Intelligent accounting platform and method for carbon emission of large-scale activities

Abstract

The invention discloses a large-scale activity carbon emission intelligent accounting platform and method, and the platform comprises an acquisition module which is used for collecting the carbon emission data of a large-scale activity; the preprocessing module is used for preprocessing the carbon emission data of the large-scale activities; the construction module is used for constructing an intelligent accounting model based on the LLM model and the dynamically updated carbon emission factor library; the accounting module is used for inputting the preprocessed carbon emission data of the large-scale activities into an intelligent accounting model for accounting and determining an accounting result; the large-scale activity carbon emission data can be comprehensively obtained, the carbon emission generated by energy use can be accurately calculated according to the carbon emission factor, and the accuracy of an accounting result is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon emission management, and particularly relates to a large-scale event carbon emission intelligent accounting platform and method. BACKGROUND

[0002] At present, the carbon emission accounting of large-scale events (such as concerts, sports events, exhibitions, conferences, etc.) generally faces problems such as low efficiency, difficult data collection, difficulty in ensuring data authenticity, and non-transparent accounting process.

[0003] The traditional accounting method mainly relies on manual collection of various types of paper tickets (such as tickets, meal coupons, transportation vouchers, etc.) and statistical data, and then professional personnel manually enter and calculate. This method has the following significant disadvantages:

[0004] Low efficiency: there are many types of tickets, and the amount is large and wide, which consumes a lot of time and effort for manual collection, entry and accounting. Especially in large-scale events, the amount of data grows exponentially, and the efficiency bottleneck is particularly prominent.

[0005] Difficult and distorted data collection: paper tickets are easy to lose, damage and may be counterfeit, resulting in incomplete and inaccurate data. For some non-ticketed emission sources (such as energy consumption, waste disposal), data acquisition is more difficult.

[0006] Non-transparent accounting process: the source of the accounting method and data is often not public, and the public cannot supervise, and the credibility of the data is questioned.

[0007] Single and static model: the existing carbon emission factor library and accounting model are mostly static updates, which are difficult to adapt to different types of activities, different regions, and changing emission standards and technological progress.

[0008] Difficult to trace responsibility: when there are data anomalies or emissions exceed the standard, it is difficult to quickly and accurately trace to the specific emission link and responsible subject.

[0009] Therefore, there is an urgent need for a large-scale event carbon emission intelligent accounting platform and method to solve the above problems. SUMMARY

[0010] The present application aims to at least one of the above technical problems. To this end, the first aspect of the present application aims to provide a large-scale event carbon emission intelligent accounting platform, which can comprehensively obtain large-scale event carbon emission data, accurately calculate the carbon emissions generated by energy use according to the carbon emission factor, and improve the accuracy of the accounting results.

[0011] The second aspect of the present application aims to provide a large-scale event carbon emission intelligent accounting method.

[0012] To achieve the above object, the first aspect of the present application provides a large-scale event carbon emission intelligent accounting platform, comprising:

[0013] A collection module is configured to collect carbon emission data of the large-scale event.

[0014] A preprocessing module is configured to preprocess the carbon emission data of the large-scale event.

[0015] A construction module is configured to construct an intelligent accounting model based on an LLM model and a dynamically updated carbon emission factor library.

[0016] An accounting module is configured to input the preprocessed carbon emission data of the large-scale event into the intelligent accounting model for accounting to determine an accounting result.

[0017] Preferably, the collection module comprises:

[0018] A first collection submodule is configured to obtain first data based on real-time data of Internet of Things device energy consumption and traffic travel.

[0019] A second collection submodule is configured to intelligently identify second data based on OCR+NLP technology on bill information.

[0020] A determination submodule is configured to determine the first data and the second data as the carbon emission data of the large-scale event.

[0021] Preferably, the construction method of the intelligent accounting model comprises:

[0022] The carbon emission factor in the carbon emission factor library is subjected to semantic understanding based on the LLM model, and the carbon emission factor text description information is converted into structured data.

[0023] The update information of the carbon emission factor library is dynamically obtained, and the update information of the carbon emission factor library is analyzed based on the LLM model to generate a correction instruction, and the factor mapping layer parameter of the LLM model is adjusted.

[0024] A bidirectional interactive collaborative training framework is constructed based on the LLM model and the dynamically updated carbon emission factor library.

[0025] A time attention mechanism is added to optimize the bidirectional interactive collaborative training framework, the self-attention layer of the bidirectional interactive collaborative training framework is fused with a carbon emission factor time decay model, and the weight coefficient of the fused model is back-propagated based on the update frequency of the dynamically updated carbon emission factor library.

[0026] In the bidirectional interactive co-training framework, a correlation matrix of carbon emission factors-emissions is constructed, with the historical version of the dynamic factor library as the horizontal axis and the activity emission results as the vertical axis. The nonlinear relationship between carbon emission factor changes and emission results is learned through the built-in basic model of the bidirectional interactive co-training framework.

