Multi-type load carbon transaction and data aggregation method and system

Abstract

The invention relates to the technical field of carbon transaction, in particular to a multi-type load-based carbon transaction and data aggregation method and system, and the method comprises the steps: S1, collecting multi-type load data; s2, data preprocessing; s3, extracting data features; s4, aggregating multi-type load data; s5, constructing a carbon transaction model; s6, optimizing a carbon transaction strategy; and S7, executing the carbon transaction strategy and performing data feedback. According to the method, acquisition equipment is deployed at each load terminal, multi-type load data are comprehensively acquired, in the acquisition process, the system extracts same-period and same-type historical data from a historical database and calculates the Euclidean distance to evaluate the data similarity, and when the similarity exceeds a set threshold value, a data anomaly early warning mechanism is triggered, so that the data anomaly early warning is realized. A detailed exception report is automatically generated, operation and maintenance personnel are assisted to quickly position the root of a problem, a reasonable threshold value is set for collected data, data integrity and accuracy are judged, and effective data are provided for subsequent carbon transaction model construction and strategy optimization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon trading, in particular to a multi-type load carbon trading and data aggregation method and system. BACKGROUND

[0002] Under the severe situation of global climate change today, carbon trading as a market mechanism aims to promote enterprises to reduce greenhouse gas emissions through economic means and achieve carbon emission reduction targets. The multi-type load carbon trading and data aggregation method and system emerged as the times require, and its core lies in collecting, processing and integrating the carbon emission data generated by different types of loads such as industry, commerce and residents, and on this basis, building a carbon trading model to provide accurate strategies and decisions for market participants.

[0003] With the improvement of data quality requirements in various industries, the existing technology has exposed many problems in processing multi-type load carbon trading data. On the one hand, there are defects such as insufficient collection accuracy and poor real-time performance in the data collection link, which is difficult to meet the strict requirements of carbon trading market on data timeliness and accuracy; on the other hand, the data processing and aggregation process lacks effective feature extraction and analysis means, which is difficult to deeply mine data value and provide effective data for carbon trading model. In addition, the existing system fails to fully consider market dynamic changes and policy influences when building a carbon trading model, resulting in insufficient adaptability and forward-looking of the model. Therefore, we propose a multi-type load carbon trading and data aggregation method and system. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a multi-type load carbon trading and data aggregation method and system, thereby solving the technical problems mentioned in the background art.

[0005] To achieve the above purpose, the present application is realized by the following technical scheme:

[0006] The multi-type load carbon trading and data aggregation method comprises the following steps:

[0007] S1, multi-type load data collection: deploying intelligent electric meters and carbon emission monitoring sensors at load terminals, collecting current data containing time stamp, load type, power consumption and carbon emission through 5G and NB-IoT technology, synchronously extracting historical data, calculating the Euclidean distance of current data and historical data to evaluate similarity, and judging the integrity and accuracy of the collected data according to the similarity;

[0008] S2, data preprocessing: cleaning the collected data to remove outliers and duplicate data, converting the data format of the cleaned data to JSON, standardizing the data by using a normalization formula, and analyzing fluctuation trend combined with the moving average benchmark of historical data;

[0009] S3. Data Feature Extraction: The sliding window method is used to extract time series features, the least squares method is used to analyze carbon emission trends, the ARIMA model is trained based on historical feature sets to predict current features, and the correlation between features and carbon trading is evaluated.

[0010] S4. Multi-type load data aggregation: The K-means clustering algorithm is used, combined with the elbow rule and historical clustering results to determine the number of clusters, calculate the distance between data points and cluster centers to perform clustering, and evaluate the rationality of the current clustering by referring to historical clustering results;

[0011] S5. Construct a carbon trading model: Based on aggregated data, establish a cost function that includes carbon price and carbon emissions, add electricity demand and carbon quota constraints, optimize model parameters using the seasonal decomposition of historical carbon prices, and construct a carbon trading model based on the optimized model parameters.

[0012] S6. Optimize carbon trading strategies: Use genetic algorithms to solve carbon trading models, evaluate the quality of strategies through fitness functions, optimize strategies by combining historical strategy libraries, and verify the effectiveness of strategies using Monte Carlo simulations.

