Electricity carbon metering system and method based on multimode data fusion
By using a multi-mode data fusion system, the problems of data heterogeneity and insufficient feature extraction in electricity carbon metering have been solved, and the nonlinear coupling relationship between electricity consumption behavior and carbon emissions has been accurately characterized, thereby improving the accuracy and robustness of carbon emission monitoring in the power system.
Patent Information
- Application Number
- CN202511513095.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-22
- Publication Date
- 2026-01-16
AI Technical Summary
Existing methods for measuring carbon emissions have shortcomings in terms of data heterogeneity, insufficient feature extraction capabilities, and inadequate adaptability to operating conditions, resulting in insufficient measurement accuracy and robustness, and failing to meet the precise measurement needs under the influence of multiple variables.
A multi-modal data fusion system is adopted, which achieves deep fusion and nonlinear correlation representation of multi-modal data by acquiring, preprocessing, extracting features, constructing tensors, rotating and transforming, and recursively reconstructing features, combined with dynamic carbon emission factors, so as to dynamically adapt to changes in operating conditions.
It achieves deep fusion of heterogeneous multimodal data, accurately captures the complex nonlinear coupling relationship between electricity consumption behavior and carbon emissions, and improves the accuracy and robustness of carbon emission monitoring in the power system.
Smart Images
Figure CN121350985A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of electric carbon metering, and in particular to an electric carbon metering system and method based on multi-modal data fusion. BACKGROUND
[0002] Under the background of the global "double carbon" strategic goal, the accurate monitoring and management of carbon emissions of the power system has become a core task in the process of energy transformation and low-carbon governance, which not only relates to the realization of regional and even global carbon emission reduction targets, but also has important significance for the green and sustainable development of the power industry. Therefore, the related electric carbon metering technology research and application have attracted widespread attention.
[0003] At present, the widely used electric carbon metering method in the industry is mainly based on traditional linear estimation. This method obtains the electricity consumption data through the electric energy meter, and performs linear multiplication operation with the fixed regional carbon emission coefficient to obtain the estimated result of carbon emissions. The advantage of this method is that the calculation logic is simple, the operation threshold is low, and the basic carbon emission accounting can be quickly completed. However, in actual application, there are obvious limitations: on the one hand, the regional carbon emission coefficient is influenced by factors such as power grid energy structure, time period power supply and demand relationship, and renewable energy generation proportion, and is in dynamic change. Using a fixed value will result in a large deviation between the estimated result and the actual carbon emission. On the other hand, this method only relies on single electricity consumption data, does not take into account environmental parameters such as temperature, humidity, and illumination, and does not consider the differences in operating characteristics of different load devices, which cannot reveal the complex coupling relationship between electricity consumption behavior and carbon emissions, and cannot meet the precise metering requirements under the influence of multiple variables.
[0004] In order to make up for the shortcomings of the traditional method, the electric carbon metering technology that introduces multi-modal data has gradually developed in the industry. This method attempts to integrate electric energy data, environmental parameter data, and equipment operating state data and other multi-source information, and expects to improve the metering accuracy through the synergistic effect of multi-dimensional data. However, from the actual application effect, this kind of technology still faces the technical problems to be solved:
[0005] Firstly, the multi-modal data has problems of structural heterogeneity and time sequence misalignment. For example, electric energy data is mostly sampled at the second level, environmental parameters are mostly sampled at the minute level, and equipment operating state data is often recorded intermittently. The dimension specifications and time scales of different data are not unified, and traditional data splicing or simple alignment methods cannot achieve deep fusion, resulting in a lack of consistent physical representation basis for subsequent modeling.
[0006] Secondly, the feature extraction capability of multi-modal data is insufficient. The existing technology mostly uses simple dimension reduction or linear transformation methods to process raw data, and cannot effectively capture the nonlinear relationship between load behavior, environmental disturbance, and energy consumption characteristics. Especially in the load type switching or environmental parameter mutation scene, the accuracy and stability of feature expression are insufficient, which affects the robustness of the metering result.