[0027] The accounting logic chain of the bidirectional interactive co-training framework is defined, and the nonlinear parameters of the correlation matrix are embedded.

[0028] Based on the accounting logic chain, the bidirectional interactive co-training framework is trained end-to-end to obtain an initial intelligent accounting model.

[0029] Obtain the training data set and test data set of the intelligent accounting model.

[0030] Iteratively train the initial intelligent accounting model based on the training data set of the intelligent accounting model to obtain a target intelligent accounting model.

[0031] Test the target intelligent accounting model based on the test data set, and obtain the trained intelligent accounting model when the test result meets the requirements.

[0032] Preferably, it further comprises a traceability module for generating traceable data ledger based on blockchain technology to record key data and results in all collection, processing and accounting processes.

[0033] Preferably, it further comprises a visualization module for constructing Web / App end interfaces for different roles of activity organizers, participants and regulatory departments, and visualizing the carbon emission overview, detailed accounting report, emission trend analysis and carbon footprint of large-scale activities.

[0034] Preferably, the preprocessing module comprises:

[0035] A data cleaning submodule for cleaning the carbon emission data of the large-scale activities.

[0036] A data verification module for verifying the cleaned carbon emission data of the large-scale activities, and taking the carbon emission data of the large-scale activities that pass the verification as the preprocessed carbon emission data of the large-scale activities.

[0037] Preferably, the data cleaning submodule comprises:

[0038] A first calculation unit for:

[0039] Calculating the abnormality possibility of each dimension of carbon emission data in the multi-dimensional carbon emission data of the large-scale activities to obtain an abnormality possibility value of each dimension of carbon emission data; and determining a plurality of to-be-cleaned dimension data based on the abnormality possibility value.

[0040] The second calculation unit is configured to:

[0041] calculate a correlation index of any two to-be-cleaned dimension data, and determine a weight of each to-be-cleaned dimension data based on the correlation index;

[0042] align a plurality of to-be-cleaned dimension data based on a time sequence; wherein parameters of different to-be-cleaned dimensions corresponding to a same time are taken as a data point;

[0043] determine a comprehensive evaluation index of each data point based on different dimension parameter values of each data point and the weight of each to-be-cleaned dimension data;

[0044] The anomaly evaluation unit is configured to:

[0045] calculate a difference value of the comprehensive evaluation indexes of any two adjacent data points, to obtain a plurality of difference values;

[0046] calculate a mean value of the plurality of difference values, to obtain a difference threshold value;

[0047] compare the difference values with the preset difference threshold value respectively, and take a data point with a difference value greater than or equal to the preset difference threshold value as an anomaly data point, to obtain a plurality of anomaly data points;

[0048] The third calculation unit is configured to calculate an anomaly degree value of each parameter in each anomaly data point, and determine anomaly parameters of a plurality of different dimensions based on the anomaly degree value;

[0049] The data cleaning unit is configured to:

[0050] obtain a preset data cleaning rule;

[0051] clean the anomaly parameters of the plurality of different dimensions based on the preset data cleaning rule, to complete data cleaning of the carbon emission data of the large-scale activity.

[0052] Preferably, the first calculation unit comprises:

[0053] The first calculation sub-unit is configured to:

[0054] take carbon emission data of any dimension as target dimension data;

[0055] obtain an extreme value in the target dimension data; and segment the target dimension data based on the extreme value, to obtain target dimension sub-data;

[0056] calculate a ratio of a number of data points in each target dimension sub-data to a number of data points in the target dimension data, to obtain a first ratio;

[0057] calculate a ratio of a standard deviation to a mean value of data points in each target dimension sub-data, to obtain a second ratio;

[0058] multiplying the first ratio and the second ratio as a stage evaluation value of the target dimension sub-data;

[0059] adding all stage evaluation values of the target dimension sub-data as an anomaly possibility value;

[0060] The first comparison sub-unit is configured to compare the anomaly possibility value with a preset anomaly possibility threshold value, and take the carbon emission data of the corresponding dimension as the to-be-cleaned dimension data when the anomaly possibility value is greater than or equal to the preset anomaly possibility threshold value, so as to obtain a plurality of to-be-cleaned dimension data.

[0061] Preferably, the third calculation unit comprises:

[0062] The second calculation sub-unit is configured to:

[0063] arbitrarily obtain a parameter of one dimension in an anomaly data point as a to-be-evaluated parameter;

[0064] obtain a data mean value of the dimension where the to-be-evaluated parameter is located;

[0065] calculate a difference value between the parameter value of each dimension in the anomaly data point and the data mean value of the same dimension, so as to obtain a plurality of difference values;

[0066] calculate a mean value of the plurality of difference values, so as to obtain a difference mean value corresponding to the anomaly data point;

[0067] determine an anomaly degree value of the to-be-evaluated parameter based on a preset algorithm according to the to-be-evaluated parameter, the data mean value and the difference mean value;

[0068] traverse all parameter values of the anomaly data point, so as to obtain an anomaly degree value corresponding to each parameter;

[0069] The second comparison sub-unit is configured to compare the anomaly degree value with a preset anomaly degree threshold value, and take the parameter when the anomaly degree value is greater than or equal to the preset anomaly degree threshold value as an anomaly parameter, so as to obtain a plurality of anomaly parameters of different dimensions.