[0013] S7. Execute carbon trading strategies and provide data feedback: Utilize blockchain technology to execute transactions, monitor data in real time and compare it with expectations, store the data in the historical database after the transaction is completed, compare the evaluation results with historical cases and update the model parameters.

[0014] Preferably, in step S1, when judging the completeness and accuracy of the collected data, the completeness threshold is... Set to 0.9, accuracy threshold Set to 0.95, if the integrity ratio ≥ And accuracy ratio ≥ If the data is found to be qualified, proceed to step S2; if the conditions are not met, the system will automatically re-execute step S1. If the data still does not meet the standards after three re-collections, the data collection operation will be stopped, and technical personnel will be notified in a timely manner to inspect the relevant equipment and investigate the data source.

[0015] Preferably, in step S2, the quartile method is used. Identify outliers in electricity consumption data; use normalization formulas Electricity consumption data Processing is carried out, among which, and These are the minimum and maximum values ​​in this set of electricity consumption data. The data is standardized; and quality assessment indicators are calculated for the preprocessed data. Set quality assessment thresholds =0.85, when ≥ If the data preprocessing is deemed satisfactory, the system can proceed to step S3; otherwise, it will return to step S1 to re-collect data and strengthen the data verification rules during the re-collection process.

[0016] Preferably, in step S3, the sliding window size is set to 1 hour, i.e., 60 minutes of data points are used to extract features, and the average value within the window is calculated. Predict features and calculate dissimilarity using the ARIMA model. ,in, To actually extract the feature set, Set the feature set length; set the feature validity threshold. The predicted difference threshold is 0.7. When it is 0.15, ≥ and ≤ hour, If the Pearson correlation coefficient is obtained, proceed to step S4; if the conditions are not met, the system will return to step S2 to re-preprocess the data and attempt to adjust the preprocessing parameters.

[0017] Preferably, in step S4, the K-means clustering algorithm is used to calculate the distance between data points and the centers of each cluster. ,in, For the first The first data point 1 eigenvalue, For the first The first data point One feature value; similarity coefficient ,in This is the current clustering result. For historical clustering results; set a threshold for the reasonableness of aggregation. The similarity threshold for historical clustering is 0.75. Set to 0.7, when the intra-cluster similarity ≥ Inter-cluster discrimination ≥ and ≥ If the aggregation result is deemed reasonable, the system can proceed to step S5; otherwise, the system will return to step S4 to readjust the clustering parameters.

[0018] Preferably, in step S5, a carbon trading cost function is established. ,in, For the first The carbon valence corresponding to each load cluster For the first Carbon emissions of a load cluster; set a feasibility evaluation index feasibility threshold is set to 0.8, when ≥ , it is considered that the model is feasible and can enter step S6; if it is not feasible, the system will return to step S5 to adjust the model parameters again, and if the model is still not feasible after multiple adjustments, it is necessary to further check whether there is deviation in the data aggregation result, and if necessary, return to step S4 to perform data aggregation again.

[0019] Preferably, in step S6, when the effectiveness of the strategy is verified by Monte Carlo simulation, the effectiveness verification threshold is set to 0.9, and the ratio of the actual return to the expected return is calculated , when ≥ , it is considered that the optimization result is effective, and step S7 can be entered; if the condition is not met, the system will return to step S6 to optimize the strategy again, and if the result is still not up to standard after multiple optimizations, the carbon trading model needs to be reexamined and returned to step S5 for model adjustment.

[0020] Preferably, in step S7, when the evaluation result is compared with the historical case, the evaluation index is set, the evaluation threshold is set to 0.85, when ≥ , it is determined that the transaction is successful, and the process is ended; if it does not meet the standard, the system returns to step S6 to optimize the transaction strategy again.

