Multi-source data fusion new energy power electricity charge intelligent accounting and energy efficiency analysis system

Abstract

The invention discloses a multi-source data fused new energy power electricity charge intelligent accounting and energy efficiency analysis system, relates to the technical field of new energy power V2G networks, and aims to solve the problems that the existing system cannot effectively fuse cross-regional heterogeneous V2G network multi-source data, complement sparse data and guarantee privacy, so that the multi-agent electricity charge allocation precision is insufficient, and the efficiency of the system is low. The system comprises a multi-region dynamic data acquisition module for acquiring real-time multi-source data and historical data of a cross-region heterogeneous V2G network, and transmitting the real-time multi-source data and historical data to a federal transfer learning processing module; and the federal transfer learning processing module is used for performing feature extraction, cross-regional common feature transfer and heterogeneous data calibration on the multi-source data, and synchronously realizing data full-life-cycle privacy protection through the self-adaptive privacy protection module. The method has the advantages that the cross-regional V2G network multi-subject electricity charge accounting precision is improved, and fair settlement support is provided for commercial operation.

Description

Technical Field

[0001] This invention relates to the field of new energy power V2G network technology, and more specifically, to a new energy power intelligent accounting and energy efficiency analysis system that integrates multi-source data. Background Technology

[0002] With the large-scale popularization of new energy vehicles and the cross-regional collaborative application of V2G (vehicle-to-grid) technology, heterogeneous V2G networks centered on metropolitan areas and economic zones have become key carriers for the dispatch and consumption of new energy power resources. In these cross-regional scenarios, there are significant differences in the distribution of charging pile types (such as the proportion of AC slow charging and DC fast charging), electricity pricing policies (peak and valley time division, cross-regional transmission loss sharing standards), and user energy consumption behaviors (commuting period charging and discharging patterns, seasonal fluctuations in electricity demand) across different regions. This has led to a multi-source heterogeneous data system that includes real-time charging and discharging data of electric vehicles, operational status data of heterogeneous charging piles, regional dynamic electricity price data, and cross-regional transmission loss data. These data are scattered and stored in local nodes in various regions, and some newly commissioned areas and remote areas often experience sudden data sparsity issues (such as interruption of charging pile data collection and insufficient sample size of user behavior data). At the same time, the cross-regional transmission of sensitive data such as user identity information and real-time charging and discharging records also faces the risk of privacy leakage.

[0003] Current electricity billing systems in the industry lack effective multi-source data fusion and privacy protection mechanisms. They cannot accurately eliminate the distribution differences of cross-regional data, nor can they reasonably supplement sparse data. This results in insufficient accuracy of electricity bill allocation among regional operators, charging pile owners, and users, which directly affects the fairness and sustainability of commercial operation of cross-regional V2G networks. In view of this, we propose a new energy power intelligent electricity billing and energy efficiency analysis system based on multi-source data fusion. Summary of the Invention

[0004] The purpose of this invention is to provide a new energy power intelligent accounting and energy efficiency analysis system that integrates multi-source data to solve the technical problem that existing systems cannot effectively integrate multi-source data from cross-regional heterogeneous V2G networks, supplement sparse data, and protect privacy, resulting in insufficient accuracy of multi-entity electricity cost allocation and difficulty in supporting cross-regional V2G commercial operation.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a multi-source data fusion intelligent accounting and energy efficiency analysis system for new energy power charges, comprising:

[0006] The multi-regional dynamic data acquisition module acquires real-time multi-source data and historical data from cross-regional heterogeneous V2G networks and transmits them to the federated transfer learning processing module.

[0007] The federated transfer learning processing module performs feature extraction, cross-regional common feature transfer, and heterogeneous data calibration on multi-source data, while simultaneously achieving full lifecycle privacy protection of data through the adaptive privacy protection module.

[0008] The intelligent electricity billing module completes accurate electricity billing for multiple entities in cross-regional V2G networks based on the processed data;

[0009] The cross-regional energy efficiency analysis module combines accounting results with multi-source data to generate regional energy efficiency difference reports and optimization strategies.

[0010] The dynamic collaborative control module responds to sudden data sparsity or heterogeneous conflict events across regions and adjusts feature migration strategies and accounting parameters in real time.

[0011] Preferably, the multi-region dynamic data acquisition module includes a regional distributed sensing unit, a data status monitoring unit, and a dynamic preprocessing unit;

[0012] Regional distributed sensing units are deployed at V2G network nodes in various regions. The data they collect includes real-time charging and discharging data of electric vehicles, operating status data of heterogeneous charging piles, regional electricity price policy data, cross-regional power transmission network loss data, and user energy consumption behavior data.

[0013] The data status monitoring unit monitors the integrity and timeliness of data in each region in real time and identifies sudden data loss or delay events;

[0014] Based on the data status monitoring results, the dynamic preprocessing unit performs format standardization on the complete data, marks missing or delayed data as abnormal, and then transmits it to the federated transfer learning processing module.

[0015] Preferably, the federated transfer learning processing module includes a dynamic feature extraction submodule, a cross-regional feature transfer submodule, a heterogeneous data calibration submodule, and a feature completion submodule;

[0016] The dynamic feature extraction submodule adaptively adjusts the feature extraction cycle according to the data density and update frequency of each region, using high-frequency extraction for data with high real-time requirements and low-frequency extraction for historical statistical data.

[0017] The cross-regional feature transfer submodule selects transferable common features from regions with sufficient data based on the similarity of regional electricity consumption patterns, establishes a feature transfer priority sequence, and prioritizes the transfer of features that are strongly related to electricity billing.

[0018] The heterogeneous data calibration submodule adopts an adversarial learning architecture, which eliminates the distribution differences of data in different regions through dynamic adversarial training of the feature transformation network and the domain discrimination network.

[0019] The feature completion submodule targets bursty sparse regions of data and generates supplementary features based on common migration features and historical correlation data to improve data dimensions.

[0020] Preferably, the cross-regional feature transfer submodule includes a regional similarity calculation unit and a transfer strategy adjustment unit;

[0021] The regional similarity calculation unit quantifies the similarity of electricity consumption patterns in different regions by analyzing user travel patterns, the distribution of charging pile types, and electricity price fluctuation patterns.

[0022] The migration strategy adjustment unit dynamically adjusts the feature migration ratio based on the similarity results, increasing the feature migration weight for high similarity regions and only migrating basic common features for low similarity regions, thus avoiding invalid feature migration from affecting the calculation accuracy.