[0070] To achieve the above object, the second aspect embodiment of the present application provides a large-scale activity carbon emission intelligent accounting method, comprising:

[0071] collecting carbon emission data of a large-scale activity;

[0072] preprocessing the carbon emission data of the large-scale activity;

[0073] constructing an intelligent accounting model based on an LLM model and a dynamically updated carbon emission factor library;

[0074] inputting the carbon emission data of the large-scale activity after preprocessing into the intelligent accounting model for accounting, and determining an accounting result.

[0075] The present application provides a large-scale event carbon emission intelligent accounting platform and method, which integrates various technologies to build a large-scale event carbon emission intelligent accounting platform. Through the Internet of Things equipment and the ticket intelligent recognition technology, the automatic collection and accounting of carbon emission data are realized, the manual intervention is greatly reduced, and the accounting efficiency is significantly improved. The ticket intelligent recognition technology can accurately extract the ticket information, and the intelligent analysis of the intelligent accounting model can improve the accuracy of data collection. The introduction of the blockchain technology ensures the non-tamperability of the data, and guarantees the authenticity of the data from the source. The intelligent accounting model can understand and analyze complex activity scenarios, and combined with the dynamically updated carbon emission factor library, the multi-dimensional and refined carbon emission calculation is carried out, and the result is more accurate. All accounting processes and data records are recorded on the blockchain, which is open and transparent, convenient for supervision and tracing by all parties, and enhances the credibility of the accounting result. The detailed data flow information recorded on the blockchain can clearly trace the responsibility subject of each emission link, which is convenient for carbon emission management and accountability. The dynamic form and the updateable carbon emission factor library enable the platform to adapt to different types and different scales of large-scale events, and update the accounting standards in time.

[0076] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be achieved and obtained by the structure particularly pointed out in the written description and the accompanying drawings.

[0077] The technical solutions of the present application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0079] Figure 1 is a block diagram of a large-scale event carbon emission intelligent accounting platform according to an embodiment of the present application;

[0080] Figure 2 is a block diagram of a collection module according to an embodiment of the present application;

[0081] Figure 3 is a flowchart of a large-scale event carbon emission intelligent accounting method according to an embodiment of the present application. DETAILED DESCRIPTION

[0082] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not limit the present application.

[0083] Embodiment 1: As shown in the figure, a large-scale event carbon emission intelligent accounting platform comprises: Figure 1

[0084] A collection module is configured to collect carbon emission data of a large-scale event.

[0085] A preprocessing module is configured to preprocess the carbon emission data of the large-scale event.

[0086] A construction module is configured to construct an intelligent accounting model based on an LLM model and a dynamically updated carbon emission factor library.

[0087] An accounting module is configured to input the preprocessed carbon emission data of the large-scale event into the intelligent accounting model for accounting to determine an accounting result.

[0088] In this embodiment, real-time data of energy and transportation is collected by Internet of Things devices (such as smart meters and RFID), and data collection automation is achieved by automatically identifying bill information with OCR+NLP technology.

[0089] In this embodiment, the dynamic factor library: real-time update of carbon emission factors, adapt to different regions, activity types and policy changes.

[0090] The working principle and beneficial effects of the above technical solution are: a large-scale event carbon emission intelligent accounting platform is constructed by integrating multiple technologies, and automatic collection and accounting of carbon emission data is achieved through Internet of Things devices and bill intelligent identification technology, which greatly reduces manual intervention and significantly improves accounting efficiency; the bill intelligent identification technology can accurately extract bill information, and the intelligent analysis of the intelligent accounting model improves the accuracy of data collection; the introduction of the blockchain technology ensures the non-tamperability of the data, and guarantees the authenticity of the data from the source; the intelligent accounting model can understand and analyze complex activity scenarios, and combined with the dynamically updated carbon emission factor library, it can perform multi-dimensional and fine-grained carbon emission calculation, and the result is more accurate; all accounting processes and data records are recorded on the blockchain, which is open and transparent, facilitating supervision and traceability by all parties, and enhancing the credibility of the accounting result; the detailed data flow information recorded on the blockchain can clearly trace the responsibility subject of each emission link, facilitating carbon emission management and accountability; the dynamic form and the updateable carbon emission factor library enable the platform to adapt to different types and scales of large-scale events, and update the accounting standards in a timely manner.

[0091] Embodiment 2: As shown in the figure, the collection module comprises: Figure 2

[0092] A first collection sub-module is configured to obtain first data based on real-time data of energy consumption and transportation of Internet of Things devices.