[0021] A multi-type load carbon trading and data aggregation system for executing the multi-type load carbon trading and data aggregation method described above, characterized in that the system comprises:

[0022] A data acquisition module for realizing real-time acquisition of industrial, commercial, and residential multi-type load data through deployment of terminal equipment, and verifying with historical data, including a data acquisition unit and a historical data comparison unit;

[0023] A data preprocessing module for cleaning, format conversion, and standardization processing of the collected data, identifying abnormalities in combination with historical data fluctuation rules, including a data cleaning and conversion unit and a data standardization unit;

[0024] A data feature extraction module for extracting load features related to carbon trading from the preprocessed data, verifying in combination with historical feature model predicted values, including a feature extraction unit and a feature prediction and verification unit;

[0025] The data aggregation module is used to cluster and aggregate multiple types of load data based on load characteristics and historical clustering results using the K-means algorithm, and to evaluate the rationality of the aggregation. It includes a clustering algorithm unit and a clustering verification unit.

[0026] The carbon trading model building module is used to build a carbon trading model containing cost functions and constraints based on aggregated data, and to optimize model parameters by combining historical carbon price trends. It includes a model building unit and a model optimization unit.

[0027] The carbon trading strategy optimization module is used to solve the carbon trading model based on the genetic algorithm, optimize the trading strategy by combining the historical strategy library, and verify the feasibility of the strategy through simulation. It includes a genetic algorithm unit and a strategy verification unit.

[0028] The transaction execution and feedback module is used to execute carbon transactions based on optimized strategies, monitor transaction data in real time, compare historical results to generate feedback, and update historical databases and model parameters. It includes a transaction execution unit and a result evaluation and feedback unit.

[0029] Beneficial effects compared to existing technologies:

[0030] 1. This invention utilizes specialized equipment deployed at various load terminals and leverages technologies such as 5G and narrowband IoT to achieve comprehensive data collection across multiple load types. During data collection, the system simultaneously extracts historical data of the same period and type from a historical database, calculates Euclidean distance to assess data similarity, and triggers a data anomaly warning mechanism when the similarity exceeds a set threshold, automatically generating a detailed anomaly report to assist maintenance personnel in quickly locating the root cause of problems. Whether it's equipment failure, data transmission errors, or special changes in actual load, these issues can be detected and addressed promptly. Simultaneously, the system assesses the completeness and accuracy of the collected data, setting reasonableness thresholds to determine data integrity and accuracy, thereby providing accurate and effective data for subsequent carbon trading model construction and strategy optimization.

[0031] 2. This invention employs the quartile method to accurately identify outliers in electricity consumption data and uses the moving average method combined with historical data fluctuation patterns to analyze current data fluctuation trends, eliminating abnormal data fluctuations caused by factors such as equipment failure and large-scale events. In the feature extraction stage, the sliding window method is used to extract time-series features and calculate statistics within the window to reflect the true load fluctuations. Simultaneously, the least squares method is used to perform trend analysis on carbon emission data, extracting characteristic parameters such as trend slope and intercept to understand the patterns of carbon emission changes. Furthermore, the ARIMA feature prediction model built based on historical data feature sets calculates the difference between actual extracted features and predicted features, and combines this with the Pearson correlation coefficient to comprehensively evaluate feature effectiveness, continuously optimizing feature extraction results and significantly improving the model's accuracy.

[0032] 3. When constructing the carbon trading model, this invention comprehensively considers multiple factors such as carbon emissions, electricity consumption, and market carbon prices for each load cluster. It utilizes seasonal difference analysis of historical carbon trading data to predict carbon price fluctuation patterns and trends. After the transaction is completed, all process data is stored in a historical database. An experience summary report is generated through comparative and attribution analysis, evaluating the trading results from multiple dimensions. Based on the trading results, the historical data summary method and prediction model parameters are adjusted to achieve continuous optimization and feedback adjustment of the carbon trading strategy. This continuously improves the efficiency and effectiveness of carbon trading, ensuring that market participants maximize their economic benefits while meeting emission reduction targets. Attached Figure Description

[0033] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0034] Figure 1 This is a schematic diagram of the multi-type load carbon trading and data aggregation method of the present invention;

[0035] Figure 2 This is a schematic diagram of the framework of the multi-type load carbon trading and data aggregation system of the present invention. Detailed Implementation

[0036] Preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention can also be implemented in various different forms, and therefore the present invention is not limited to the embodiments described below. In addition, for the purpose of more clearly describing the present invention, parts not connected to the invention will be omitted from the drawings.