[0023] The regional similarity calculation unit uses the following formula to quantify the similarity of regional electricity consumption patterns:

[0024] ;

[0025] in, For the region With the region Similarity of electricity consumption patterns The weighting coefficients for the similarity of users' travel patterns. The weighting coefficients represent the similarity in the distribution of charging pile types. The weighting coefficients for the similarity of electricity price fluctuation patterns. For the region With the region The similarity of user travel patterns For the region With the region Similarity in the distribution of charging pile types For the region With the region Similarity of electricity price fluctuation patterns;

[0026] Migration strategy adjustment unit according to Adjust the feature transfer ratio using the following formula:

[0027] ;

[0028] in, For the region To the region Feature transfer weights, The highest migration weight, For the lowest migration weight, The similarity threshold;

[0029] when When the similarity is above the threshold, the highest transfer weight is used; when the similarity is below the threshold, the weight is adjusted linearly according to the similarity; when the similarity is 0, no feature transfer is performed.

[0030] Preferably, the feature completion submodule includes a missing pattern recognition unit and an associated feature reasoning unit;

[0031] The missing pattern recognition unit identifies missing patterns as random or systematic missing by analyzing the temporal and regional distribution of bursty sparse data.

[0032] For random missing features, the related feature inference unit interpolates and completes them based on historical related data from the same region. For systematic missing features, it calls highly similar migration features from other regions and combines them with local historical data to reconstruct the features, ensuring that the completed features are consistent with the actual electricity consumption trend.

[0033] The interpolation completion formula for randomly missing data in the association feature inference unit is as follows:

[0034] ;

[0035] in, for Data completion in real time Select the window length for historical data. for The weight of historical data at any given moment for Historical data related to a specific moment;

[0036] The formula for feature reconstruction of systematically missing data is:

[0037] ;

[0038] in, For the region exist Data reconstruction and completion in real time These are the weight coefficients for the transfer features. For the region To the region Feature transfer weights, For the region exist Transition feature data at time points For the region exist Historical average data for the same period at that time.

[0039] Preferably, the adaptive privacy protection module includes a hierarchical encryption unit, a dynamic key management unit, and a local training isolation unit;

[0040] The hierarchical encryption unit adopts differentiated encryption strategies based on the data sensitivity level, using high-strength encryption for user identity information and real-time charging and discharging data, and lightweight encryption for non-sensitive device operating status data;

[0041] The dynamic key management unit updates the encryption key regularly based on the frequency of cross-regional data interaction, and different regions use independent key pools;

[0042] The local training isolation unit limits the data of each region to complete feature extraction and preliminary model training only on the local node, and only uploads the feature parameters to the federated transfer learning processing module for global aggregation.

[0043] The hierarchical encryption unit uses differentiated encryption strength calculation, and the specific formula is as follows:

[0044] ;

[0045] in, Data encryption strength level, This represents the weighting coefficient for the data sensitivity level. Data sensitivity level, This is the data interaction frequency coefficient.

[0046] Preferably, the smart electricity billing module includes a dynamic adaptation unit for billing rules, a multi-subject apportionment coefficient calculation unit, and a billing result verification unit;

[0047] The dynamic adaptation unit for accounting rules automatically updates the electricity billing rules based on real-time electricity pricing policies in each region, real-time transmission loss data, and changes in V2G participation modes.

[0048] The multi-entity cost-sharing coefficient calculation unit calculates the electricity cost-sharing coefficients for regional operators, charging pile owners, and users based on the calibrated fusion characteristics and the participation, contribution, and loss-bearing ratio of each entity in the V2G network.

[0049] The verification unit compares the real-time verification results with historical data from the same period and data from the same type of region. If the deviation exceeds the preset range, the federated transfer learning processing module is triggered to recalibrate and verify the features.

[0050] The multi-entity sharing coefficient calculation unit uses the following formula to calculate the electricity cost sharing coefficient for each entity:

[0051] ;

[0052] in, For the first Electricity cost sharing coefficient for the main body of the category, As part of the participation weight, As a weight for contribution, Assign a proportional weight to losses. For the first V2G network participation of the main body, For the first The contribution of the subject to the V2G network, For the first The proportion of losses borne by the main body of the class;

[0053] The formula for calculating and judging the deviation of the verification unit of the accounting results is as follows:

[0054] ;

[0055] when At that time, it triggers re-feature calibration and recalculation;

[0056] in, To account for the relative deviation between the real-time calculation results and the reference results, For real-time calculation of electricity bills, For reference regarding electricity bill results, This is the preset deviation threshold.

[0057] Preferably, the cross-regional energy efficiency analysis module includes a dynamic energy efficiency index calculation unit, a regional difference tracing unit, and an optimization strategy generation unit;

[0058] The dynamic energy efficiency index calculation unit calculates energy efficiency indicators such as unit power loss rate, new energy consumption rate, charging pile load rate and user participation in each region based on real-time data, and dynamically corrects them as the data is updated.

[0059] The regional difference tracing unit compares energy efficiency indicators in different regions and combines data feature migration records and accounting parameters to identify the core factors that cause energy efficiency differences, including data quality, electricity pricing mechanisms, or equipment operating efficiency.

[0060] Based on the results of differential source tracing, the optimization strategy generation unit generates targeted cross-regional resource scheduling suggestions, charging pile operation mode adjustment plans, and user energy consumption incentive strategies, and quantifies the economic benefits of strategy implementation by linking them to electricity billing results.

[0061] The formula for calculating the renewable energy absorption rate of the dynamic energy efficiency index calculation unit is as follows:

[0062] ;

[0063] in, For the regional renewable energy consumption rate, The amount of renewable energy generated for the regional V2G network. The electricity obtained by the regional V2G network from the main power grid;

[0064] The formula for quantifying the economic benefits of the optimization strategy generation unit is:

[0065] ;

[0066] in, For the first time after the strategy is implemented The economic benefits of the subject For the first Electricity cost sharing coefficient for the main body of the category, The unit electricity cost before the strategy is implemented. The unit electricity cost after the strategy is implemented. This represents the total electricity consumption during the strategy implementation period.

[0067] Preferably, the dynamic collaborative control module includes an event response unit, a parameter adaptive adjustment unit, and a linkage execution unit;

[0068] The event response unit monitors sudden sparse events, heterogeneous conflict events, and accounting deviation events in cross-regional data transmission in real time and triggers corresponding processing mechanisms.

[0069] The parameter adaptive adjustment unit increases the feature transfer frequency and completion weights for bursty data sparsity events, optimizes the training parameters of the adversarial learning network for heterogeneous conflict events, and adjusts the calculation logic of the multi-subject apportionment coefficient for accounting bias events.

[0070] The linkage execution unit will adjust parameters and synchronize them to the federated transfer learning processing module, the smart electricity billing module, and the cross-regional energy efficiency analysis module to ensure that each module responds to event processing in a coordinated manner.

[0071] The formula for adjusting the feature completion weights in response to bursty data sparsity events is as follows:

[0072] ;

[0073] in, To complete the weights for the adjusted features, To complete the weights for the initial features, To adjust the coefficient, Data missing rate;

[0074] The formula for adjusting the weighting of the multi-entity allocation coefficients in response to accounting deviation events is as follows:

[0075] ;

[0076] ;

[0077] By reducing the weight of participation and increasing the weight of contribution, the impact of bias on the apportionment coefficient is corrected.