[0007] Furthermore, when establishing the correlation between energy consumption and carbon emissions, simply mapping electrical parameters such as voltage and current to establish a relationship with carbon factors cannot adapt to the differences in energy consumption-carbon emission response characteristics under different operating conditions, further restricting the improvement of measurement accuracy. Summary of the Invention
[0008] This invention proposes an electric carbon metering system and method based on multi-modal data fusion. Its purpose is to realize an electric carbon metering method that can effectively fuse heterogeneous time-series multi-modal data, accurately capture the complex nonlinear coupling relationship between electricity consumption behavior and carbon emissions, and dynamically adapt to different operating conditions, so as to significantly improve the accuracy and robustness of carbon emission monitoring in the power system.
[0009] The technical solution of this invention is as follows:
[0010] A multi-mode data fusion-based carbon metering system includes a multi-mode data acquisition module, a data preprocessing and feature extraction module, a tensor construction module, a tensor transformation module, a recursive feature reconstruction module, a dynamic carbon emission factor construction module, a chimeric mapping processing module, and a carbon metering estimation module.
[0011] The raw multi-mode data obtained by the multi-mode data acquisition module is sequentially processed by the data preprocessing and feature extraction module, tensor construction module, tensor transformation module, and recursive feature reconstruction module to obtain recursive reconstruction vectors. The chimeric mapping processing module obtains the chimeric mapping vectors of each electrical device based on the recursive reconstruction vectors and the dynamic carbon emission factor constructed by the dynamic carbon emission factor construction module. The electricity carbon metering estimation module obtains the electricity carbon metering estimation results based on the chimeric mapping vectors of each electrical device.
[0012] As a further improvement to the aforementioned electricity carbon metering system based on multi-mode data fusion: a multi-mode data acquisition module is used to synchronously acquire raw multi-mode data through a multi-mode sensor acquisition device deployed at the electricity terminal, and send the raw multi-mode data to the data preprocessing and feature extraction module;
[0013] The data preprocessing and feature extraction module is used to preprocess the original multi-modal data, then extract features from the preprocessed multi-modal data, and then perform dimensionality reduction, normalization and standardization to obtain multi-modal data vectors, which are then sent to the tensor construction module.
[0014] As a further improvement to the multi-mode data fusion-based carbon metering system: the tensor construction module is used to construct tensors from the multi-mode data vectors output by the data preprocessing and feature extraction module using the outer product operation, and then send the constructed tensors to the tensor transformation module.
[0015] As a further improvement to the aforementioned carbon metering system based on multi-mode data fusion: the tensor transformation module is used to construct a tensor consistency mapping by introducing rotation tensor operations based on the constructed tensor, so as to ensure that the modes have temporal mapping isomorphism, remove the nonlinear coupling bias between modes, and obtain the basic mode rotation vector.
[0016] As a further improvement to the aforementioned multi-mode data fusion-based carbon metering system, the recursive feature reconstruction module is used to introduce a recursive feature reconstruction mechanism to reconstruct the basic mode rotation vector, obtain the recursive reconstruction vector, and send it to the chimeric mapping processing module.
[0017] As a further improvement to the aforementioned electric carbon metering system based on multi-mode data fusion: the dynamic carbon emission factor construction module is used to construct dynamic carbon emission factors and send them to the chimeric mapping processing module;
[0018] The chimeric mapping processing module is used to perform recursive mode mapping based on the recursive reconstruction vector and the constructed dynamic carbon emission factor to form the final feature chimeric representation, which serves as the chimeric mapping vector for each electrical device and is sent to the electricity carbon metering estimation module.
[0019] 7. The carbon metering system based on multi-mode data fusion as described in claim 6, characterized in that: the carbon metering estimation module obtains the carbon metering estimation result by introducing the carbon metering estimation formula for explicit calculation, thereby realizing carbon metering based on multi-mode data fusion.