[0037] The technical solutions in this application are designed to address the problems described in the background, and are generally as follows:

[0038] Example 1:

[0039] This embodiment introduces a method for multi-type load carbon trading and data aggregation, including the following steps:

[0040] S1. Multi-type load data acquisition:

[0041] The core task of this step is to collect comprehensive and high-precision load data from various types of loads participating in carbon trading, including industrial, commercial, and residential loads. At each load terminal, specialized equipment such as smart meters and carbon emission monitoring sensors are deployed based on different load characteristics and monitoring needs. For example, large industrial enterprises will install high-precision smart meters and multi-parameter carbon emission monitoring equipment to monitor their electricity consumption and carbon emissions in real time during production. In residential communities, distributed smart meters and simplified carbon emission estimation devices are used to monitor residents' electricity consumption and corresponding carbon emissions. These devices, leveraging technologies such as 5G and Narrowband Internet of Things (NB-IoT), transmit the collected electricity consumption and carbon emission data to the data center in real time.

[0042] The collected data includes key elements such as timestamps, load types, electricity consumption, and carbon emissions, represented by sets. ,in, Represents a timestamp, accurate to the minute. It is used to accurately record the time of data collection, providing a time dimension reference for subsequent data analysis and processing; Indicates the load type. Where 1 corresponds to industrial load, 2 corresponds to commercial load, 3 corresponds to residential load, etc., clearly defining the load category to which the data belongs, which facilitates classification analysis and processing; This indicates electricity consumption, measured in kilowatt-hours (kWh). It reflects the power consumption of the load within a specific time period; It represents carbon emissions, and the unit is kilograms of carbon dioxide equivalent (kgCO2e). This reflects the carbon emissions generated by the corresponding electricity consumption behavior.

[0043] During the data acquisition process, in order to better determine the rationality and anomalies of the currently collected data, the system will simultaneously extract historical data with the same time period and the same load type from the historical database to construct a historical dataset. For example, when collecting minute-by-minute electricity consumption and carbon emission data for an industrial park on a given day, the system will extract historical data from the same dates (distinguishing between weekdays and weekends) and time periods over the past year. This data is then used to calculate the Euclidean distance between the currently collected data and the historical data. To assess the similarity between the two, The historical data similarity threshold set for the current data set. If it is 0.8, If the similarity to historical data exceeds the set threshold of 0.8, the system will immediately trigger a data anomaly warning mechanism. At this time, the system will automatically generate a detailed anomaly report. The report will not only list the specific abnormal data points, but also compare and analyze the differences between these data points and historical data, prompting maintenance personnel to conduct manual verification to determine whether the data anomaly is due to equipment failure, data transmission errors, or special changes in the actual load, such as industrial enterprises temporarily adding production shifts.

[0044] After data collection is complete, the completeness and accuracy of the collected data are assessed. A completeness threshold is set. and accuracy threshold Generally speaking, Set it to 0.9. Set it to 0.95. Calculate the integrity ratio of the collected data. For example, if a region should have 1000 data points collected, but only 950 valid data points were actually collected, and none of these data points are missing key fields, then... =0.95; Accuracy Ratio The accuracy of the data is determined through comparison with data from authoritative metering equipment and logical verification. Taking electricity consumption and carbon emissions as examples, there is a theoretical correlation between the two; if the collected data does not conform to this correlation, it is considered inaccurate. ≥ and ≥ If the data is deemed acceptable, proceed to step S2; otherwise, the system automatically re-executes step S1, up to a maximum of three times. If the data still fails to meet the standards after three re-collections, the data collection operation is stopped, and technical personnel are promptly notified to inspect the relevant equipment and investigate the data source to ensure the quality of subsequent data collection.

[0045] S2, Data Preprocessing:

[0046] The data preprocessing step involves cleaning, transforming, and standardizing the collected data. Cleaning removes noise, outliers, and duplicate data; transformation unifies data from different formats into a system-recognizable format; and standardization maps the data to specific ranges for easier subsequent processing.

[0047] For electricity consumption data, the quartile method is used to identify outliers. Specifically, the quartiles of the data are first calculated. It will exceed or below Data is considered outliers and processed accordingly; for duplicate data, all fields such as timestamp and load type are compared, and if the data are completely identical, only one record is retained.

