A method and system for carbon emission estimation based on non-intrusive power load monitoring

Abstract

The present application relates to a kind of carbon emission estimation method and system based on non-invasive power load monitoring, method includes: through non-invasive monitoring equipment, real-time acquisition enterprise overall power load data, through feature extraction and machine learning classification model, obtain the real-time power consumption state of each enterprise equipment, obtain corresponding real-time power, and combine with corresponding carbon emission intensity factor, calculate the direct carbon emission total amount of enterprise equipment real-time operation;Through regional power grid dispatching system, obtain the real-time marginal carbon emission factor of power grid operation, and combine with the real-time power of each enterprise equipment, calculate the indirect carbon emission amount caused by the power consumption behavior of enterprise;The direct carbon emission total amount and indirect carbon emission amount obtained are analyzed in real time, obtain the real-time carbon emission amount of enterprise overall.Compared with prior art, the present application realizes high-precision, low-cost real-time estimation of carbon emission, and provides effective data support for energy saving and emission reduction.

Description

Technical Field

[0001] This invention relates to the field of carbon emission monitoring technology, and in particular to a carbon emission estimation method and system based on non-intrusive power load monitoring. Background Technology

[0002] As a major source of greenhouse gas emissions, the monitoring and control of carbon emissions from the power industry are crucial. Enterprises, as significant electricity consumers, contribute substantially to carbon emissions through their daily electricity consumption. Real-time monitoring of enterprise carbon emissions and guidance for low-carbon behaviors among residents are essential for achieving overall carbon reduction goals.

[0003] Existing carbon emission monitoring methods mainly include two types: Intrusive Load Monitoring (ILM) and Non-Intrusive Load Monitoring (NILM). Intrusive load monitoring requires the installation of individual sensors or metering devices on each electrical appliance. Although it offers high accuracy, it is costly, complex to install and maintain, and difficult to popularize. In contrast, NILM technology only requires the installation of a single monitoring device at the user's power inlet. It uses advanced machine learning algorithms to decompose the overall electricity consumption data of the enterprise, accurately identifying the electricity consumption characteristics of each electrical appliance. It has significant advantages such as low cost, convenient deployment, and ease of user acceptance.

[0004] However, current non-intrusive load monitoring methods mostly focus on classifying and identifying equipment energy consumption, with insufficient research on combining them with real-time carbon emission monitoring. Existing methods typically suffer from low real-time performance, insufficient model generalization ability, and strong data dependence. Furthermore, most current carbon emission estimation research focuses on direct carbon emission monitoring, and there is still a lack of effective methodologies for indirect carbon emissions generated by enterprise users, especially for real-time estimation through marginal carbon emission factors.

[0005] On the other hand, with the continuous improvement of the electricity market mechanism and the promotion of carbon trading mechanisms, corporate users have gradually become important participants in carbon emission reduction activities. Real-time monitoring of the direct and indirect carbon emissions from corporate equipment can not only encourage users to actively optimize their electricity consumption behavior, but also provide more accurate data support for the carbon trading market. However, at present, there is a lack of a method that can achieve both real-time monitoring and accurate differentiation between direct and indirect carbon emissions, which restricts the enthusiasm and accuracy of enterprises in participating in low-carbon actions. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the existing technology, where most current carbon emission estimation studies focus on direct carbon emission monitoring, and there is still a lack of effective methodologies for indirect carbon emissions generated by enterprise users, especially for real-time estimation through marginal carbon emission factors. This invention provides a carbon emission estimation method and system based on non-intrusive power load monitoring.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] A carbon emission estimation method based on non-intrusive power load monitoring includes the following steps: real-time collection of overall enterprise power load data through non-intrusive monitoring equipment;

[0009] Based on the collected overall electricity load data of enterprises, the real-time electricity status of each enterprise's equipment is obtained through pre-trained feature extraction and machine learning classification models;

[0010] The real-time power consumption of each enterprise's equipment is obtained based on its real-time power consumption status, and combined with the corresponding carbon emission intensity factor to calculate the total direct carbon emissions of the enterprise's equipment during real-time operation.