[0078] in, The adjusted participation weights, As the initial participation weight, The adjusted contribution weight. As the initial contribution weight, This represents the relative deviation of the calculation results.

[0079] A method for intelligent accounting and energy efficiency analysis of new energy power costs through multi-source data fusion includes the following steps:

[0080] S1. Acquire real-time multi-source data and historical data of cross-regional heterogeneous V2G networks through the multi-regional dynamic data acquisition module, monitor data status and perform dynamic preprocessing;

[0081] S2. The federated transfer learning processing module is used to dynamically extract features from the preprocessed data, transfer common features based on regional similarity, calibrate the differences in the distribution of heterogeneous data through adversarial learning, and complete the features in the bursty sparse regions of the data.

[0082] S3. Throughout the entire data processing process, adaptive privacy protection is achieved through hierarchical encryption, dynamic key management, and local training isolation.

[0083] S4. Based on the calibrated fusion features, combined with dynamically adapted accounting rules and multi-subject apportionment coefficients, complete cross-regional electricity fee accounting and verify the results;

[0084] S5. Calculate the dynamic energy efficiency index of each region, trace the regional differences, and generate optimization strategies;

[0085] S6 responds to emergencies through the dynamic collaborative control module, adjusts the parameters of each module in real time, ensures the continuous and stable operation of the system, synchronously stores data and analysis results, and enables cross-platform interaction.

[0086] Compared with the prior art, the beneficial effects of the present invention are:

[0087] 1. This invention utilizes the dynamic feature extraction, cross-regional common feature transfer, and heterogeneous data calibration functions of the federated transfer learning processing module, combined with the dynamic adaptation rules and multi-subject apportionment coefficient algorithm of the intelligent electricity billing module. This effectively integrates cross-regional multi-source heterogeneous data, accurately completes the sparse data dimensions, reduces the deviation in multi-subject electricity billing in cross-regional V2G networks, solves the problem of unfair cross-regional electricity billing, and provides fair and accurate settlement support for the commercial operation of cross-regional V2G.

[0088] 2. This invention also relies on the dynamic energy efficiency index calculation and regional difference tracing function of the cross-regional energy efficiency analysis module. On the basis of achieving accurate electricity cost accounting, it can further locate the core causes of energy efficiency differences in various regions, generate targeted cross-regional resource scheduling suggestions and charging pile operation optimization schemes, improve the regional new energy consumption rate and the balance of charging pile load rate, and solve the problem of insufficient cross-regional V2G network energy efficiency optimization after accurate accounting.

[0089] 3. This invention also achieves full lifecycle privacy protection for data by using a hierarchical encryption, dynamic key management, and local training isolation mechanism in the adaptive privacy protection module. This ensures multi-source data fusion and accurate accounting while preventing the leakage of sensitive user information. Furthermore, the differentiated encryption strategy can balance data security and interaction efficiency, solving the problem of balancing privacy protection and system operating efficiency during accurate accounting and energy efficiency optimization. Attached Figure Description

[0090] Figure 1 This is a schematic diagram of the system framework of the present invention;

[0091] Figure 2 This is a schematic diagram of the system flow of the present invention;

[0092] Figure 3 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0093] To facilitate understanding of the technical solution of the present invention by those skilled in the art, the technical solution of the present invention will now be further described in conjunction with the accompanying drawings.

[0094] Example 1, such as Figure 1 and Figure 2 As shown, this invention provides a multi-source data fusion-based intelligent accounting and energy efficiency analysis system for new energy power charges, comprising:

[0095] The multi-regional dynamic data acquisition module acquires real-time multi-source data and historical data from cross-regional heterogeneous V2G networks and transmits them to the federated transfer learning processing module.

[0096] The federated transfer learning processing module performs feature extraction, cross-regional common feature transfer, and heterogeneous data calibration on multi-source data, while simultaneously achieving full lifecycle privacy protection of data through the adaptive privacy protection module.

[0097] The intelligent electricity billing module completes accurate electricity billing for multiple entities in cross-regional V2G networks based on the processed data;

[0098] The cross-regional energy efficiency analysis module combines accounting results with multi-source data to generate regional energy efficiency difference reports and optimization strategies.

[0099] The dynamic collaborative control module responds to sudden data sparsity or heterogeneous conflict events across regions, and adjusts feature migration strategies and accounting parameters in real time. The modules work together to solve the dynamic collaborative problems of data privacy protection, accounting accuracy, and energy efficiency analysis effectiveness in cross-regional heterogeneous V2G networks.

[0100] In an embodiment of the present invention, the multi-region dynamic data acquisition module includes a regional distributed sensing unit, a data status monitoring unit, and a dynamic preprocessing unit.

[0101] Regional distributed sensing units are deployed at V2G network nodes in various regions. The data they collect includes real-time charging and discharging data of electric vehicles, operating status data of heterogeneous charging piles, regional electricity price policy data, cross-regional power transmission network loss data, and user energy consumption behavior data.

[0102] The data status monitoring unit monitors the integrity and timeliness of data in each region in real time and identifies sudden data loss or delay events;

[0103] Based on the data status monitoring results, the dynamic preprocessing unit performs format standardization on the complete data, marks missing or delayed data as abnormal, and then transmits it to the federated transfer learning processing module.

[0104] In an embodiment of the present invention, the federated transfer learning processing module includes a dynamic feature extraction submodule, a cross-regional feature transfer submodule, a heterogeneous data calibration submodule, and a feature completion submodule.

[0105] The dynamic feature extraction submodule adaptively adjusts the feature extraction cycle according to the data density and update frequency of each region, using high-frequency extraction for data with high real-time requirements and low-frequency extraction for historical statistical data.

[0106] The cross-regional feature transfer submodule selects transferable common features from regions with sufficient data based on the similarity of regional electricity consumption patterns, establishes a feature transfer priority sequence, and prioritizes the transfer of features that are strongly related to electricity billing.

[0107] The heterogeneous data calibration submodule adopts an adversarial learning architecture, which eliminates the distribution differences of data in different regions through dynamic adversarial training of the feature transformation network and the domain discrimination network.

[0108] The feature completion submodule targets bursty sparse regions of data and generates supplementary features based on common migration features and historical correlation data to improve data dimensions.

[0109] In an embodiment of the present invention, the cross-regional feature transfer submodule includes a region similarity calculation unit and a transfer strategy adjustment unit;

[0110] The regional similarity calculation unit quantifies the similarity of electricity consumption patterns in different regions by analyzing user travel patterns, the distribution of charging pile types, and electricity price fluctuation patterns.

[0111] The migration strategy adjustment unit dynamically adjusts the feature migration ratio based on the similarity results, increasing the feature migration weight for high similarity regions and only migrating basic common features for low similarity regions, thus avoiding invalid feature migration from affecting the calculation accuracy.