[0020] A method for metering carbon electricity based on multi-mode data fusion includes the following steps:
[0021] S1. Obtain the original multimodal data, perform feature extraction and dimensionality reduction after preprocessing to obtain multimodal data vectors, construct tensor representation based on multimodal data vectors, introduce rotation angle vectors, perform rotation transformation to obtain basic modal rotation vectors, and then introduce a recursive feature reconstruction mechanism to obtain recursive reconstruction vectors.
[0022] S2. Based on the dynamic carbon emission factor constructed by recursive reconstruction vector combination, recursive mode mapping is performed to form the final feature-based embedding representation, which serves as the embedding mapping vector for each electrical device. Then, the carbon emission estimation formula is introduced for explicit calculation to obtain the carbon emission measurement result, thus realizing carbon emission measurement through multi-mode data fusion.
[0023] As a further improvement to the multi-mode data fusion-based carbon metering method, step S1 specifically includes:
[0024] S1-1. Acquire raw multi-mode data synchronously through a multi-mode sensor acquisition device deployed at the power consumption terminal;
[0025] S1-2. Preprocess the raw multi-modal data;
[0026] S1-3. After feature extraction of the preprocessed multi-mode data, dimensionality reduction is performed to obtain multi-mode data vectors, which include electrical parameter mode vectors, environmental mode vectors, and equipment state mode vectors.
[0027] S1-4. Perform tensor construction operation on multimodal data vectors, and combine vectors of different modes into a higher-order tensor through outer product operation, that is, construct tensor;
[0028] S1-5. Introduce rotation tensor operations to construct tensor consistency mapping to ensure temporal mapping isomorphism between modes, remove nonlinear coupling bias between modes, obtain rotation tensor after rotation, expand rotation tensor into vector form to obtain basic mode rotation vector;
[0029] S1-6. A recursive feature reconstruction mechanism is introduced, which performs multi-level recursive calculations based on the basic mode rotation vector. The influence of ambient humidity and light intensity is considered during the calculation process to obtain the recursive reconstruction vector.
[0030] As a further improvement to the multi-mode data fusion-based carbon metering method, step S2 specifically includes:
[0031] S2-1. Based on the recursive reconstruction vector, combined with the constructed dynamic carbon emission factor, recursive mode mapping is performed to form the final feature embedding representation, which serves as the embedding mapping vector for each electrical device.
[0032] The calculation process for dynamic carbon emission factors incorporates the region's renewable energy ratio and electricity intensity per unit area;
[0033] The calculation process of the chimeric mapping vector incorporates the instantaneous voltage and instantaneous current values of the electrical equipment;
[0034] S2-2. The embedded mapping vector of each electrical device is used as a complete input for the final carbon emission estimation. The carbon emission is calculated by introducing the electrical carbon estimation formula.
[0035] The carbon dioxide estimation formula incorporates a weighting mechanism for electrical equipment.
[0036] Compared with the prior art, the present invention has the following beneficial effects:
[0037] 1. This invention transforms heterogeneous data with different dimensions and sampling frequencies, such as electrical parameters, environmental parameters, and equipment status, into a unified vector space representation with structural alignment and consistent physical meaning through tensor construction and rotation tensor transformation. This achieves deep and structured fusion of heterogeneous multimodal data, thereby fundamentally solving the information fragmentation and distortion problems caused by simple data splicing or alignment in traditional methods, and laying a solid foundation for subsequent accurate modeling.