[0048] Format conversion is performed to address data format differences between different device manufacturers, converting all data into a standardized format recognizable by the system, such as JSON, to facilitate data storage, transmission, and processing. The standardization process employs a normalization formula. Electricity consumption data Processing is carried out, among which and These are the minimum and maximum values ​​in the set of electricity consumption data, respectively. This method maps the data to... This range facilitates subsequent data analysis and model building; for carbon emission data A similar normalization method is also used for processing.

[0049] During data preprocessing, historical data is combined with the moving average method to analyze the fluctuation trend of the current data in depth. Taking a commercial office building as an example, the system will calculate the moving average of the building's electricity consumption during the same period over the past week and use it as a reference benchmark. If the deviation of the currently collected electricity consumption data from this benchmark exceeds twice the historical standard deviation, the system will further check whether the data is a true outlier, thereby effectively identifying abnormal data fluctuations caused by equipment failure, large-scale events, etc.

[0050] After data preprocessing is complete, a quality assessment of the processed data is necessary. This assessment should be conducted using multiple dimensions, including data consistency, standardization, and completeness. Data consistency is primarily assessed by checking whether related data from different data sources match. For example, does the electricity consumption data recorded by smart meters match the carbon emission data calculated by carbon emission monitoring sensors based on electricity consumption? Normative checks verify whether data fields conform to established standard format requirements. Integrity assessment focuses on checking for missing or incomplete data after preprocessing. ≥ At that time, among which, the quality assessment threshold If the value is set to 0.85, the data preprocessing is considered qualified, and the process can proceed to step S3. If the value is not qualified, since the data preprocessing failure is likely due to fundamental problems with the original collected data, the system will return to step S1 to re-collect the data and strengthen the data verification rules during the re-collection, such as adding a device self-checking step, to ensure that the quality of the collected data meets the requirements.

[0051] S3. Data Feature Extraction:

[0052] The main objective of this step is to extract key features closely related to carbon trading and load characteristics from the preprocessed data. A sliding window method is used to extract time series features, with the sliding window size set to [value missing]. For example, will Set to 1 hour, or 60 minute data points. For electricity consumption sequences... In the Calculate the average value within each window. It also calculates statistics such as the maximum, minimum, and standard deviation within the window, which can intuitively reflect the fluctuation of the load during that time period. In addition, it uses the least squares method to perform trend analysis on carbon emission data, extracting the changing trend of carbon emissions and obtaining characteristic parameters such as trend slope and intercept, thereby understanding the changing pattern of carbon emissions over time.

[0053] Based on historical data feature sets, an Autoregressive Integrated Moving Average (ARIMA) model is used to construct a feature prediction model. Taking the daily peak electricity consumption characteristics of a factory as an example, the system trains the ARIMA model based on the factory's daily peak electricity consumption data over the past 30 days. Then, the trained model is used to predict the peak electricity consumption for the current day, thereby obtaining the prediction feature set. Next, the difference between the actual extracted features and the predicted features is calculated. ,in, To actually extract the feature set, is the length of the feature set.

[0054] After feature extraction, the validity of the features needs to be rigorously assessed. A feature validity threshold should be set. The predicted difference threshold is 0.7. The correlation coefficient is 0.15, calculated by Pearson correlation coefficient between characteristics and carbon trading and load characteristics. And, combined with the degree of difference from the predicted features, a comprehensive evaluation is performed. ≥ and ≤ If the extracted features are deemed valid, the process proceeds to step S4. If the conditions are not met, it indicates that the data preprocessing may not have fully explored the potential value of the data, resulting in poor feature extraction. In this case, the system will return to step S2 to re-perform data preprocessing and attempt to adjust the preprocessing parameters, such as changing the threshold range for outlier handling, in order to obtain more effective data features.