[0112] The regional similarity calculation unit uses the following formula to quantify the similarity of regional electricity consumption patterns:

[0113] ;

[0114] in, For the region With the region The similarity of electricity consumption patterns is defined in the range of [0,1]. The closer the value is to 1, the more similar the electricity consumption patterns of the two regions are; the closer the value is to 0, the greater the difference is.

[0115] This is a weighting coefficient for the similarity of user travel patterns, used to adjust the degree of influence of this dimension in the overall similarity.

[0116] This is a weighting coefficient for the similarity of charging pile type distribution, which serves to balance the contribution of this dimension to the assessment of electricity consumption patterns.

[0117] The weighting coefficients for the similarity of electricity price fluctuation patterns, and , Together they constitute a weighting system and satisfy... Ensure that the weight allocation is reasonable;

[0118] For the region With the region The similarity of user travel patterns is calculated using the cosine similarity algorithm, reflecting the consistency of electricity consumption behaviors related to travel, such as charging time and frequency, between users in the two regions.

[0119] For the region With the region The similarity of the charging pile type distribution is obtained by using the cosine similarity algorithm, which reflects the matching degree of the distribution of different types of equipment such as AC charging piles and DC charging piles in the two areas.

[0120] For the region With the region The similarity of electricity price fluctuation patterns is obtained through the cosine similarity algorithm, which represents the similarity of price mechanisms such as peak-valley electricity price period division and electricity price adjustment range between the two regions.

[0121] This formula uses a weighted summation method to quantify the correlation between electricity consumption patterns in two regions by integrating the similarity of three core dimensions: user travel patterns, charging pile type distribution, and electricity price fluctuation patterns. First, the cosine similarity of the three dimensions is calculated separately. Then, weights are assigned according to the importance of each dimension to electricity consumption patterns, ultimately yielding the overall similarity score. This ensures that the similarity assessment is comprehensive and closely reflects the actual operating characteristics of V2G networks.

[0122] Migration strategy adjustment unit according to Adjust the feature transfer ratio using the following formula:

[0123] ;

[0124] in, For the region To the region Feature transfer weights, The highest migration weight, For the lowest migration weight, The similarity threshold;

[0125] when When the similarity is above the threshold, the highest transfer weight is used; when the similarity is below the threshold, the weight is adjusted linearly according to the similarity; when the similarity is 0, no feature transfer is performed.

[0126] Based on the above calculations, the similarity of regional electricity consumption patterns The feature transfer weights are dynamically adjusted using a piecewise function. When the similarity is higher than a preset threshold, it indicates that the electricity consumption patterns of the two regions are highly compatible, and the highest transfer weight is used to fully utilize the features of the region with sufficient data. When the similarity is between the threshold and 0, the weights are adjusted linearly according to the similarity to avoid invalid feature transfer. When the similarity is 0, feature transfer is stopped to prevent irrelevant features from interfering with the calculation accuracy and to achieve accurate adaptation of the transfer strategy.

[0127] Through the synergistic effect of the two algorithms mentioned above, regions with similar electricity consumption patterns in cross-regional V2G networks can be accurately identified, and feature migration weights can be reasonably allocated. This ensures that sparse data regions can make full use of the effective features of data-rich regions to supplement their own data dimensions, while avoiding invalid feature migration caused by differences in regional electricity consumption patterns. This provides high-quality fused features for subsequent electricity billing, while reducing unnecessary data interaction and improving the system's data processing efficiency.

[0128] In an embodiment of the present invention, the feature completion submodule includes a missing pattern recognition unit and an associated feature reasoning unit;

[0129] The missing pattern recognition unit identifies missing patterns as random or systematic missing by analyzing the temporal and regional distribution of bursty sparse data.

[0130] For random missing features, the related feature inference unit interpolates and completes them based on historical related data from the same region. For systematic missing features, it calls highly similar migration features from other regions and combines them with local historical data to reconstruct the features, ensuring that the completed features are consistent with the actual electricity consumption trend.

[0131] The interpolation completion formula for randomly missing data in the association feature inference unit is as follows:

[0132] ;

[0133] in, for The completed data for each time step is the missing time step data obtained through interpolation calculation;

[0134] The window length for historical data determines the number of historical data points to be included in the interpolation calculation. This should be set reasonably based on the data update frequency and electricity consumption patterns.

[0135] for The weight of historical data at any given time, the magnitude of the weight is related to the sum of the historical data and the weight of the historical data. The correlation between the time-series data is positive, and satisfies the following conditions: This ensures the reasonableness of weight allocation;

[0136] for Historical data related to a given moment, i.e., the number of moments before the missing moment. Historical data at any given moment serves as the foundation for interpolation and completion.

[0137] For scenarios with randomly missing data, a weighted interpolation method is used to complete the missing data. First, data before the time of the missing data is selected. The system uses historical data from the same period to correlate with each other, and then assigns weights based on the correlation between each historical data and the missing data. The missing data is then calculated by weighted summation to ensure that the completed data matches the historical electricity consumption trend of the region and avoids the impact of random missing data on data integrity.

[0138] The formula for feature reconstruction of systematically missing data is:

[0139] ;

[0140] in, For the region exist Real-time reconstruction and data completion, i.e., completion results in scenarios with systemic missing data;

[0141] The weight coefficient for the transfer feature, with a value range of [0,1], is used to adjust the degree of influence of the transfer feature in the reconstruction and completion process;

[0142] For the region To the region Feature transfer weights, reflecting regional characteristics Features for regions The contribution of data completion;

[0143] For the region exist The migration feature data at each time point, i.e., used to supplement the region. Missing cross-regional feature data;

[0144] For the region exist Historical average data for the same period at any given time reflects the region The historical average electricity consumption at the same point in time provides a local benchmark for data reconstruction;

[0145] For systematically missing data, reconstruction is performed by combining cross-regional feature migration with local historical data. First, highly similar migration features from other regions are introduced, and then combined with the historical average data of the same period at the time of the missing data in that region. The reconstructed and completed data is generated by weighted summation. This approach not only utilizes effective cross-regional features to compensate for the lack of local data, but also relies on local historical data to ensure that the completed data conforms to the region's own electricity consumption characteristics.

[0146] The two data completion algorithms described above are designed with appropriate completion logic for different missing patterns, effectively solving the problem of bursty data sparsity in cross-regional V2G networks. Random missing interpolation completion ensures short-term data continuity, while systematic missing feature reconstruction combines cross-regional resources with local historical data to ensure the accuracy and rationality of the completed data. This provides complete and reliable data support for subsequent privacy protection, electricity billing, and other processes, avoiding accounting deviations or energy efficiency analysis distortions caused by data missing information.

[0147] In embodiments of the present invention, the adaptive privacy protection module includes a hierarchical encryption unit, a dynamic key management unit, and a local training isolation unit;

[0148] The hierarchical encryption unit adopts differentiated encryption strategies based on the data sensitivity level, using high-strength encryption for user identity information and real-time charging and discharging data, and lightweight encryption for non-sensitive device operating status data;

[0149] The dynamic key management unit updates the encryption key regularly based on the frequency of cross-regional data interaction, and different regions use independent key pools to avoid the risk of a single key leak causing data risks across the entire region.