[0038] 2. This invention further introduces a recursive feature reconstruction mechanism. This mechanism integrates multiple factors such as time-based incentives, environmental regulation, and energy constraints to perform multiple rounds of nonlinear iterative optimization of features. This ensures that the final feature vector not only accurately represents the current state but also effectively captures the dynamic evolution trend of electricity consumption behavior and its response pattern to external disturbances. It effectively captures the nonlinear correlation between load behavior, environmental disturbances, and energy consumption characteristics. The recursive reconstruction vector formed through this iterative approach not only possesses high expressive power for the current state but also reflects the nonlinear interlocking characteristics between time-based evolution trends, abnormal disturbance responses, and energy regulation structures. Therefore, it maintains stable and accurate metering performance even under complex scenarios such as load type switching and sudden changes in environmental parameters, significantly improving the model's representational ability and robustness under complex operating conditions.
[0039] 3. This invention abandons the approach of using fixed regional carbon emission coefficients and innovatively constructs a dynamic carbon emission factor. This factor can reflect changes in key variables such as the proportion of renewable energy in a region, electricity intensity, and time period in real time, enabling the metrology model to adapt to the dynamic evolution of the power grid's operating status.
[0040] 4. This invention significantly enhances the characterization ability of the response characteristics between energy consumption and carbon emissions under different operating conditions by embedding and mapping dynamic carbon emission factors with recursive features and introducing joint modulation of parameters such as voltage, current, and carbon factor through nonlinear combination. Attached Figure Description
[0041] Figure 1 This is a schematic diagram of the structure of the electric carbon metering system based on multi-mode data fusion in this invention;
[0042] Figure 2 This is a flowchart of the carbon metering method based on multi-mode data fusion in this invention. Detailed Implementation
[0043] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0045] Example 1
[0046] This embodiment provides an electric carbon metering system based on multi-mode data fusion, the structure of which is as follows: Figure 1 As shown, the system includes: a multi-mode data acquisition module, a data preprocessing and feature extraction module, a tensor construction module, a tensor transformation module, a recursive feature reconstruction module, a dynamic carbon emission factor construction module, a chimeric mapping processing module, and an electric carbon metering and estimation module.
[0047] The raw multi-mode data obtained by the multi-mode data acquisition module is sequentially processed by the data preprocessing and feature extraction module, tensor construction module, tensor transformation module, and recursive feature reconstruction module to obtain recursive reconstruction vectors. The chimeric mapping processing module obtains the chimeric mapping vectors of each electrical device based on the recursive reconstruction vectors and the dynamic carbon emission factor constructed by the dynamic carbon emission factor construction module. The electricity carbon metering estimation module obtains the electricity carbon metering estimation results based on the chimeric mapping vectors of each electrical device.
[0048] Specifically:
[0049] The multi-mode data acquisition module synchronously acquires raw multi-mode data through a multi-mode sensor acquisition device deployed at the power consumption terminal, and sends the raw multi-mode data to the data preprocessing and feature extraction module.
[0050] The data preprocessing and feature extraction module preprocesses the original multi-modal data, extracts features from the preprocessed multi-modal data using feature engineering techniques such as statistical analysis, and performs dimensionality reduction, normalization, and standardization using methods such as principal component analysis to obtain multi-modal data vectors, which are then sent to the tensor construction module.
[0051] The tensor construction module constructs tensors from multimodal data vectors using outer product operations, obtains the constructed tensors, and sends them to the tensor transformation module.
[0052] The tensor transformation module introduces rotation tensor operations based on the constructed tensor to construct a tensor consistency mapping, ensuring temporal mapping isomorphism between modes, removing nonlinear coupling bias between modes, and obtaining the rotation vector of the basic mode.
[0053] The recursive feature reconstruction module introduces a recursive feature reconstruction mechanism to reconstruct the basic mode rotation vector, obtain the recursive reconstruction vector, and send it to the chimeric mapping processing module.
[0054] The dynamic carbon emission factor construction module constructs dynamic carbon emission factors based on basic information such as the proportion of regional renewable energy and regional electricity intensity, and sends them to the chimeric mapping processing module.