[0055] S4. Multi-type load data aggregation:

[0056] Based on the load type and extracted features, the K-means clustering algorithm is used to aggregate the data, grouping load data with similar characteristics into the same cluster, thereby achieving effective data classification and integration. The number of clusters is then determined. At the same time, the Elbow Method and historical clustering results are used in combination. First, from... Starting with =2, calculate different... The sum of squared errors within a cluster (SSE) under the given value is plotted as follows: For changing curves, select the inflection point corresponding to the curve. The value is used as the initial number of clusters; at the same time, the optimal number of clusters in similar scenarios based on historical data is used to refine the initially determined number. The values ​​are adjusted by calculating the distance between the data points and the centers of each cluster. ,in, For the first The first data point 1 eigenvalue, For the first The first data point Each feature value is used to assign data points to the nearest cluster.

[0057] During the data aggregation process, the Jaccard similarity coefficient between the current clustering result and the historical clustering results is calculated by referring to the clustering results of historical data. ,in This is the current clustering result. The results are historical clustering findings, which are used to evaluate the stability and rationality of the current clustering.

[0058] After data aggregation is completed, the similarity within the cluster is used to determine the data. Inter-cluster discrimination Similarity to historical clustering The reasonableness of the aggregation results is evaluated from three aspects. A reasonableness threshold for aggregation is set. The value is 0.75, which is evaluated by calculating the average distance between data points within a cluster. The smaller the average distance, the better. A larger value indicates a higher degree of similarity among data within a cluster; this is assessed by calculating the average distance between different cluster centers. The greater the average distance, the The larger the value, the better the distinction between clusters. When ≥ , ≥ and ≥ At that time, the similarity threshold for historical clustering If the value is set to 0.7, the aggregation result is considered reasonable, and the process proceeds to step S5. If the condition is not met, the system will return to step S4 to readjust the clustering parameters, such as changing the initial cluster center selection method or adjusting the number of clustering iterations. If the aggregation result still does not meet the standard after multiple adjustments, a maximum of three re-aggregations will be performed. In this case, manual intervention will be prompted, and professionals will conduct in-depth analysis of the data characteristics to determine if there are any unreasonable aspects and take appropriate action.

[0059] S5. Constructing a carbon trading model:

[0060] Based on the aggregated data, a carbon trading model is constructed, which considers multiple factors such as carbon emissions, electricity consumption, and market carbon prices for each load cluster. A carbon trading cost function is established. ,in, For the first The carbon price corresponding to each load cluster needs to be determined by comprehensively analyzing information such as publicly available data from the carbon trading market, relevant policy guidance, and industry forecast reports. At the same time, the carbon price difference corresponding to different time periods (such as peak-valley-flat electricity price periods) must also be considered. For the first The carbon emissions of each load cluster. In addition, the model also needs to include electricity demand constraints and carbon quota constraints for the loads. The electricity demand constraints are mainly to ensure the normal operation of industrial production and other activities, and to ensure that the minimum electricity demand is met; the carbon quota constraints are to reasonably decompose the annual carbon quotas allocated by the government, so as to achieve effective control of carbon emissions.

[0061] To make the carbon price parameters in the model more reasonable and accurate, historical carbon trading data is used to conduct an in-depth analysis of the fluctuation patterns and trends of market carbon prices, employing seasonal differences. On a quarterly basis, carbon price data is decomposed into trend, seasonal, and stochastic components. By analyzing and predicting these components, the range of carbon price fluctuations over a future period is determined, thus providing a more scientific basis for setting the carbon price parameters in the model. For example, if it is predicted that carbon prices will rise in the next quarter due to policy tightening, the carbon price parameters will be adjusted upwards in advance in the model to optimize subsequent carbon trading strategies.

[0062] Once the model is built, its feasibility needs to be comprehensively assessed. Feasibility assessment indicators should be set. The model is comprehensively evaluated from multiple aspects, including whether it meets various constraints and conforms to the logic of actual market transactions. First, a preliminary solution is obtained using a mathematical programming solver (such as Gurobi or CPLEX) to check for feasible solutions. Second, experts in the carbon trading field are invited to review the model, evaluating its rationality from the perspectives of practical operation and market principles. ≥ At that time, feasibility threshold If the value is set to 0.8, the model is considered feasible, and the process proceeds to step S6. If it is not feasible, the system will return to step S5 to readjust the model parameters, such as modifying the range of constraints or adjusting the weights of the cost function. If the model is still not feasible after multiple adjustments, it is necessary to further check whether there are any deviations in the data aggregation results, and if necessary, return to step S4 to re-aggregate the data.