[0011] The real-time marginal carbon emission factor of the power grid operation is obtained through the regional power grid dispatch system and combined with the real-time power of each enterprise's equipment to calculate the indirect carbon emissions caused by the enterprise's electricity consumption behavior.

[0012] The total direct and indirect carbon emissions are aggregated and analyzed in real time to obtain the overall real-time carbon emissions of the enterprise.

[0013] Furthermore, the formula for calculating the total direct carbon emissions from the real-time operation of the enterprise's equipment is as follows:

[0014]

[0015] In the formula, For enterprise equipment within a time range Total direct carbon emissions within the region; This indicates the current time of the nth device. t Real-time power; This represents the real-time carbon emission intensity factor corresponding to the nth device. d This indicates the number of identified devices within the company.

[0016] Furthermore, the formula for calculating the indirect carbon emissions caused by the enterprise's electricity consumption behavior is as follows:

[0017]

[0018] In the formula, For enterprise equipment within a time range Indirect carbon emissions within the region for t Time of the first i Real-time marginal carbon emission factors connected to the device For the nth device at the current moment t Real-time power, dThis indicates the number of identified devices within the enterprise;

[0019] The process of obtaining the real-time marginal carbon emission factor is as follows:

[0020] The real-time output data of different generator sets within the enterprise's region is obtained through the regional power grid dispatch system, and the real-time marginal carbon emission factor is calculated by combining the carbon emission characteristic curves of the corresponding generator sets.

[0021] Furthermore, the calculation expression for the real-time marginal carbon emission factor is as follows:

[0022]

[0023] In the formula, for t Time of the first i Real-time marginal carbon emission factors connected to the device The emission characteristic curve of unit A is shown. To meet the real-time load demand of the region, The marginal emission rate of unit R. This is the participation factor of the generating unit for a unit load increment. For the unit t Carbon emission intensity at any given time.

[0024] Furthermore, the method assigns dynamic weights to both direct and indirect carbon emissions based on equipment usage intensity, time sensitivity factors, and regional power grid carbon intensity fluctuation parameters for weighted fusion; the calculation expression for the weighted fusion is as follows:

[0025]

[0026] In the formula, Let be the weighted carbon emissions at time t. Total direct carbon emissions Indirect carbon emissions, As a weighting of total direct carbon emissions, As a weight for indirect carbon emissions, .

[0027] Furthermore, the real-time collection of the enterprise's overall electricity load data through non-invasive monitoring equipment specifically involves:

[0028] A single-point load monitoring device is installed at the main circuit inlet of the enterprise. The device records the overall power load data of the enterprise and performs data cleaning.

[0029] Furthermore, the method also includes dividing the cleaned enterprise's overall electricity load data into different load types using load curve analysis technology and clustering algorithms; extracting the characteristic load curves of each type of load equipment under typical operating modes; obtaining regional electricity carbon emission intensity data; calculating the typical carbon emission levels of different types of loads in different time periods; and establishing an enterprise electricity carbon emission benchmark data model for comparison with the enterprise's overall real-time carbon emissions to formulate energy-saving behavior adjustment and carbon footprint control strategies for the enterprise.

[0030] Furthermore, the feature extraction and machine learning classification model processing steps include:

[0031] The overall electricity load data of the enterprise is preprocessed. The preprocessing operation includes data denoising, normalization and feature extraction. The features extracted include steady-state features and transient features.

[0032] The extracted features are input into a trained deep learning load identification model to obtain the real-time power consumption status of each enterprise's equipment; the training process of the deep learning load identification model includes:

[0033] The deep learning load identification model is iteratively trained using labeled historical enterprise overall electricity load data, and the generalization performance of the model is evaluated periodically during the training process using cross-validation and early stopping of training.

[0034] Furthermore, the method also includes a comprehensive analysis based on the enterprise's overall real-time carbon emission data and the enterprise's electricity consumption behavior, from the dimensions of current electricity consumption structure, load period characteristics, and high carbon emission behavior, to generate personalized and scenario-based energy-saving and carbon reduction optimization suggestions.