[0055] The chimeric mapping processing module, based on the recursive reconstruction vector and combined with the constructed dynamic carbon emission factor, performs recursive mode mapping to form the final feature chimeric representation, which serves as the chimeric mapping vector for each electrical device and is sent to the electricity carbon metering and estimation module.
[0056] The carbon metering estimation module uses an explicit calculation based on the carbon metering estimation formula to obtain the carbon metering estimation result, thus achieving carbon metering based on multi-mode data fusion.
[0057] Example 2
[0058] This embodiment proposes an electric carbon metering method based on multi-mode data fusion, the flowchart of which is as follows: Figure 2 As shown, the method includes the following steps:
[0059] S1. Obtain the original multimodal data, perform feature extraction and dimensionality reduction after preprocessing to obtain multimodal data vectors, construct tensor representations based on multimodal data vectors, introduce rotation angle vectors, perform rotation transformation to obtain basic modal rotation vectors, and then introduce a recursive feature reconstruction mechanism to obtain recursive reconstruction vectors.
[0060] The specific process is as follows:
[0061] S1-1. Acquire raw multi-mode data synchronously through a multi-modal sensor acquisition device deployed at the power consumption terminal. The raw multi-mode data includes electrical parameters such as voltage, current, and power factor, environmental parameters such as temperature and humidity, and load data such as load type label vector.
[0062] S1-2. Perform preprocessing on the original multi-modal data, such as cleaning, denoising, outlier handling, standardization, and normalization, to obtain preprocessed multi-modal data.
[0063] The preprocessing procedures all employ techniques well-known to those skilled in the art, and will not be elaborated upon here.
[0064] S1-3. After extracting features from the preprocessed multi-mode data using feature engineering techniques such as statistical analysis, and then performing dimensionality reduction, normalization, and standardization using methods such as principal component analysis, the multi-mode data vector is obtained, including the electrical parameter mode vector. Environment mode vector Device state mode vector .
[0065] S1-4. To avoid strong heterogeneity in temporal sequence, distribution, and physical meaning among multimodal data vectors, a tensor construction operation is performed on the multimodal data vectors. This involves combining vectors of different modes into a higher-order tensor through outer product operations, i.e., constructing the tensor. The specific formula is as follows:
[0066] ;
[0067] in, It is the constructed tensor, which contains the cross-interaction relationships of all modal variables within the same sampling time window; It is an outer product operation that combines vectors of different modalities into a higher-order tensor to explicitly store the interaction relationships between variables of different modalities.
[0068] S1-5. To avoid the lack of global structural consistency in the positional representation of the tensor constructed above, a rotation tensor operation is introduced to construct a tensor consistency mapping, ensuring temporal mapping isomorphism between modes, removing nonlinear coupling biases between modes, and obtaining a rotation tensor after rotation. Expanding the rotation tensor into vector form yields the rotation vector of the fundamental mode. .
[0069] This step introduces a rotation angle vector. ,like For the above tensor A rotation transformation is performed to preserve the nonlinear relative positions of each dimension of the tensor, resulting in a rotated tensor. This rotated tensor is then expanded into a vector form, yielding the unified eigenvectors after the modal rotation mapping, which are the fundamental modal rotation vectors. .
[0070] The technical objective of this step is to preserve the coupling information between modes while ensuring that they have a consistent spatial structure that can be processed, thus laying the foundation for subsequent feature reconstruction.
[0071] S1-6. Obtaining the fundamental mode rotation vector Subsequently, to enable it to have recursive feature evolution capability, a recursive feature reconstruction mechanism was introduced, based on the fundamental mode rotation vector. Multi-level recursive calculations are performed, taking into account the effects of ambient humidity and light intensity, to obtain the recursive reconstruction vector. .