[0093] ​​The second acquisition sub-module is configured to intelligently recognize the bill information based on an OCR+NLP technology to obtain second data.

[0094] The determining sub-module is configured to take the first data and the second data as carbon emission data of the large-scale event.

[0095] The working principle and beneficial effects of the above technical solution are as follows: the first acquisition sub-module acquires real-time data of the Internet of Things device, and can obtain dynamic information of energy consumption and traffic travel, which reflects the instant carbon emission during the large-scale event. The second acquisition sub-module starts from bill information, and can mine out supplementary information such as historical records of energy consumption and carbon emission related to the event. The combination of the two makes the carbon emission data more comprehensive, avoids the possible omissions of a single data source, and thus more accurately reflects the overall carbon emission of the large-scale event.

[0096] Embodiment 3: a method for constructing an intelligent accounting model, comprising:

[0097] Based on the LLM model, the carbon emission factor in the carbon emission factor library is subjected to semantic understanding, and the carbon emission factor text description information is converted into structured data;

[0098] The updating information of the carbon emission factor library is dynamically acquired, the updating information of the carbon emission factor library is analyzed based on the LLM model to generate a correction instruction, and the factor mapping layer parameter of the LLM model is adjusted;

[0099] A bidirectional interactive collaborative training framework is constructed based on the LLM model and the dynamically updated carbon emission factor library;

[0100] A time attention mechanism is added to optimize the bidirectional interactive collaborative training framework, the self-attention layer of the bidirectional interactive collaborative training framework is fused with a carbon emission factor time decay model, and the weight coefficient of the learning model fusion is inversely propagated based on the updating frequency of the dynamically updated carbon emission factor library;

[0101] A carbon emission factor-emission correlation matrix is constructed in the bidirectional interactive collaborative training framework, with the historical version of the dynamic factor library as the horizontal axis and the activity emission result as the vertical axis, and the nonlinear relationship between the carbon emission factor change and the emission result is learned by the built-in basic model of the bidirectional interactive collaborative training framework;

[0102] The accounting logic chain of the bidirectional interactive collaborative training framework is defined, and the nonlinear parameters of the correlation matrix are embedded;

[0103] The bidirectional interactive collaborative training framework is trained end to end based on the accounting logic chain to obtain an initial intelligent accounting model;

[0104] The training data set and the test data set of the intelligent accounting model are acquired;

[0105] iteratively training the initial intelligent accounting model based on the training data set of the intelligent accounting model to obtain a target intelligent accounting model;

[0106] testing the target intelligent accounting model based on the test data set, and obtaining the trained intelligent accounting model when the test result meets the requirements.

[0107] In this embodiment, the basic model is a neural network model.

[0108] In this embodiment, the example accounting logic chain is: logic for calculating carbon emissions, such as: calculating participant transportation emissions: number of people x average mileage x transportation factor; superimposing venue power emissions: power consumption x real-time power factor.

[0109] The working principle and beneficial effects of the above technical solution are: through dynamic updating of the carbon emission factor library and corresponding adjustment of the LLM model, the changes in carbon emission factors can be timely adapted. This enables the intelligent accounting model to use the latest and most accurate carbon emission factor data when calculating carbon emissions, improving the accuracy of carbon emission accounting; adding a time attention mechanism and integrating a carbon emission factor time decay model can fully consider the changes in carbon emission factors over time. For some factors that change significantly over time (such as the reduction in carbon emission factors due to the promotion of new energy sources), they can be more accurately reflected in the accounting results, improving the adaptability of the model to carbon emission accounting at different time stages; constructing a carbon emission factor-emission correlation matrix and learning non-linear relationships can help capture complex internal relationships in carbon emission calculations. This is very important for accurately accounting for carbon emissions in complex situations such as large-scale events, as carbon emissions are often influenced by multiple factors. The capture of this non-linear relationship can improve the accuracy of the accounting results; through end-to-end training and iterative training and testing based on the training data set and the test data set, the performance of the intelligent accounting model can be continuously optimized. Ensuring that the model has high accuracy, stability, and reliability in actual application, thereby providing an effective tool for carbon emission accounting.

[0110] Embodiment 4: further comprising a traceability module for generating traceable data ledgers based on blockchain technology to record key data and results during all collection, processing, and accounting processes.

[0111] Embodiment 5: further comprising a visualization module for constructing Web / App end interfaces for different roles of event organizers, participants, and regulatory authorities to visually display carbon emission overviews, detailed accounting reports, emission trend analysis, and carbon footprints of large-scale events.

[0112] Embodiment 6: the preprocessing module includes:

[0113] a data cleaning submodule, configured to clean the carbon emission data of the large-scale event;

[0114] a data verification module, configured to verify the carbon emission data of the large-scale event after cleaning, and take the carbon emission data of the large-scale event that passes the verification as the preprocessed carbon emission data of the large-scale event.