[0035] The present invention also provides a carbon emission estimation system based on non-intrusive power load monitoring, including a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the method described above.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] (1) This invention proposes to collect overall electricity consumption data by installing non-intrusive monitoring equipment on the enterprise bus, and combine it with machine learning classification model to decompose the equipment operation status, thereby determining the real-time power of each enterprise's equipment. Then, the carbon emission intensity factor corresponding to the equipment is obtained from an authoritative institution to realize the calculation of the total direct carbon emissions of the enterprise's equipment. Based on the real-time output data of different generator sets in the region, combined with the carbon emission characteristic curve of the corresponding generator set, the real-time changing marginal carbon emission factor of the power grid is calculated, so as to realize the accurate quantification of the indirect carbon emissions caused by the enterprise's electricity consumption behavior, and finally realize the second-level integrated monitoring of direct and indirect carbon emissions.

[0038] (2) This invention achieves high-precision, low-cost real-time carbon emission estimation through non-invasive monitoring and deep learning technology, providing effective data support for energy conservation and emission reduction. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating a carbon emission estimation method based on non-intrusive power load monitoring provided in an embodiment of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0041] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0042] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0043] Example 1

[0044] like Figure 1 As shown, this embodiment provides a carbon emission estimation method based on non-intrusive power load monitoring, including the following steps:

[0045] S1: Real-time collection of overall enterprise power load data through non-intrusive monitoring equipment, such as by installing non-intrusive monitoring equipment at a single point at the enterprise's main power inlet to collect overall enterprise power load data in real time;

[0046] S2: Using deep learning algorithms, with the overall electricity load data of the enterprise as input, the real-time electricity status of each enterprise's equipment is obtained through pre-trained feature extraction and machine learning classification models;

[0047] S3: Obtain the corresponding real-time power based on the real-time power consumption status of each enterprise's equipment, and combine it with the corresponding carbon emission intensity factor to calculate the total direct carbon emissions of the enterprise's equipment in real-time operation.

[0048] S4: Obtain the real-time marginal carbon emission factor of the power grid operation through the regional power grid dispatch system, and combine it with the real-time power of each enterprise's equipment to calculate the indirect carbon emissions caused by the enterprise's electricity consumption behavior.

[0049] S5: Perform real-time summary and analysis of the total direct and indirect carbon emissions obtained in steps S3 and S4 to obtain the overall real-time carbon emissions of the enterprise, and generate optimization suggestions to guide enterprise users to implement effective energy conservation and emission reduction measures.

[0050] In step S1, the real-time collection of the enterprise's overall power load data using non-invasive monitoring equipment specifically involves:

[0051] A single-point load monitoring device is installed at the main circuit inlet of the enterprise. The device records the overall power load data of the enterprise and performs data cleaning.

[0052] Preferably, the method further includes dividing the cleaned enterprise's overall electricity load data into different load types using load curve analysis technology and clustering algorithms; extracting the characteristic load curves of each type of load equipment under typical operating modes; obtaining regional electricity carbon emission intensity data; calculating the typical carbon emission levels of different types of loads at different time periods; and establishing an enterprise electricity carbon emission benchmark data model for comparison with the enterprise's overall real-time carbon emissions to formulate energy-saving behavior adjustment and carbon footprint control strategies for the enterprise.

[0053] This embodiment specifically includes the following sub-steps:

[0054] S11: Collecting enterprise electricity consumption data is fundamental for real-time carbon emission estimation. To fully reflect the actual electricity consumption behavior and equipment operating status of enterprise users, this invention employs Non-Intrusive Power Load Monitoring (NILM) technology for data collection. Specifically, a single-point load monitoring device is installed at the enterprise's main circuit inlet. This device records the enterprise's overall electricity consumption data in real time, including but not limited to key electrical parameters such as voltage, current, total active power, reactive power, and power factor. It is recommended that the data collection time granularity be at the second or minute level to capture the instantaneous characteristics of equipment startup and shutdown, as well as load fluctuations. Furthermore, to ensure that the monitoring data accurately reflects the seasonal changes and daily patterns of the enterprise's load, it is recommended that the data collection time span cover a typical annual period. Through the above implementation method, high-quality, high-time-granularity overall enterprise electricity consumption data can be obtained, laying a reliable data foundation for subsequent load identification and carbon emission analysis.