[0072] The recursive feature reconstruction mechanism is implemented through the following formula:
[0073] ;
[0074] in, It is the first The result of the layer recursive reconstruction; It is a recursion level index. This is the total number of recursion levels, determined according to specific needs, such as 3 to 5. It is the first The result of the layer recursive reconstruction, in the first iteration, Take the fundamental mode rotation vector ; It is the first The time-based activation factor of the layer, used to control time-periodic activation, is determined based on expert experience, with a reference value range of [value missing]. ; It is a local time point at the modal sampling moment, a time attribute during multi-modal data sampling, specifically the hour value in the local timestamp (GNSS time synchronization or NTP synchronization), with a value of... ; It is the first The environmental regulation factor of the layer, used to control the intensity of the humidity effect, is determined based on expert experience, with a reference range of values. ; It is the standardized real-time ambient humidity, obtained through a humidity sensor; It is the standardized real-time light intensity, obtained through a light sensor; It is the first The power exponential attenuation factor of the layer, used to suppress the influence of high power, is determined based on expert experience, with a reference range of values. ; It is the current standardized instantaneous active power, obtained through an energy meter; It is an L1 norm; It is an L2 norm; Used to perform nonlinear synthesis of the current characteristic state and time-modulated excitation; Used to achieve nonlinear suppression and characteristic stability control of environmental factors; It is an energy constraint term used to perform exponential suppression under high power loads and high feature energy densities, preventing the feature from exploding after multiple recursions.
[0075] After a predetermined number of recursive levels, the recursive reconstruction vector is obtained. .
[0076] S2. Based on the dynamic carbon emission factor constructed by recursive reconstruction vector combination, recursive mode mapping is performed to form the final feature-based embedding representation, which serves as the embedding mapping vector for each electrical device. Then, the carbon emission estimation formula is introduced for explicit calculation to obtain the carbon emission measurement result, thus realizing carbon emission measurement through multi-mode data fusion.
[0077] The specific process is as follows:
[0078] S2-1. Based on the recursive reconstruction vector and combined with the constructed dynamic carbon emission factor, recursive mode mapping is performed to form the final feature-based embedding representation, which serves as the embedding mapping vector for each electrical device. The calculation process of the dynamic carbon emission factor incorporates the region's renewable energy ratio and electricity intensity per unit area. The calculation process of the embedding mapping vector incorporates the instantaneous voltage and instantaneous current values of the electrical devices.
[0079] The dynamic carbon emission factor for the region is calculated as follows:
[0080] ;
[0081] in, It is in time Area code Regional power consumption per unit area Dynamic carbon emission factors under; It is a global time variable of the carbon factor, used to drive the periodic structure of the carbon emission factor over time, and represents the center moment of the current calculation cycle; It is a region code, either the State Grid regional division or the ISO region code; The area code is The proportion of renewable energy in a region is obtained in real time from an energy structure database or a power grid dispatch center. It is the angular frequency of the periodic excitation, determined according to expert experience, such as... ; This is the time decay coefficient, used to adjust the time-series decay of the carbon factor. It is determined based on expert experience, and the reference range is [range missing]. ; It is the regional electricity intensity per unit area, obtained from existing energy consumption maps or load density maps. The electricity intensity represents the regional electricity consumption density, that is, the ratio of the total electricity consumption in a certain region to the area of that region within a statistical period. This refers to the average electricity intensity of the reference area, which is determined based on the specific application scenario and will not be elaborated here. It represents the square root of the proportion of renewable energy (such as wind power, solar power, and hydropower) in the total power supply in the current region. The square root is used to gradually reduce sensitivity when the proportion is high. It simulates the periodic correlation between daily 24-hour electricity supply and demand and carbon emission factors; It describes that the higher the electricity intensity, the more the regional energy dispatch relies on peak-shaving power sources (such as thermal power units), and therefore the carbon factor increases accordingly; This indicates the natural decay of carbon emission factors over a continuous period of time.