[0150] The local training isolation unit limits the data of each region to complete feature extraction and preliminary model training only on the local node, and only uploads the feature parameters to the federated transfer learning processing module for global aggregation. The original data is not transmitted across regions.

[0151] The hierarchical encryption unit uses differentiated encryption strength calculation, and the specific formula is as follows:

[0152] ;

[0153] in, This represents the data encryption strength level, with a value range of [1, 5]. A higher level indicates more stringent encryption protection.

[0154] This is the data sensitivity level weighting coefficient, with a value range of [0,1], used to adjust the impact of data sensitivity on encryption strength;

[0155] The data sensitivity level is set according to the data type, and the value range is [1,5]. For example, user identity information and real-time charging and discharging data have a high sensitivity level, while equipment operating status data has a low sensitivity level.

[0156] This is the data interaction frequency coefficient, calculated based on the number of cross-regional data interactions. Its value ranges from [1,5]. The higher the interaction frequency, the larger the coefficient, which represents the higher the security risks faced during cross-regional data transmission.

[0157] This formula determines the encryption strength level of data by weightedly combining data sensitivity level and data interaction frequency. First, a sensitivity level is set based on the data type. Then, an interaction frequency coefficient is calculated based on the number of cross-regional data interactions. Finally, the encryption strength level is obtained by summing the two according to their weighted proportions. This achieves differentiated protection: "high encryption for highly sensitive, highly interactive data, and lightweight encryption for low-sensitivity, low-interaction data," ensuring both privacy and system efficiency.

[0158] By employing differentiated encryption strength algorithms, precise encryption strategies can be formulated to address the security needs and transmission characteristics of different types of data in cross-regional V2G networks. This provides strong protection for user privacy data and frequently interacting data, effectively mitigating the risk of data leakage, while using lightweight encryption for non-sensitive data to reduce the system resource consumption of encryption operations. This balances privacy protection and system efficiency, ensuring both security and smoothness during cross-regional data interaction.

[0159] In an embodiment of the present invention, the smart electricity billing module includes a dynamic adaptation unit for billing rules, a multi-subject apportionment coefficient calculation unit, and a billing result verification unit;

[0160] The dynamic adaptation unit for accounting rules automatically updates the electricity billing rules based on real-time electricity pricing policies in each region, real-time transmission loss data, and changes in V2G participation modes.

[0161] The multi-entity cost-sharing coefficient calculation unit calculates the electricity cost-sharing coefficients for regional operators, charging pile owners, and users based on the calibrated fusion characteristics and the participation, contribution, and loss-bearing ratio of each entity in the V2G network.

[0162] The verification unit compares the real-time verification results with historical data from the same period and data from the same type of region. If the deviation exceeds the preset range, the federated transfer learning processing module is triggered to recalibrate and verify the features.

[0163] The multi-entity sharing coefficient calculation unit uses the following formula to calculate the electricity cost sharing coefficient for each entity:

[0164] ;

[0165] in, For the first Electricity cost sharing coefficient for the main body of the category, Representing regional operators, Representing the owner of the charging station, Representing the user and meeting the requirements This determines the proportion of electricity costs borne by each entity;

[0166] The participation weight is used to adjust the impact of the degree of participation of the subject in V2G network operation on the sharing coefficient;

[0167] As a contribution weight, the contribution of the balancing subject to the V2G network (such as charging piles and electricity regulation) to the apportionment coefficient.

[0168] Assigning a proportional weight to losses, and , Together they constitute a weighting system and satisfy... Ensure that the weight allocation is scientific;

[0169] For the first The V2G network participation of the subject is calculated based on the calibrated fusion characteristics, reflecting the depth of the subject's participation in network operation;

[0170] For the first The V2G network contribution of the subject is obtained by the fusion feature analysis, which reflects the resource value provided by the subject to the network;

[0171] For the first The loss-bearing ratio of a subject is calculated based on the transmission loss data in the fusion characteristics, representing the share of network operation losses borne by that subject.

[0172] This formula is based on the participation, contribution, and loss-bearing ratio of multiple entities in the V2G network. Through weighted summation and normalization, it calculates the electricity cost sharing coefficient for each entity. First, participation, contribution, and loss-bearing ratio are weighted separately to obtain a comprehensive score for each entity. Then, the comprehensive score of each entity is divided by the sum of the comprehensive scores of all entities to normalize the sharing coefficient, ensuring that the sum of the sharing coefficients is 1 and objectively reflecting the actual role and responsibility of each entity in the V2G network.

[0173] The formula for calculating and judging the deviation of the verification unit of the accounting results is as follows:

[0174] ;

[0175] when At that time, it triggers re-feature calibration and recalculation;

[0176] in, The relative deviation between the real-time calculation results and the reference results reflects the degree of deviation between the calculation results and the benchmark value;

[0177] This refers to the real-time calculated electricity bill result, that is, the electricity bill data calculated by the system for the current time period;

[0178] To reference the electricity bill results, the average of the historical calculation results for the same period and the calculation results for similar areas is taken as the benchmark for judging whether the real-time calculation results are reasonable.

[0179] The preset deviation threshold is the critical value for determining whether the accounting results need to be recalibrated. Exceeding this threshold indicates that there may be errors in the accounting results.

[0180] The accuracy of the calculation results is determined by calculating the relative deviation between the real-time calculation results and the reference results. First, the average value of the historical calculation results for the same period and the calculation results for the same type of area is selected as the reference result. Then, the absolute difference between the real-time calculation result and the reference result is calculated. Finally, the relative deviation is obtained by dividing by the reference result. If the deviation exceeds a preset threshold, re-feature calibration and calculation are triggered to ensure the reliability of the electricity bill calculation results.

[0181] The multi-entity cost-sharing coefficient algorithm can objectively and fairly allocate the electricity cost-sharing ratio among various entities in a cross-regional V2G network, avoiding the imbalance of interests caused by a "one-size-fits-all" approach, and ensuring the participation of multiple entities such as operators, charging pile owners, and users. The calculation result deviation judgment algorithm, by comparing with historical data and data from similar regions, can promptly identify calculation errors and trigger correction mechanisms to ensure the accuracy of electricity cost calculation results, providing a fair and reliable basis for cost settlement for the commercial operation of cross-regional V2G networks.

[0182] In an embodiment of the present invention, the cross-regional energy efficiency analysis module includes a dynamic energy efficiency index calculation unit, a regional difference tracing unit, and an optimization strategy generation unit.

[0183] The dynamic energy efficiency index calculation unit calculates energy efficiency indicators such as unit power loss rate, new energy consumption rate, charging pile load rate and user participation in each region based on real-time data, and dynamically corrects them as the data is updated.

[0184] The regional difference tracing unit compares energy efficiency indicators in different regions and combines data feature migration records and accounting parameters to identify the core factors that cause energy efficiency differences, including data quality, electricity pricing mechanisms, or equipment operating efficiency.