[0055] S12: Based on the high-precision electricity consumption data collected in step S11, this invention further preprocesses and performs feature analysis on the data. First, the raw data is cleaned to remove noise and missing data caused by equipment malfunctions, external interference, etc. If necessary, missing data is supplemented using interpolation methods to ensure data integrity and consistency. Second, using load curve analysis technology and clustering algorithms, the total load data of enterprise electrical equipment is divided into different load types, including rigid loads (such as air conditioners and water heaters with fixed operating times and difficult to adjust flexibly) and transferable loads (such as washing machines and rice cookers with flexible operating times and adjustable capabilities). Then, characteristic load curves of each type of load equipment under typical operating modes are extracted, and combined with regional electricity carbon emission intensity data obtained from power grid companies or government departments, the typical carbon emission levels of different types of loads at different time periods are calculated. Finally, based on the above results, a complete benchmark data model for enterprise electricity carbon emissions is established, providing accurate reference for real-time carbon emission monitoring and subsequent optimized scheduling.

[0056] In step S2, the feature extraction and machine learning classification model processing includes:

[0057] The overall electricity load data of the enterprise is preprocessed. The preprocessing operation includes data denoising, normalization and feature extraction. The features extracted include steady-state features and transient features.

[0058] The extracted features are input into the trained deep learning load identification model to obtain the real-time power consumption status of each enterprise's equipment; the training process of the deep learning load identification model includes:

[0059] The deep learning load identification model is iteratively trained using labeled historical enterprise overall electricity load data, and the generalization performance of the model is evaluated periodically during the training process using cross-validation and early stopping of training.

[0060] This embodiment specifically includes the following sub-steps:

[0061] S21: Non-invasive load identification modeling is a crucial foundation for accurate identification of enterprise electrical equipment. First, the overall enterprise electricity consumption data collected in real-time in step S1 is preprocessed, including data denoising, normalization, and feature extraction. Specifically, the extracted features include steady-state features (such as active power, reactive power, apparent power, and power factor) and transient features (such as starting current spikes and power gradient changes). Subsequently, this invention employs deep learning algorithms to process and analyze these load features to construct a load identification model. The preferred deep learning algorithm is a combination of Long Short-Term Memory (LSTM) or Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BLSTM). The CNN module efficiently extracts temporal and frequency domain features from the load data, while the BLSTM module effectively captures the bidirectional time dependence of the load data, improving the accuracy and generalization ability of the identification.

[0062] S22: Based on the deep learning load identification model constructed in step S21, train and optimize the enterprise's electrical equipment classification model. The specific implementation includes the following processes: (1) Use the labeled historical dataset to train the identification model, and obtain a stable and reliable classification model through repeated iterations and model parameter optimization; (2) To avoid overfitting, cross-validation and early stopping of training are adopted during the training process, and the generalization performance of the model is evaluated regularly; (3) After the model training is completed, the real-time collected overall electricity consumption data is input, and the model can output the energy consumption status and operating mode of each equipment in the enterprise in real time, realizing real-time and accurate identification of the energy consumption of each equipment; (4) The model is verified and updated regularly to adapt to dynamic scenarios such as the addition or replacement of equipment or changes in electricity consumption habits, and to maintain the high accuracy and strong generalization performance of the load identification model. This implementation method can provide accurate energy consumption data at the equipment level for subsequent real-time carbon emission calculation.