[0063] S6. Optimize carbon trading strategies:

[0064] A genetic algorithm is used to solve the carbon trading model to obtain the optimal carbon trading strategy. Let the carbon trading strategy be... ,in, This represents a decision variable in the strategy, such as the adjustment of carbon emissions for a load cluster over a certain time period. It is achieved through the fitness function. ,in, To adopt a strategy The fitness value is used to evaluate the effectiveness of a strategy by assessing the carbon trading cost at the time of trading; a higher fitness value indicates a better strategy. In the operation of the genetic algorithm, roulette wheel selection is used for selection, single-point crossover for crossover, and random mutation for mutation. These operations continuously optimize the carbon trading strategy.

[0065] To improve the efficiency and accuracy of strategy optimization, a historical strategy-result database is established, combining historical carbon trading strategies and their execution results. Through text mining and machine learning algorithms (such as cosine similarity calculation), the database is used to identify historical scenarios and corresponding optimal strategies similar to the current market environment (including carbon price fluctuation trends, policy changes, industry dynamics, etc.). These scenarios serve as important references for the initial population or optimization direction of the genetic algorithm. For example, if the current market carbon price is in a phase of rapid increase, the system will select successful trading strategies from the historical database during periods of rising carbon prices, extract key decision variables, and use these as part of the initial population of the genetic algorithm. This accelerates the algorithm's convergence speed and increases the likelihood of finding the optimal strategy.

[0066] Once the strategy optimization is complete, the effectiveness of the optimization results needs to be verified. A validity verification threshold should be set. The optimized strategy, with a value of 0.9, is applied to a simulated carbon trading scenario. This scenario includes various possible situations, such as sudden fluctuations in carbon prices and temporary policy adjustments. A large amount of simulated data is generated using the Monte Carlo simulation method to calculate the ratio of actual to expected returns. .when ≥ If the optimization result is deemed valid, the process proceeds to step S7. If the conditions are not met, the system returns to step S6 to re-optimize the strategy, adjusting the parameters of the genetic algorithm (such as crossover probability and mutation probability) or changing the search space of the strategy. If the result is still unsatisfactory after multiple optimizations (maximum of 3 optimizations), the carbon trading model needs to be re-examined, and the system returns to step S5 for model adjustment.

[0067] S7. Execute carbon trading strategies and provide data feedback:

[0068] Carbon trading operations are executed according to the optimized carbon trading strategy. During execution, blockchain technology is used to ensure the transparency, immutability, and security of the transactions. Various data points are monitored in real time, including carbon trading volume, trading price, and actual carbon emissions, and these data are compared and analyzed with expected data. A real-time data monitoring dashboard is established to display key data indicators and their trends in intuitive charts. When the deviation between actual and expected data exceeds a set threshold (5%), an early warning mechanism is immediately triggered, prompting traders to pay close attention and take appropriate countermeasures.

[0069] After the transaction is completed, all data from the entire transaction process will be stored in a historical database. Comparative and attribution analyses will be used to compare the actual transaction results with similar historical cases, analyzing the reasons for differences in transaction costs, emission reduction effects, etc., and generating an experience summary report. Evaluation indicators will be set from dimensions such as transaction costs, emission reduction effects, and compliance. Set evaluation thresholds It is 0.85, when ≥ If the transaction is successful, the process ends; if it fails to meet the criteria, the system returns to step S6 to re-optimize the trading strategy and adjusts the historical data summary method and prediction model parameters based on the transaction results.

[0070] Example 2:

[0071] This embodiment, based on Embodiment 1, introduces a multi-type load carbon trading and data aggregation system. The system includes a data acquisition module, a data preprocessing module, a data feature extraction module, a data aggregation module, a carbon trading model construction module, a carbon trading strategy optimization module, and a transaction execution and feedback module.

[0072] I. Data Acquisition Module

[0073] This module enables real-time collection of various load data, including industrial, commercial, and residential loads, by deploying terminal devices and comparing and verifying them with historical data to ensure the integrity and accuracy of the original data. It includes a data acquisition unit and a historical data comparison unit.