[0082] For any electrical device within this area, its chimeric mapping vector The calculation method is as follows:
[0083] ;
[0084] in, It is a chimeric mapping vector; It is the normalized instantaneous voltage value of the current electrical equipment; It is the normalized instantaneous current value of the current electrical equipment; for the recursive reconstruction vector The purpose of raising the value to the power of 1.5 is to enhance the distinguishability of mid-to-high amplitude features while maintaining the stability of low amplitude features, thus forming a non-linear amplification mechanism. Based on the joint modulation terms of electrical parameters and dynamic carbon factors, a nonlinear coupling control mechanism is formed between energy consumption parameters and carbon factors, emphasizing the carbon emission response regulation effect under different operating conditions (such as low voltage and high current vs. high voltage and low current).
[0085] S2-2. Mapping vectors of each electrical device Providing complete input for the final carbon emission estimate, the estimation process does not employ traditional black-box neural networks. Instead, it uses an explicit calculation method based on an electrical carbon estimation formula to obtain the carbon emissions. This formula incorporates a weighting mechanism for electrical equipment.
[0086] Final estimation results The carbon emissions, i.e., the results of electricity carbon metering, are calculated as follows:
[0087] ;
[0088] in, It is an estimated value of carbon emissions, i.e., the result of carbon metering. It represents the total number of devices within the area; It is the first The embedding mapping vector of each device The norm; It is the first The weighting factors for each device are determined by expert experience based on the importance of the device's rated power, operating time, or device type, and the sum is 1. It is the baseline carbon emission factor (regional long-term average), obtained from existing carbon factor databases (such as publicly available data from national / regional power grids); This is the renewable energy reverse adjustment coefficient, which adjusts the carbon factor as the proportion of renewable energy decreases. It is determined based on expert experience, with a reference value of [value missing]. ; It is a minimal stable term, preventing division by zero.
[0089] It should be noted that Embodiments 1 and 2 are not independent of each other. In implementation, the corresponding modules in Embodiment 1 can be implemented by referring to the method steps in Embodiment 2, or the module division method in Embodiment 1 can be used to plan and design the program product for implementing Embodiment 2.
[0090] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0091] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0092] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics thereof. The scope of the invention is defined by the claims rather than the foregoing description.
Claims
1. A multi-mode data fusion based electrical carbon metering system, characterized by: The multi-modal data acquisition module, the data preprocessing and feature extraction module, the tensor construction module, the tensor transformation module, the recursive feature reconstruction module, the dynamic carbon emission factor construction module, the chimeric mapping processing module, and the electric carbon metering estimation module are included. The original multi-modal data obtained by the multi-modal data acquisition module is sequentially subjected to the data preprocessing and feature extraction module, the tensor construction module, the tensor transformation module, and the recursive feature reconstruction module to obtain a recursive reconstruction vector.
2. The multi-modal data fusion based electro-carbon metering system as claimed in claim 1, wherein: The multi-modal data acquisition module is configured to synchronously acquire original multi-modal data through a multi-modal sensor acquisition device deployed at an electric terminal, and send the original multi-modal data to the data preprocessing and feature extraction module. The data preprocessing and feature extraction module is configured to preprocess the original multi-modal data, then extract features from the preprocessed multi-modal data, and then perform dimension reduction processing, normalization and standardization processing to obtain a multi-modal data vector, and send the multi-modal data vector to the tensor construction module.
3. The multi-modal data fusion based electro-carbon metering system as claimed in claim 1 or 2, wherein: The tensor construction module is configured to use an outer product operation to construct a tensor from the multi-modal data vector output by the data preprocessing and feature extraction module, and send the constructed tensor to the tensor transformation module.
4. The multi-modal data fusion based electro-carbon metering system as claimed in claim 3, wherein: The tensor transformation module is configured to introduce a rotation tensor operation based on the constructed tensor to construct a tensor consistency mapping, to ensure mapping isomorphism in time sequence between modalities, to remove nonlinear coupling bias between modalities, and to obtain a basic modality rotation vector.