[0063] In this embodiment, step S3 specifically includes the following steps:

[0064] S31: Real-time calculation of direct carbon emissions is based on refined analysis results from non-intrusive load identification. Specifically, firstly, based on the real-time power consumption status of each enterprise's equipment identified in step S2, real-time operating power data for each piece of equipment is obtained. Secondly, a database of carbon emission intensity factors corresponding to the equipment is obtained and maintained from authoritative institutions (such as power grid companies or government departments). These factors represent the direct carbon emissions (usually kg CO2 / kWh) corresponding to each unit of electricity consumed by various types of equipment. It is recommended that the factor database be updated regularly to reflect the impact of factors such as changes in power supply structure and equipment technology improvements on emission intensity, ensuring the accuracy and reliability of carbon emission calculations.

[0065] S32: Based on the real-time operating power data and corresponding carbon emission factors of the equipment, the total direct carbon emissions of various types of equipment are calculated using a real-time multiplicative accumulation method. Specifically, the real-time carbon emission calculation formula is as follows:

[0066] (1)

[0067] in, For enterprise equipment within a time range The total direct carbon emissions within the region, expressed in kgCO2; This represents the real-time power of the nth device at the current moment, in kW; This represents the real-time carbon emission intensity factor for the nth device, expressed in kgCO2 / kWh. d This indicates the number of identified devices within the enterprise. Using the above methods, accurate real-time calculation of a company's direct carbon emissions can be achieved, providing strong data support for assessing low-carbon electricity consumption behavior among enterprise users.

[0068] Step S4 specifically includes the following sub-steps:

[0069] S41: Real-time acquisition of marginal carbon emission factors of regional power grids:

[0070] Based on real-time electricity market clearing data, the real-time marginal carbon emission factor (MCEF) of the power grid operation is obtained through the regional power grid dispatching system. Specifically, based on the real-time output data of different generating units within the region, combined with the corresponding carbon emission characteristic curves of the units, the real-time changing marginal carbon emission factor of the power grid is calculated to accurately reflect the indirect carbon emission level caused by changes in electricity load, providing accurate data support for the real-time estimation of indirect carbon emissions from enterprise electricity consumption. The calculation formula is as follows:

[0071]

[0072] in, This represents the active power output of unit i at time t; The emission characteristic curve of unit A is a function of the emission rate per unit time as a function of the output. This indicates the marginal emission rate of unit A; This indicates the real-time load / net output demand of the region; This represents the participation factor (AGC / dispatch allocation factor) of the generating unit for 1 unit load increment. For the unit t Carbon emission intensity at any given time.

[0073] S42: Real-time calculation of indirect carbon emissions by enterprises:

[0074] Using the real-time overall electricity consumption data of the enterprise obtained in step S1, and combined with the regional power grid marginal carbon emission factor obtained in step S41, the indirect carbon emissions caused by changes in the enterprise's electricity load are calculated through real-time multiplication. Specifically, this includes multiplying the enterprise's real-time electricity load with the regional power grid's real-time marginal carbon emission factor point by point to obtain the indirect carbon emission value at the corresponding time, and accumulating the calculation results at each time in real time, thereby realizing real-time, dynamic quantitative monitoring of the enterprise's indirect carbon emissions and supporting comprehensive carbon emission analysis of electricity consumption behavior.

[0075] (2)

[0076] in, For enterprise equipment within a time range Indirect carbon emissions within the region, expressed in kgCO2; For the first t Time of the first i The real-time marginal carbon emission factor of the connected devices is expressed in kgCO2 / kWh. Through the above methods, the indirect carbon emissions of enterprises can be accurately calculated in real time, providing strong data support for assessing the low-carbon electricity consumption behavior of enterprise users.

[0077] Preferably, the method also includes a comprehensive analysis based on the enterprise's overall real-time carbon emission data and the enterprise's electricity consumption behavior, from the dimensions of current electricity consumption structure, load period characteristics and high carbon emission behavior, to generate personalized and scenario-based energy-saving and carbon reduction optimization suggestions.