[0115] The working principle and beneficial effects of the above technical solution are as follows: through the operation of the data cleaning submodule, noise in the data can be removed, and missing data can be supplemented, so that the data is more accurate and complete. The data verification module further ensures the rationality of the data, thereby improving the quality of the large-scale event carbon emission data as a whole; high-quality data is the basis for subsequent accurate carbon emission analysis, accounting, etc. The data verification process is based on strict rules and standards, which can exclude data that does not conform to logic or is unreasonable, so that the preprocessed carbon emission data of the large-scale event obtained finally is more reliable; reliable data can provide strong basis for relevant decisions (such as the formulation of carbon emission management strategies for large-scale events, the setting of emission reduction targets, etc.), avoiding decision-making errors caused by data errors; the processing of unit inconsistency and other issues in the data cleaning process, as well as the inspection of data format and other aspects in the data verification process, all help to enhance the consistency of the large-scale event carbon emission data; in the subsequent data processing, analysis and sharing process, consistent data facilitates the interaction and collaborative work between different systems or modules, improving the efficiency of the entire carbon emission management process.

[0116] Embodiment 7: The data cleaning submodule comprises:

[0117] a first calculation unit, configured to:

[0118] calculate the abnormality possibility of the carbon emission data of each dimension in the multi-dimensional carbon emission data of the large-scale event carbon emission data, to obtain an abnormality possibility value of the carbon emission data of each dimension; and determine a plurality of dimension data to be cleaned based on the abnormality possibility value;

[0119] a second calculation unit, configured to:

[0120] calculate the correlation index of any two dimension data to be cleaned, and determine the weight of each dimension data to be cleaned based on the correlation index;

[0121] align the plurality of dimension data to be cleaned based on time series; wherein the parameters of different dimensions to be cleaned corresponding to the same time are taken as a data point;

[0122] determine a comprehensive evaluation index of each data point based on the different dimension parameter values of each data point and the weight of each dimension data to be cleaned;

[0123] an abnormality evaluation unit, configured to:

[0124] a difference value of the comprehensive evaluation index of any two adjacent data points is calculated to obtain a plurality of difference values;

[0125] a mean value of the plurality of difference values is calculated to obtain a difference value threshold;

[0126] the difference values are compared with the preset difference value threshold respectively, and the data points with a difference value greater than or equal to the preset difference value threshold are taken as abnormal data points to obtain a plurality of abnormal data points;

[0127] a third calculation unit is configured to calculate an abnormal degree value of each parameter in each abnormal data point, and determine a plurality of abnormal parameters in different dimensions based on the abnormal degree value;

[0128] a data cleaning unit is configured to:

[0129] obtain a preset data cleaning rule;

[0130] perform data cleaning on the plurality of abnormal parameters in different dimensions based on the preset data cleaning rule, to complete data cleaning of the carbon emission data of the large-scale event.

[0131] The working principle and beneficial effects of the above technical solutions are as follows: first, for the multi-dimensional carbon emission data in the large event carbon emission data (for example, it can include different dimensions such as event site energy consumption, personnel transportation carbon emission, and event material production carbon emission), the abnormal possibility of each dimension carbon emission data is calculated; this can be determined by some statistical method or compared with historical data; for example, if the data of a certain dimension deviates too much from the historical average data of the dimension, it can be determined to have a higher abnormal possibility; according to the calculated abnormal possibility value, the dimension data to be cleaned that may exist abnormality is determined; for the determined dimension data to be cleaned, the correlation index between them is calculated; if there is a strong correlation between two dimension data, such as the lighting electricity and air conditioning electricity of the event site, which can have similar change trend in time, the correlation between them is higher. According to the correlation index, the weight of each dimension data to be cleaned is determined, and the dimension data with higher correlation can have higher weight in the comprehensive evaluation; then the dimension data to be cleaned is aligned according to the time sequence, and different dimensions corresponding to the same time are combined into a data point; for example, at a certain time, the electric power carbon emission of the event site, the fuel oil carbon emission of personnel transportation and other parameters form a data point; finally, according to the different dimension parameter values of each data point and the weight of each dimension data to be cleaned, the comprehensive evaluation index of each data point is determined by weighted calculation and other ways; the difference between the comprehensive evaluation indexes of any two adjacent data points is calculated to obtain a series of difference values; then the mean value of the difference values is calculated as the difference threshold value; each difference value is compared with the preset difference threshold value, and if the difference value is greater than or equal to the preset difference threshold value, the data point is determined as an abnormal data point; in this way, the data point with large fluctuation in the comprehensive evaluation index can be found, and these data points can exist data abnormality; for each abnormal data point, the abnormal degree value of each parameter is calculated. This can be determined by comparing the difference between the parameter and the same dimension parameter in other normal data points; according to the abnormal degree value, the parameters existing abnormality in different dimensions are determined; through multi-dimensional analysis and cleaning, the abnormal value in the carbon emission data can be more accurately identified and corrected; compared with single dimension judgment, this multi-dimensional comprehensive consideration method can avoid misjudgment, so that the cleaned data can more accurately reflect the real carbon emission of the large event, and provide more reliable data support for the evaluation, management and reduction strategy of carbon emission.

