calculation module to construct the dynamic carbon emission factor, and after path loss correction and equipment conversion efficiency correction, finally synthesize the overall dynamic carbon emission factor of the transformer area. S3. Analysis of the sensitivity of the electrocarbon response: The sensitivity analysis module generates a library of electrocarbon response sensitivity functions; S4. Construction of Causal Network: Through the causal network construction module and the multi-method fusion strategy, a multi-node causal network is constructed to identify the real causal relationship between each node, which is used to reveal the inherent coupling mechanism between power flow and carbon emissions in the transformer area. S5. Prediction of dynamic carbon emission factors: Input key features in the causal network into the prediction module of dynamic carbon emission factors to perform short-term feature prediction, and predict the dynamic carbon emission factors at future times accordingly. S6. Platform Implementation and Visualization: Integrate the above modules into the central control platform.

2. The transformer substation carbon coupling modeling method according to claim 1, characterized in that, In S1, the database construction module includes multi-source data. The multi-source data is processed by a data preprocessor, and then the multi-source data is cleaned, interpolated, and spatiotemporally aligned by the database construction module. The data cleaning includes using interpolation rules to straighten non-equal interval data and marking and reasonably filling in missing data.

3. The transformer substation carbon coupling modeling method according to claim 2, characterized in that... In S1, the multi-source data includes load curves, photovoltaic power generation, charging pile power behavior, energy storage operation status, meteorological data, and main grid power supply structure, and is collected by data acquisition terminals configured in the transformer substation area.

4. The transformer substation carbon coupling modeling method according to claim 3, characterized in that, In S2, dynamic carbon emission factors are constructed for four types of power supply sources: main grid power supply, photovoltaic power generation, energy storage discharge, and external power transmission. The dynamic carbon emission factor calculation module for main grid power supply adopts a real-time marginal carbon emission factor calculation model and combines it with the real-time power structure of the main grid for dynamic updates. The dynamic carbon emission factor calculation module for photovoltaic power generation calculates the instantaneous carbon emission factor based on the carbon emission data of the component life cycle and real-time output data. The dynamic carbon emission factor calculation module for energy storage system establishes carbon tags and performs dynamic calculations. The dynamic carbon emission factor calculation module for external power transmission calculates the weighted average carbon emission factor when the distribution area is used as a power source. Subsequently, a method combining PCA principal component analysis and entropy weight method is used to determine the dynamic contribution weight of carbon emission factors of various power supply sources, and the dynamic carbon emission factor of the distribution area as a whole is calculated accordingly.

5. The transformer substation carbon coupling modeling method according to claim 4, characterized in that, In S2, the path loss correction includes amplifying the carbon emission factor to reflect the effective carbon emission intensity on the user side, wherein the line loss rate is the line loss rate measured in real time.

6. The transformer substation carbon coupling modeling method according to claim 5, characterized in that, In S3, the sensitivity analysis module uses deep learning to analyze the marginal impact of key variables on carbon emissions. The deep learning adopts the gradient boosting tree (GBDT) regression model, uses the collected and processed data indicators as features, uses deep learning to predict the overall dynamic carbon emission factor of the power station area, and uses SHAP to analyze the marginal impact of each feature on the model output.

7. The transformer substation carbon coupling modeling method according to claim 6, characterized in that, In S4, the causal network construction module first determines the influencing factors with a correlation degree greater than 0.6 through grey relational analysis, then eliminates spurious correlation interference through time-series partial correlation analysis, and finally constructs a structured causal graph and quantifies the causal strength.

8. The transformer substation carbon coupling modeling method according to claim 7, characterized in that, In S5, the prediction module for dynamic carbon emission factors includes a causal feature data prediction module and a dynamic carbon emission factor calculation module. Based on the key feature indicators obtained from the causal network construction module, the causal feature data prediction module is used to predict future data, which is then input into the neural network model in the dynamic carbon emission factor calculation module for prediction, thereby obtaining the future number of dynamic carbon emission factors. In S6, the dynamic carbon emission factor calculation module, sensitivity analysis module, causal network construction module, and dynamic carbon emission factor prediction module are integrated into the central control platform. The central control platform is used to realize data management, modeling operation, result visualization, and model effectiveness monitoring.

9. A carbon emission factor prediction system, characterized in that, The system includes a central control platform connected to a data acquisition terminal and a data preprocessor, and integrated modules for database construction, dynamic carbon emission factor calculation, sensitivity analysis, causal network construction, and dynamic carbon emission factor prediction. The data acquisition terminal collects multi-source data in real time, including data on transformer load, photovoltaic power generation, energy storage discharge, charging piles, and meteorological data, and uploads the collected data to the central control platform. The data preprocessor cleans, interpolates, aligns, and normalizes the collected data to build a high-quality fusion database. The dynamic carbon emission factor calculation module calculates the dynamic carbon emission factor based on multi-source data and, through path loss and equipment efficiency correction, synthesizes the overall dynamic carbon emission factor for the transformer area. The sensitivity analysis module analyzes the marginal impact of key variables on the overall carbon emission factor of the transformer area and generates a library of electrical carbon response sensitivity functions. The causal network construction module constructs a multi-node causal network based on a multi-method fusion strategy to identify the inherent coupling mechanism between transformer power flow and carbon emissions. The dynamic carbon emission factor prediction module uses key features selected from the causal network to make short-term predictions of the dynamic carbon emission factor at future times.

10. The carbon emission factor prediction system according to claim 9, characterized in that, The central control platform is connected to a monitoring terminal, which is used to monitor data fluctuations during the modeling process and feed the collected modeling data back to the central control platform for subsequent modeling accuracy verification and method optimization.

Images