[0078] Specifically, step S5 includes the following sub-steps:

[0079] S51: Carbon Emission Data Aggregation and Real-Time Assessment Model Construction

[0080] After completing the real-time calculation of direct and indirect carbon emissions, this step introduces a carbon emission data fusion and assessment mechanism to comprehensively understand the enterprise-level carbon emission status. By constructing a real-time carbon emission assessment model, the direct carbon emissions of each electrical device obtained in step S3 and the overall indirect carbon emissions estimated in step S4 are integrated in a time series manner to form the enterprise's full lifecycle time-series carbon emission data. This model, based on a weighted fusion algorithm, incorporates equipment usage intensity, time sensitivity factors, and regional power grid carbon intensity fluctuation parameters, assigning dynamic weights to direct and indirect carbon emissions respectively, thereby more scientifically reflecting the enterprise's carbon emission performance under different energy consumption scenarios. The model supports minute-level updates, has dynamic response characteristics, and can capture carbon emission change trends in real time. Simultaneously, the system also integrates a historical emission benchmark database, enabling comparison of carbon emission levels with the same period in the past, different seasons, or user-defined time periods, generating visual analysis charts to support users in adjusting energy-saving behaviors and formulating carbon footprint control strategies. The dynamic weight calculation formula is as follows:

[0081]

[0082] in, Indicates dynamic weighted carbon emissions; and This represents direct / indirect carbon emissions (e.g., on-site fuel and electricity purchase); the basic weight is: .

[0083] S52: Generation and Push of Personalized Energy Saving and Carbon Reduction Optimization Suggestions

[0084] After completing the real-time integration and evaluation of enterprise carbon emission data, this step generates personalized, scenario-based energy-saving and carbon-reduction optimization suggestions based on dynamic carbon emission data and electricity consumption behavior analysis results to further enhance the system's practical value and promote user participation in emission reduction. This suggestion system comprehensively analyzes multiple dimensions, including current electricity consumption structure (such as the proportion of high-carbon equipment such as air conditioners and water heaters), load period characteristics (such as peak-valley electricity consumption ratio), and high-carbon-emission behaviors (such as prolonged operation of high-power equipment). It combines local carbon emission factor fluctuation patterns and weather forecasts to output highly executable control strategies. For example, it suggests that users operate high-energy-consuming equipment during periods of lower marginal carbon emission factors, or prompts users to turn off equipment that has been idle for extended periods and adjust air conditioner temperature settings. The system also provides a simulation function for emission reduction effects, allowing users to view the carbon emission reduction corresponding to different behavioral adjustments, thus enhancing the incentive effect of emission reduction behavior. Furthermore, optimization suggestions are pushed to users in real-time via mobile apps or enterprise smart terminals in the form of charts, voice, or text, constructing a "data-suggestion-feedback" closed loop to guide residents to actively participate in energy conservation and carbon reduction.

[0085] Example 2

[0086] This embodiment provides a carbon emission estimation system based on non-intrusive power load monitoring, including a memory and a processor. The memory stores a computer program, and the processor calls the computer program to execute the steps of the carbon emission estimation method based on non-intrusive power load monitoring as described in Embodiment 1.