[0132] Embodiment 8: a first calculation unit, comprising:

[0133] A first calculation subunit is configured to:

[0134] Any one dimension of carbon emission data is taken as target dimension data;

[0135] obtaining an extreme value in the target dimension data; segmenting the target dimension data based on the extreme value to obtain target dimension sub-data;

[0136] calculating a ratio of a number of data points in each target dimension sub-data to a number of data points in the target dimension data to obtain a first ratio;

[0137] calculating a ratio of a standard deviation to a mean of data points in each target dimension sub-data to obtain a second ratio;

[0138] multiplying the first ratio and the second ratio to obtain a stage evaluation value of the target dimension sub-data;

[0139] summing the stage evaluation values of all target dimension sub-data to obtain an anomaly likelihood value;

[0140] a first comparison subunit, configured to compare the anomaly likelihood value with a preset anomaly likelihood threshold value, and take carbon emission data of a corresponding dimension as to-be-cleaned dimension data when the anomaly likelihood value is greater than or equal to the preset anomaly likelihood threshold value, to obtain a plurality of to-be-cleaned dimension data.

[0141] The working principle and beneficial effects of the above technical solution are as follows: for carbon emission data of any target dimension, first, an extreme value is obtained; the target dimension data is segmented using the extreme values to divide the entire target dimension data into different intervals to obtain target dimension sub-data; a ratio of a number of data points in each target dimension sub-data to a number of data points in the target dimension data (a first ratio) is calculated, which reflects the proportion of each sub-data in the entire data set; at the same time, a ratio of a standard deviation to a mean of data points in each target dimension sub-data (a second ratio) is calculated, which can measure the dispersion degree of the data; the first ratio and the second ratio are multiplied to obtain a stage evaluation value of the target dimension sub-data; the stage evaluation value comprehensively considers the size and dispersion degree of the sub-data; finally, the stage evaluation values of all target dimension sub-data are summed to obtain an anomaly likelihood value of the target dimension data. The anomaly likelihood value comprehensively reflects the anomaly likelihood degree of the entire target dimension data; in calculating the anomaly likelihood value, the size and dispersion degree of the data are comprehensively considered; the comprehensive consideration of these multiple factors can more comprehensively evaluate the anomaly likelihood of each dimension data.

[0142] Embodiment 9: the third calculation unit comprises:

[0143] the second calculation subunit is configured to:

[0144] arbitrarily obtaining a parameter of a dimension in an anomaly data point as a to-be-evaluated parameter;

[0145] obtaining a mean value of data of the dimension in which the to-be-evaluated parameter is located;

[0146] Calculate the difference between the parameter value of each dimension in the abnormal data point and the data mean of the same dimension, to obtain a plurality of difference values;

[0147] Calculate the mean of the plurality of difference values to obtain the difference mean corresponding to the abnormal data point;

[0148] According to the to-be-evaluated parameter, the data mean and the difference mean, determine the abnormal degree value of the to-be-evaluated parameter based on a preset algorithm;

[0149] Traverse all parameter values of the abnormal data point to obtain the abnormal degree value corresponding to each parameter;

[0150] The second comparison subunit is configured to compare the abnormal degree value with a preset abnormal degree threshold, and take the parameter when the abnormal degree value is greater than or equal to the preset abnormal degree threshold as an abnormal parameter, to obtain a plurality of abnormal parameters of different dimensions.

[0151] In this embodiment, the preset algorithm comprises:

[0152]

[0153] Wherein, M i,j represents the abnormal degree value of the jth to-be-evaluated parameter in the ith abnormal data point; T i,j represents the data value of the jth to-be-evaluated parameter in the ith abnormal data point; μ i represents the difference mean corresponding to the ith abnormal data point; represents the data mean of the dimension where the jth to-be-evaluated parameter in the ith abnormal data point is located; T i,j,k represents the data value of the same dimension as the jth to-be-evaluated parameter in the kth neighborhood data point of the ith abnormal data point; σ i represents the standard deviation of the ith abnormal data point; N represents the total number of neighborhood data points.

[0154] The working principle and beneficial effects of the above technical solution are: by comprehensively considering the mean value (first mean value) of the dimension where the to-be-evaluated parameter is located and the comprehensive degree (second mean value) of the entire abnormal data point deviating from the mean value of each dimension, the abnormal degree value is determined, which can more accurately quantify the abnormal degree of each parameter in the abnormal data point; this is more comprehensive and accurate than simply comparing the difference between the parameter and the mean value, because it considers the overall deviation of the entire abnormal data point; comparison with the preset abnormal degree threshold can effectively screen out the real abnormal parameters; this helps to focus the data cleaning on those parameters with greater impact on data quality and higher abnormal degree, improves the pertinence and effectiveness of data cleaning, and avoids unnecessary cleaning operation on parameters with lower abnormal degree.