5. The multi-modal data fusion based electro-carbon metering system as claimed in claim 4, wherein: The recursive feature reconstruction module is configured to introduce a recursive feature reconstruction mechanism to reconstruct the basic modality rotation vector to obtain a recursive reconstruction vector, and send the recursive reconstruction vector to the chimeric mapping processing module.
6. The multi-modal data fusion based electro-carbon metering system as claimed in claim 5, wherein: The dynamic carbon emission factor construction module is configured to construct a dynamic carbon emission factor, and send the dynamic carbon emission factor to the chimeric mapping processing module. The chimeric mapping processing module is configured to perform recursive modality mapping based on the recursive reconstruction vector and the constructed dynamic carbon emission factor to form a final feature chimeric representation as a chimeric mapping vector of each electric device, and send the chimeric mapping vector to the electric carbon metering estimation module.
7. The multi-modal data fusion based electro-carbon metering system as claimed in claim 6, wherein: The electric carbon metering estimation module is configured to perform explicit calculation by introducing an electric carbon estimation formula to obtain an electric carbon metering estimation result, and to realize electric carbon metering based on multi-modal data fusion.
8. A method for electric carbon metering based on multi-modal data fusion, characterized in that The method comprises the following steps: S1. Obtain original multi-modal data, perform feature extraction and dimension reduction processing after preprocessing to obtain a multi-modal data vector, construct a tensor representation based on the multi-modal data vector, introduce a rotation angle vector, perform rotation transformation to obtain a basic modality rotation vector, and then introduce a recursive feature reconstruction mechanism to obtain a recursive reconstruction vector. S2. Perform recursive modality mapping based on the recursive reconstruction vector and the constructed dynamic carbon emission factor to form a final feature chimeric representation as a chimeric mapping vector of each electric device, then introduce an electric carbon estimation formula to perform explicit calculation to obtain an electric carbon metering result, and realize electric carbon metering based on multi-modal data fusion.
9. The multi-modal data fusion based electrical carbon metering method as claimed in claim 8, wherein, The step S1 specifically comprises: S1-1. Synchronize acquisition of original multi-modal data through multi-modal sensor acquisition devices deployed at power consumption terminals; S1-2. Preprocess the original multi-modal data; S1-3. Perform feature extraction on the preprocessed multi-modal data, and then perform dimension reduction processing to obtain a multi-modal data vector, which includes an electrical parameter modal vector, an environmental modal vector, and a device state modal vector; S1-4. Perform a tensor construction operation on the multi-modal data vector, and combine vectors of different modalities into a high-order tensor through an outer product operation, i.e., construct a tensor; S1-5. Introduce a rotation tensor operation to construct a tensor consistency mapping to ensure mapping isomorphism between modalities in time sequence, and remove nonlinear coupling bias between modalities. After rotation, a rotation tensor is obtained, which is expanded into a vector form to obtain a basic modal rotation vector; S1-6. Introduce a recursive feature reconstruction mechanism, and perform multi-layer recursive calculation based on the basic modal rotation vector. During the calculation process, the influence of environmental humidity and light intensity is considered to obtain a recursive reconstruction vector.
10. The multi-modal data fusion based electrical carbon metering method as claimed in claim 8 or 9, wherein, The step S2 specifically includes: S2-1. Based on the recursive reconstruction vector, combine the constructed dynamic carbon emission factor to perform recursive modal mapping, and form a final feature chimeric representation as a chimeric mapping vector of each power consumption device; The calculation process of the dynamic carbon emission factor introduces the renewable energy proportion of the region and the unit area power consumption intensity; The calculation process of the chimeric mapping vector introduces the instantaneous voltage value and the instantaneous current value of the power consumption device; S2-2. Take the chimeric mapping vector of each power consumption device as a complete input for final estimated carbon emission, and perform calculation through an introduced electric carbon estimation formula to obtain the carbon emission; The electric carbon estimation formula introduces a weighting mechanism for the power consumption device.