[0087] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for carbon emission estimation based on non-intrusive power load monitoring, characterized in that, The method comprises the following steps: real-time acquisition of enterprise overall power consumption load data by a non-intrusive monitoring device; obtaining real-time power consumption states of various enterprise devices according to the acquired enterprise overall power consumption load data through a pre-trained feature extraction and machine learning classification model; obtaining corresponding real-time power according to the real-time power consumption states of various enterprise devices, and combining the real-time power with corresponding carbon emission intensity factors to calculate the total direct carbon emission of enterprise devices in real-time operation; obtaining the real-time marginal carbon emission factor of power grid operation through a regional power grid dispatching system, and combining the real-time power of various enterprise devices to calculate the indirect carbon emission of enterprises caused by power consumption behavior; real-time collection and analysis of the obtained total direct carbon emission and indirect carbon emission to obtain the real-time carbon emission of the enterprise as a whole; the calculation expression of the indirect carbon emission of the enterprise caused by power consumption behavior is: wherein, is the indirect carbon emission of the enterprise device in the time interval , is the real-time marginal carbon emission factor of the t 𝑖th device at the time instant i , is the real-time power of the t 𝑖th device at the current time instant d denotes the number of identified devices in the enterprise; the obtaining process of the real-time marginal carbon emission factor is specifically: obtaining real-time output data of different generating units in the region where the enterprise is located through a regional power grid dispatching system, and combining the carbon emission characteristic curve of the corresponding generating unit to calculate the real-time marginal carbon emission factor which changes in real time; the calculation expression of the real-time marginal carbon emission factor is: In the formula, is t the moment i the real-time marginal carbon emission factor of the access of the device, is the emission characteristic curve of the unit j is the marginal emission rate of the unit is the emission characteristic curve of the unit j is the marginal emission rate of the unit 2. The method of estimating carbon emission based on non-intrusive power load monitoring according to claim 1, wherein, the calculation expression of the total direct carbon emission of the enterprise devices in real-time operation is: wherein, is the total amount of direct carbon emissions of the enterprise devices within the time interval ; represents the real-time power of the 𝑖th device at the current time t ; is the real-time carbon emission intensity factor corresponding to the 𝑖th device; d represents the number of identified devices in the enterprise.

3. The method of estimating carbon emission based on non-intrusive power load monitoring according to claim 1, wherein, The method assigns dynamic weights to the total direct carbon emission and indirect carbon emission respectively according to the device usage intensity, time sensitivity factor and regional power grid carbon intensity fluctuation parameter, and performs weighted fusion; the calculation expression of the weighted fusion is: In the formula, is the total amount of direct carbon emissions, is the total amount of direct carbon emissions, is the amount of indirect carbon emissions, is the total amount of direct carbon emissions weight, is the amount of indirect carbon emissions weight, .

4. The method of estimating carbon emission based on non-intrusive power load monitoring according to claim 1, wherein, The real-time acquisition of enterprise overall power consumption load data by a non-intrusive monitoring device is specifically: installing a single-point load monitoring device at the total entrance of the enterprise circuit, recording the enterprise overall power consumption load data through the single-point load monitoring device, and performing data cleaning.

5. The method for carbon emission estimation based on non-intrusive power load monitoring according to claim 4, characterized in that, The method further comprises dividing the enterprise overall power consumption load data after data cleaning into different load types by using load curve analysis technology and clustering algorithm; extracting the characteristic load curve of each type of load device under the typical operation mode, obtaining regional power carbon emission intensity data, calculating the typical carbon emission level of different types of load in different time periods, establishing an enterprise power carbon emission benchmark data model, and comparing the enterprise overall real-time carbon emission to formulate the energy-saving behavior adjustment and carbon footprint control strategy of the enterprise.

6. The method of estimating carbon emission based on non-intrusive power load monitoring according to claim 1, wherein, The processing process of the feature extraction and machine learning classification model comprises: preprocessing the enterprise overall power consumption load data, which includes data denoising, normalization processing and feature extraction, wherein the extracted features include steady-state features and transient-state features; inputting the extracted features into a trained deep learning load identification model to obtain the real-time power consumption state of each enterprise device; the training process of the deep learning load identification model comprises: iteratively training the deep learning load identification model using the labeled historical enterprise overall power consumption load data, and periodically evaluating the generalization performance of the model during the training process by using cross-validation and early stopping training methods.

7. The method of estimating carbon emission based on non-intrusive power load monitoring as claimed in claim 1, wherein, The method further comprises comprehensively analyzing the current power consumption structure, load period characteristics and high-carbon emission behavior dimensions according to the overall real-time carbon emission data of the enterprise and the power consumption behavior of the enterprise, and generating personalized and scenario-based energy saving and carbon reduction optimization suggestions.

8. A carbon emission estimation system based on non-intrusive power load monitoring, characterized by, The computer program product comprises a memory and a processor, the memory stores a computer program, and the processor calls the computer program to execute the steps of the method according to any one of claims 1 to 7.

Images