[0155] To achieve the above object, the first aspect of the present application provides a large-scale event carbon emission intelligent accounting method, comprising steps S1-S4:

[0156] S1: collecting carbon emission data of a large-scale event;

[0157] S2: preprocessing the carbon emission data of the large-scale event;

[0158] S3: constructing an intelligent accounting model based on an LLM model and a dynamically updated carbon emission factor library;

[0159] S4: inputting the preprocessed carbon emission data of the large-scale event into the intelligent accounting model for accounting to determine an accounting result.

[0160] The working principle and beneficial effects of the above technical solution are as follows: a large-scale event carbon emission intelligent accounting platform is constructed by integrating multiple technologies, automatic collection and accounting of carbon emission data are realized through Internet of Things devices and bill intelligent recognition technology, manual intervention is greatly reduced, and accounting efficiency is significantly improved; the bill intelligent recognition technology can accurately extract bill information, and the intelligent analysis of the intelligent accounting model improves the accuracy of data collection; the introduction of the blockchain technology ensures the non-tamperability of the data, and guarantees the authenticity of the data from the source; the intelligent accounting model can understand and analyze complex activity scenarios, and combined with the dynamically updated carbon emission factor library, multi-dimensional and fine carbon emission calculation is performed, and the result is more accurate; all accounting processes and data records are recorded on the blockchain, which is open and transparent, convenient for supervision and tracing by all parties, and enhances the credibility of the accounting result; the detailed data flow information recorded on the blockchain can clearly trace the responsibility subject of each emission link, which is convenient for carbon emission management and accountability; the dynamic form and the updateable carbon emission factor library enable the platform to adapt to large-scale events of different types and different scales, and update the accounting standards in a timely manner.

[0161] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A large-scale event carbon emission intelligent accounting platform, characterized in that, include: The data acquisition module is used to collect carbon emission data from large-scale events. A preprocessing module is used to preprocess the carbon emission data of the large-scale event; The module is used to build an intelligent accounting model based on the LLM model and a dynamically updated carbon emission factor library; The accounting module is used to input the pre-processed carbon emission data of large-scale events into the intelligent accounting model for accounting and to determine the accounting results.

2. The large-scale event carbon emission intelligent accounting platform of claim 1, wherein, The data acquisition module includes: The first data acquisition submodule is used to obtain the first data based on real-time data of energy consumption and traffic of IoT devices; The second data acquisition submodule is used to intelligently identify ticket information based on OCR+NLP technology to obtain the second data. The determination submodule is used to use the first and second data as carbon emission data for large-scale events.

3. The large-scale event carbon emission intelligent accounting platform of claim 1, wherein, The construction methods for intelligent accounting models include: Based on the LLM model, semantic understanding of carbon emission factors in the carbon emission factor database is performed, and the textual description information of carbon emission factors is transformed into structured data. Dynamically acquire updated information of the carbon emission factor database, parse the updated information of the carbon emission factor database based on the LLM model to generate correction instructions, and adjust the parameters of the factor mapping layer of the LLM model. A bidirectional interactive collaborative training framework is constructed based on the LLM model and a dynamically updated carbon emission factor library. A time attention mechanism is added to optimize the bidirectional interactive collaborative training framework. The self-attention layer of the bidirectional interactive collaborative training framework is fused with the carbon emission factor time decay model. The weight coefficients of the fused model are learned by backpropagation based on the update frequency of the dynamically updated carbon emission factor library. Within the bidirectional interactive collaborative training framework, a correlation matrix between carbon emission factors and emissions is constructed, with the historical versions of the dynamic factor library as the horizontal axis and the active emission results as the vertical axis. The nonlinear relationship between changes in carbon emission factors and emission results is learned through the built-in basic model of the bidirectional interactive collaborative training framework. The computational logic chain of the bidirectional interactive collaborative training framework is defined, and the nonlinear parameters of the correlation matrix are embedded. The bidirectional interactive collaborative training framework is trained end-to-end based on the accounting logic chain to obtain the initial intelligent accounting model. Obtain the training and test datasets for the intelligent accounting model; The initial intelligent accounting model is iteratively trained based on the training dataset of the intelligent accounting model to obtain the target intelligent accounting model; The target intelligent accounting model is tested based on the test dataset. When the test results meet the requirements, the trained intelligent accounting model is obtained.

4. The large-scale activity carbon emission intelligent accounting platform of claim 1, wherein, Also includes: The traceability module is used to generate a traceable data ledger based on blockchain technology, recording key data and results in all collection, processing and accounting processes. 5.The large-scale event carbon emission intelligent accounting platform of claim 1, wherein, Also includes: The visualization module is used to build web / app interfaces for different roles such as event organizers, participants, and regulatory authorities, and to visualize the carbon emission overview, detailed accounting reports, emission trend analysis, and carbon footprint of large-scale events.