[0185] Based on the results of differential source tracing, the optimization strategy generation unit generates targeted cross-regional resource scheduling suggestions, charging pile operation mode adjustment plans, and user energy consumption incentive strategies, and quantifies the economic benefits of strategy implementation by linking them to electricity billing results.

[0186] The formula for calculating the renewable energy absorption rate of the dynamic energy efficiency index calculation unit is as follows:

[0187] ;

[0188] in, The regional renewable energy absorption rate reflects the extent to which the regional V2G network utilizes renewable energy for power generation; the higher the value, the better the renewable energy absorption effect.

[0189] The amount of renewable energy power consumed by the regional V2G network, including the consumption of renewable energy power such as photovoltaic and wind power in the network, is obtained by the multi-regional dynamic data acquisition module.

[0190] The amount of electricity that the regional V2G network obtains from the main power grid, i.e. the amount of traditional electricity supplied in the network, is also obtained through statistics from the data acquisition module.

[0191] This formula quantifies a region's capacity to absorb renewable energy by calculating the proportion of renewable energy generation consumed by the regional V2G network to the total electricity consumption (the sum of renewable energy generation and the main grid supply). The calculation, which directly divides renewable energy generation by total electricity consumption, is simple and intuitive, quickly reflecting the proportion of renewable energy in V2G network electricity consumption and providing core indicators for energy efficiency analysis.

[0192] The formula for quantifying the economic benefits of the optimization strategy generation unit is:

[0193] ;

[0194] in, For the first time after the strategy is implemented The economic benefits of the subject group, namely the amount of electricity saved by each subject due to the implementation of the strategy;

[0195] For the first The electricity cost sharing coefficient for each entity determines the proportion of electricity cost savings that each entity shares.

[0196] The unit electricity cost before the strategy is implemented, i.e., the unit electricity cost before the optimization strategy is executed;

[0197] The unit electricity cost after the strategy is implemented is typically lower than the unit electricity cost after the strategy is executed. ;

[0198] This refers to the total electricity consumption during the policy implementation period, i.e., the total electricity consumption of the regional V2G network during the policy execution period.

[0199] By combining the electricity cost sharing coefficients of multiple entities, the difference in electricity cost per unit of electricity before and after the implementation of the strategy, and the total electricity consumption, the economic benefits of each entity are calculated. First, the reduction in electricity cost per unit of electricity is calculated, then multiplied by the total electricity consumption to obtain the total electricity cost savings. Finally, the savings are allocated according to the sharing coefficients of each entity, quantifying the economic benefits brought to each entity by the implementation of the strategy, and providing data support for the promotion of the optimization strategy.

[0200] The renewable energy absorption rate algorithm provides key quantitative indicators for energy efficiency analysis of cross-regional V2G networks, which can clearly reflect the region's utilization level of renewable energy and help identify areas with potential for energy efficiency improvement. The strategy implementation economic benefit algorithm, on the other hand, quantifies the economic benefits of each entity, intuitively reflecting the actual value of optimization strategies. This not only provides persuasiveness for the implementation of strategies such as grid dispatching and charging pile operation adjustment, but also further incentivizes multiple entities to participate in energy efficiency optimization, promoting the development of cross-regional V2G networks towards high efficiency and economy.

[0201] In embodiments of the present invention, the dynamic collaborative control module includes an event response unit, a parameter adaptive adjustment unit, and a linkage execution unit;

[0202] The event response unit monitors sudden sparse events, heterogeneous conflict events, and accounting deviation events in cross-regional data transmission in real time and triggers corresponding processing mechanisms.

[0203] The parameter adaptive adjustment unit increases the feature transfer frequency and completion weights for bursty data sparsity events, optimizes the training parameters of the adversarial learning network for heterogeneous conflict events, and adjusts the calculation logic of the multi-subject apportionment coefficient for accounting bias events.

[0204] The linkage execution unit will adjust parameters and synchronize them to the federated transfer learning processing module, the smart electricity billing module, and the cross-regional energy efficiency analysis module to ensure that each module responds to event processing in a coordinated manner.

[0205] The formula for adjusting the feature completion weights in response to bursty data sparsity events is as follows:

[0206] ;

[0207] in, The adjusted feature completion weights are the completion weights used after a burst of data sparsity events.

[0208] The initial feature is used to complete the weights, and the complete weights are used as the baseline under normal data conditions;

[0209] The adjustment coefficient is used to adjust the degree of influence of the data missing rate on the completion weight;

[0210] The data missing rate is calculated as the ratio of the amount of missing data to the total amount of data, as recorded by the data status monitoring unit. It reflects the severity of the data missing. The larger, The higher the value, the greater the weight of feature completion.

[0211] The feature completion weights are dynamically adjusted based on the data missing rate. The adjusted weights are calculated using the initial completion weights as a base, combined with the data missing rate and adjustment coefficients. A higher data missing rate results in a larger adjusted weight, meaning that the feature completion process relies more heavily on transfer features or historical data. This ensures that when data suddenly becomes sparse, the reliability of the completed data can be improved by increasing the completion weights, guaranteeing the normal operation of subsequent steps.

[0212] The formula for adjusting the weighting of the multi-entity allocation coefficients in response to accounting deviation events is as follows:

[0213] ;

[0214] ;

[0215] By reducing the weight of participation and increasing the weight of contribution, the impact of bias on the apportionment coefficient is corrected.

[0216] in, The adjusted participation weight is the participation weight used to calculate the allocation coefficient after the accounting deviation event occurs.

[0217] The initial participation weight is the participation weight benchmark under normal accounting conditions;

[0218] The adjusted contribution weight is the contribution weight used after the occurrence of the accounting deviation event;

[0219] The initial contribution weight is the baseline contribution weight under normal accounting conditions;

[0220] The relative deviation of the accounting results reflects the severity of the accounting deviation;

[0221] To address accounting deviation events, the impact of deviations is corrected by adjusting the participation and contribution weights in the multi-entity cost allocation coefficients. When accounting deviations occur, the participation weight is reduced and the contribution weight is increased to reduce the interference of participation assessment deviations on the cost allocation coefficients. At the same time, the role of contribution in the cost allocation coefficients is strengthened, making the cost allocation coefficients more closely reflect the actual contributions of each entity, thereby indirectly correcting the impact of accounting deviations on cost allocation.

[0222] The two dynamic adjustment algorithms described above can quickly respond to sudden events in cross-regional V2G networks. When data becomes suddenly sparse, the feature completion weights are adjusted to ensure data completion quality and prevent data problems from propagating to subsequent stages. When there are accounting deviations, the allocation coefficient weights are adjusted to reduce the impact of deviations on cost allocation and ensure fairness among multiple stakeholders. Through real-time parameter adjustments, the system's adaptability to sudden situations is enhanced, ensuring the continuous stability of electricity cost accounting and energy efficiency analysis in cross-regional V2G networks, and improving the overall reliability and robustness of the system.

[0223] Example 2, as follows Figure 3 As shown, a method for intelligent accounting and energy efficiency analysis of new energy power charges based on multi-source data fusion includes the following steps:

[0224] S1. Acquire real-time multi-source data and historical data of cross-regional heterogeneous V2G networks through the multi-regional dynamic data acquisition module, monitor data status and perform dynamic preprocessing;

[0225] S2. The federated transfer learning processing module is used to dynamically extract features from the preprocessed data, transfer common features based on regional similarity, calibrate the differences in the distribution of heterogeneous data through adversarial learning, and complete the features in the bursty sparse regions of the data.

[0226] S3. Throughout the entire data processing process, adaptive privacy protection is achieved through hierarchical encryption, dynamic key management, and local training isolation.

[0227] S4. Based on the calibrated fusion features, combined with dynamically adapted accounting rules and multi-subject apportionment coefficients, complete cross-regional electricity fee accounting and verify the results;

[0228] S5. Calculate the dynamic energy efficiency index of each region, trace the regional differences, and generate optimization strategies;

[0229] S6 responds to emergencies through the dynamic collaborative control module, adjusts the parameters of each module in real time, ensures the continuous and stable operation of the system, synchronously stores data and analysis results, and enables cross-platform interaction.

[0230] The embodiments disclosed in this invention are preferred embodiments, but are not limited thereto. Those skilled in the art can easily understand the spirit of this invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of this invention, they are all within the protection scope of this invention.

Claims

1. A new energy power electricity fee intelligent accounting and energy efficiency analysis system for multi-source data fusion, characterized in that, The method comprises the following steps: a multi-region dynamic data acquisition module acquires real-time multi-source data and historical data of a cross-region heterogeneous V2G network and transmits the data to a federal transfer learning processing module; the federal transfer learning processing module performs feature extraction, cross-region common feature transfer, and heterogeneous data calibration on the multi-source data, and synchronously realizes data lifecycle privacy protection through an adaptive privacy protection module; an intelligent electricity fee accounting module completes cross-region V2G network multi-subject electricity fee accurate accounting based on the processed data; a cross-region energy efficiency analysis module generates a region energy efficiency difference report and an optimization strategy in combination with the accounting result and the multi-source data; a dynamic collaborative regulation module responds to cross-region data burst sparsity or heterogeneous conflict events, and adjusts the feature transfer strategy and the accounting parameters in real time. 2.The new energy power electricity bill intelligent accounting and energy efficiency analysis system of multi-source data fusion of claim 1, characterized in that, The multi-region dynamic data acquisition module comprises a region distributed sensing unit, a data state monitoring unit, and a dynamic preprocessing unit; The region distributed sensing unit is deployed at each region V2G network node, and the acquired data comprises real-time electric vehicle charging and discharging data, heterogeneous charging pile operation state data, regional electricity price policy data, cross-region power transmission network loss data, and user energy consumption behavior data; The data state monitoring unit monitors the integrity and timeliness of the data of each region in real time, and identifies data burst missing or delay events; The dynamic preprocessing unit performs format standardization processing on the complete data according to the data state monitoring result, and transmits the missing or delayed data marked with an abnormal state to the federal transfer learning processing module. 3.The new energy power electricity bill intelligent accounting and energy efficiency analysis system of multi-source data fusion of claim 1, characterized in that, The federal transfer learning processing module comprises a dynamic feature extraction submodule, a cross-region feature transfer submodule, a heterogeneous data calibration submodule, and a feature completion submodule; The dynamic feature extraction submodule adaptively adjusts the feature extraction period according to the data density and update frequency of each region, and adopts high-frequency extraction for data with high real-time requirements and low-frequency extraction for historical statistical data; The cross-region feature transfer submodule filters transferable common features from data sufficient regions based on the region electricity consumption mode similarity, establishes a feature transfer priority sequence, and preferentially transfers features strongly related to electricity fee accounting; The heterogeneous data calibration submodule adopts an adversarial learning architecture, and eliminates the distribution difference of different region data through dynamic adversarial training of a feature conversion network and a domain discrimination network; The feature completion submodule generates supplementary features to perfect the data dimension based on the transferred common features and historical associated data for data burst sparse regions.