6. The large-scale event carbon emission intelligent accounting platform of claim 1, wherein, The preprocessing module includes: The data cleaning submodule is used to clean the carbon emission data of the large-scale event. The data verification module is configured to verify the carbon emission data of the large-scale event after cleaning, and the carbon emission data of the large-scale event that passes the verification is taken as the preprocessed carbon emission data of the large-scale event.

7. The large-scale event carbon emission intelligent accounting platform according to claim 6, wherein, The data cleaning sub-module comprises: The first calculation unit is configured to: calculate the abnormal possibility of the carbon emission data of each dimension in the multi-dimensional carbon emission data of the large-scale event, to obtain an abnormal possibility value of the carbon emission data of each dimension; and determine a plurality of dimension data to be cleaned based on the abnormal possibility value. The second calculation unit is configured to: calculate the correlation index of any two dimension data to be cleaned, and determine the weight of each dimension data to be cleaned based on the correlation index; align the plurality of dimension data to be cleaned based on time series; wherein parameters of different dimensions corresponding to the same time point are taken as a data point; determine a comprehensive evaluation index of each data point based on the different dimension parameter values of each data point and the weight of each dimension data to be cleaned; The abnormal evaluation unit is configured to: calculate the difference value of the comprehensive evaluation indexes of any two adjacent data points, to obtain a plurality of difference values; calculate the mean value of the plurality of difference values, to obtain a difference threshold value; compare the difference values with the preset difference threshold value respectively, and take the data points with a difference value greater than or equal to the preset difference threshold value as abnormal data points, to obtain a plurality of abnormal data points; The third calculation unit is configured to calculate the abnormal degree value of each parameter in each abnormal data point; and determine a plurality of abnormal parameters of different dimensions based on the abnormal degree value. The data cleaning unit is configured to: obtain a preset data cleaning rule; clean the abnormal parameters of different dimensions based on the preset data cleaning rule, to complete the data cleaning of the carbon emission data of the large-scale event.

8. The large-scale event carbon emission intelligent accounting platform of claim 7, wherein, The first calculation unit comprises: The first calculation sub-unit is configured to: take carbon emission data of any dimension as target dimension data; obtain the extreme value in the target dimension data; and segment the target dimension data based on the extreme value, to obtain target dimension sub-data; calculate the ratio of the number of data points in each target dimension sub-data to the number of data points in the target dimension data, to obtain a first ratio value; calculate the ratio of the standard deviation to the mean value of the data points in each target dimension sub-data, to obtain a second ratio value; take the product of the first ratio value and the second ratio value as the stage evaluation value of the target dimension sub-data; take the sum value of the stage evaluation values of all target dimension sub-data as the abnormal possibility value; The first comparison sub-unit is configured to compare the abnormal possibility value with a preset abnormal possibility threshold value, and take the carbon emission data of the corresponding dimension as dimension data to be cleaned when the abnormal possibility value is greater than or equal to the preset abnormal possibility threshold value, to obtain a plurality of dimension data to be cleaned. 9.The large-scale event carbon emission intelligent accounting platform of claim 8, wherein, The third calculation unit comprises: The second calculation sub-unit is configured to: arbitrarily take a parameter of a dimension in an abnormal data point as a parameter to be evaluated; take the mean value of the data in the dimension of the parameter to be evaluated; calculate the difference value of the parameter value of each dimension in the abnormal data point and the mean value of the data of the same dimension, to obtain a plurality of difference values; calculate the mean value of the plurality of difference values, to obtain the difference value mean value corresponding to the abnormal data point; and determine the abnormal degree value of each parameter in the abnormal data point based on the difference value mean value corresponding to the abnormal data point and the mean value of the data of the same dimension. According to the to-be-evaluated parameter, the data mean value and the difference mean value, an abnormality degree value of the to-be-evaluated parameter is determined based on a preset algorithm; All parameter values of the abnormal data points are traversed to obtain an abnormality degree value corresponding to each parameter; The second comparison subunit is configured to compare the abnormality degree value with a preset abnormality degree threshold value, and take the parameter when the abnormality degree value is greater than or equal to the preset abnormality degree threshold value as an abnormal parameter to obtain abnormal parameters of different dimensions. 10.The accounting method of the large-scale event carbon emission intelligent accounting platform according to any one of claims 1-9, characterized in that, It comprises: Collecting carbon emission data of large-scale activities; Preprocessing the carbon emission data of the large-scale activities; Constructing an intelligent accounting model based on an LLM model and a dynamically updated carbon emission factor library; Inputting the preprocessed carbon emission data of the large-scale activities into the intelligent accounting model for accounting to determine an accounting result.

Images