4. The new energy power electricity fee intelligent accounting and energy efficiency analysis system of multi-source data fusion according to claim 3, characterized in that, The cross-region feature transfer submodule comprises a region similarity calculation unit and a transfer strategy adjustment unit; The region similarity calculation unit quantifies the electricity consumption mode similarity of different regions by analyzing user travel regularity, charging pile type distribution, and electricity price fluctuation mode; The transfer strategy adjustment unit dynamically adjusts the feature transfer proportion according to the similarity result, increases the feature transfer weight for high-similarity regions, and only transfers basic common features for low-similarity regions, so as to avoid invalid feature transfer affecting the accounting accuracy; The region similarity calculation unit quantifies the region electricity consumption mode similarity by using the following formula: ； wherein, is the similarity of the use mode of the region is the similarity of the use mode of the region is the similarity of the use mode of the region is the weight coefficient of the user travel regularity similarity, is the weight coefficient of the charging pile type distribution similarity, is the weight coefficient of the electricity price fluctuation mode similarity, is the similarity of the user travel regularity of the region is the similarity of the user travel regularity of the region is the similarity of the user travel regularity of the region is the similarity of the charging pile type distribution of the region is the similarity of the charging pile type distribution of the region is the similarity of the charging pile type distribution of the region is the similarity of the electricity price fluctuation mode of the region is the similarity of the electricity price fluctuation mode of the region is the similarity of the electricity price fluctuation mode of the region The migration policy adjustment unit adjusts the migration proportion of the feature according to The adjustment feature migration proportion, the specific formula is: ； wherein, is a region to a region characteristic transfer weight, is a highest transfer weight, is a lowest transfer weight, is a similarity threshold; When The highest migration weight is used when the threshold is exceeded, the weight is linearly adjusted according to the similarity when the threshold is not exceeded, and no feature migration is performed when the similarity is 0. 5.The new energy power electricity bill intelligent accounting and energy efficiency analysis system of multi-source data fusion of claim 3, characterized in that, The feature completion submodule comprises a missing mode identification unit and an associated feature reasoning unit; The missing pattern recognition unit recognizes the missing pattern as random missing or systematic missing by analyzing the time distribution and regional distribution of data burst sparsity; The correlation feature reasoning unit performs interpolation completion based on historical correlation data in the same region for random missing, and calls other regions with high similarity to transfer features, and combines local historical data to reconstruct features for systematic missing, to ensure that the completed features are consistent with the actual electricity consumption trend; The interpolation completion formula of the correlation feature reasoning unit for random missing data is: ； wherein, is complementary data of the time point, is a window length for selecting historical data, is is a weight of historical data of the time point, is is historical correlation data of the time point; The feature reconstruction formula for systematic missing data is: ； wherein, is a region at a reconstruction completion data at a time point, is a weight coefficient of a migration feature, is a region to a region a feature migration weight of, is a region at a migration feature data at a time point, is a region at a history same period average data at a time point. 6.The new energy power electricity bill intelligent accounting and energy efficiency analysis system of multi-source data fusion of claim 1, characterized in that, The adaptive privacy protection module includes a hierarchical encryption unit, a dynamic key management unit, and a local training isolation unit; The hierarchical encryption unit adopts a differential encryption strategy according to the data sensitivity level, and uses high-strength encryption for user identity information and real-time charging and discharging data, and uses lightweight encryption for non-sensitive device operating state data; The dynamic key management unit regularly updates the encryption key according to the cross-regional data interaction frequency, and different regions use independent key pools; The local training isolation unit limits the data of each region to complete feature extraction and preliminary model training only in the local node, and only uploads the feature parameters to the federal transfer learning processing module for global aggregation; The hierarchical encryption unit adopts a differential encryption strength calculation, and the specific formula is: ； wherein, is a data encryption strength level, is a data sensitivity level weight coefficient, is a data sensitivity level, is a data interaction frequency coefficient. 7.The new energy power electricity bill intelligent accounting and energy efficiency analysis system of multi-source data fusion of claim 1, characterized in that, The intelligent electricity cost accounting module includes an accounting rule dynamic adaptation unit, a multi-subject apportionment coefficient calculation unit, and an accounting result verification unit; The accounting rule dynamic adaptation unit automatically updates the electricity cost accounting rules according to the real-time electricity price policy of each region, real-time power transmission loss data, and V2G participation mode changes; The multi-subject apportionment coefficient calculation unit calculates the electricity cost apportionment coefficients of the regional operator, charging pile owner and user based on the calibrated fusion features, combined with the participation, contribution and loss bearing proportion of each subject in the V2G network; The accounting result verification unit compares the real-time accounting result with the historical same period data and the same type regional data, if the deviation exceeds the preset range, the federal transfer learning processing module is triggered to recalibrate the features and account again; The multi-subject apportionment coefficient calculation unit calculates the electricity cost apportionment coefficients of each subject using the following formula: ； wherein, is the electricity cost sharing coefficient of the first class subject, is the participation weight, is the contribution weight, is the loss bearing proportion weight, is the V2G network participation of the first class subject, is the V2G network contribution of the first class subject, is the loss bearing proportion of the first class subject. The deviation calculation and judgment formula of the accounting result verification unit is: ； When re-calibration of the features and accounting is triggered; wherein, is a relative deviation of the real-time accounting result from the reference result, is a real-time accounting electricity fee result, is a reference electricity fee result, is a preset deviation threshold. 8.The new energy power electricity bill intelligent accounting and energy efficiency analysis system of multi-source data fusion of claim 1, characterized in that, The cross-regional energy efficiency analysis module includes a dynamic energy efficiency index calculation unit, a regional difference tracing unit, and an optimization strategy generation unit; The dynamic energy efficiency index calculation unit calculates the energy efficiency indexes of unit power loss rate, new energy consumption rate, charging pile load rate and user participation rate based on real-time data, and dynamically corrects them with data updates; The regional difference tracing unit locates the core factors causing the energy efficiency difference by comparing the energy efficiency indexes of different regions, combined with the data feature transfer record and accounting parameters, including data quality, electricity price mechanism or equipment operation efficiency; The optimization strategy generation unit generates targeted cross-regional resource scheduling suggestions, charging pile operation mode adjustment schemes and user energy use incentive strategies according to the difference tracing results, and quantifies the economic benefits of implementing the strategies based on the electricity cost accounting results; The new energy consumption rate calculation formula of the dynamic energy efficiency index calculation unit is: ； wherein, is the regional new energy consumption rate, is the new energy generation capacity consumed by the regional V2G network, is the power obtained by the regional V2G network from the large power grid; The formula for quantifying the economic benefits of the optimization strategy generation unit is: ； wherein, the total electricity consumption after the strategy implementation, the economic benefits of the class subject, the total electricity consumption before the strategy implementation, the electricity cost allocation coefficient of the class subject, the electricity cost per unit of electricity before the strategy implementation, the electricity cost per unit of electricity after the strategy implementation, the total electricity consumption during the strategy implementation period. 9.The new energy power electricity bill intelligent accounting and energy efficiency analysis system of multi-source data fusion of claim 1, characterized in that, The dynamic coordination module comprises an event response unit, a parameter adaptive adjustment unit and a linkage execution unit. The event response unit monitors sudden sparse events, heterogeneous conflict events and accounting deviation events in cross-regional data transmission in real time, and triggers corresponding processing mechanisms. The parameter adaptive adjustment unit increases the feature migration frequency and the completion weight for data sudden sparse events, optimizes the training parameters of the adversarial learning network for heterogeneous conflict events, and adjusts the multi-agent allocation coefficient calculation logic for accounting deviation events. The linkage execution unit synchronizes the adjustment parameters to the federal transfer learning processing module, the intelligent electricity fee accounting module and the cross-regional energy efficiency analysis module to ensure that each module responds to event processing in coordination. For data sudden sparse events, the formula for adjusting the feature completion weight is: ； wherein, is the adjusted feature completion weight, is the initial feature completion weight, is the adjustment coefficient, is the data missing rate; For accounting deviation events, the formula for adjusting the multi-agent allocation coefficient weight is: ； ； By reducing the participation weight and increasing the contribution weight, the influence of deviation on the allocation coefficient is corrected. wherein, is the adjusted engagement weight, is the initial engagement weight, is the adjusted contribution weight, is the initial contribution weight, is the relative deviation of the accounting result.

10. A new energy power electricity fee intelligent accounting and energy efficiency analysis method of multi-source data fusion, comprising the new energy power electricity fee intelligent accounting and energy efficiency analysis system of multi-source data fusion according to any one of claims 1-9, characterized in that, The method comprises the following steps: S1, acquiring real-time multi-source data and historical data of a cross-regional heterogeneous V2G network through a multi-region dynamic data acquisition module, monitoring data status and performing dynamic preprocessing; S2, using a federal transfer learning processing module to perform dynamic feature extraction on the preprocessed data, migrating common features based on regional similarity, calibrating heterogeneous data distribution differences through adversarial learning, and performing feature completion on data sudden sparse areas; S3, during the whole data processing process, adaptive privacy protection is realized through hierarchical encryption, dynamic key management and local training isolation; S4, based on the calibrated fusion features, combined with dynamically adapted accounting rules and multi-agent allocation coefficients, cross-regional electricity fee accounting is completed and the results are verified; S5, calculating the dynamic energy efficiency indicators of each region, tracing the regional differences and generating optimization strategies; S6, responding to sudden events through the dynamic coordination module, adjusting the parameters of each module in real time to ensure the continuous and stable operation of the system, synchronously storing data and analysis results and realizing cross-platform interaction.

Images