[0074] 1. Data Acquisition Unit

[0075] By using devices such as smart meters and carbon emission monitoring sensors, combined with 5G and NB-IoT communication technologies, electricity consumption data and carbon emission data from industrial, commercial, and residential loads are collected in real time at a sampling frequency of minutes, and processed into a structured current dataset. As in the set in Example 1 After data collection, the data integrity ratio is calculated. With accuracy ratio Set threshold =0.9、 =0.95, if ≥ and ≥ If the data is qualified, it will be entered into the historical data comparison unit; otherwise, it will be re-collected, repeating a maximum of 3 times. If it still fails to meet the standard, an equipment maintenance signal will be output.

[0076] 2. Historical Data Comparison Unit

[0077] Extract historical data of the same type and period from historical databases. The similarity between the current data and historical data is calculated using the Euclidean distance formula as shown in Example 1, and then processed into a similarity evaluation value. Set a similarity threshold. =0.8, if ≤ If the data is valid, it will enter the data preprocessing module normally; if it exceeds the threshold, an anomaly report will be generated and manual verification will be triggered. If the data is valid after verification, it will be forced to pass; otherwise, it will be returned to the data acquisition unit for re-acquisition.

[0078] II. Data Preprocessing Module

[0079] This module cleans, converts, and standardizes the collected data, identifies anomalies by combining historical data fluctuation patterns, and improves data quality. It includes a data cleaning and conversion unit and a data standardization unit.

[0080] 1. Data Cleaning and Transformation Unit

[0081] Noise and outliers in the data are removed using the quartile method, duplicate data is removed by full-field comparison, and data from different manufacturers' devices are uniformly converted to the JSON standard format. After cleaning and conversion, data consistency is checked (such as the logical correlation between electricity consumption and carbon emissions). If the check passes, the data enters the data standardization unit; otherwise, it is cleaned again, repeated a maximum of 2 times. If it still fails to meet the requirements, it is fed back to the data acquisition module to check the source data.

[0082] 2. Data Standardization Unit

[0083] Using the normalization formula as shown in Example 1, the cleaned data is mapped to... Interval, and calculate the moving average of historical data. with standard deviation Determine if the current data exceeds 2 The scope is processed into a standardized data sequence, and fluctuation trends are assessed. Quality assessment indicators are set. threshold =0.85, if ≥ If the preprocessing is successful, the data will proceed to the data feature extraction module; otherwise, the data will return to the data cleaning and transformation unit, where parameters will be adjusted and the data will be reprocessed.

[0084] III. Data Feature Extraction Module

[0085] This module extracts load characteristics related to carbon trading from preprocessed data and verifies them by combining them with historical feature model predictions. It includes a feature extraction unit and a feature prediction verification unit.

[0086] 1. Feature Extraction Unit

[0087] Using the sliding window method as described in Example 1 (window size) (60 minutes) Extract time series features, calculate statistics such as the mean and maximum value within the window, and use the least squares method to analyze carbon emission trends, processing the data into a feature set containing statistics and trend parameters. Calculate the correlation coefficient between the characteristics and carbon trading. Set threshold =0.7, if ≥ If the condition is met, proceed to the feature prediction and verification unit; otherwise, return to this unit, adjust the window size, or re-extract the feature using the new analysis method.

[0088] 2. Feature Prediction Validation Unit

[0089] An ARIMA model is trained based on historical feature sets to predict current features. The difference is then calculated using the difference formula shown in Example 1, and processed into a feature evaluation value. A prediction difference threshold is set. =0.15, if ≥ and ≤ If the data is valid, the feature is entered into the data aggregation module; otherwise, it is returned to the data preprocessing module, where the standardized parameters are optimized and the data is reprocessed.

[0090] IV. Data Aggregation Module

[0091] This module uses the K-means algorithm to cluster and aggregate multiple types of load data based on load characteristics and historical clustering results, and evaluates the rationality of the aggregation. It includes a clustering algorithm unit and a clustering verification unit.

[0092] 1. Clustering Algorithm Unit

[0093] Determining the number of clusters by combining the elbow rule and the historical best The data points are divided into optimal clusters by calculating the distance between the data points and the cluster centers using the K-means distance formula as shown in Example 1, and then processed into an aggregated load cluster set. Calculate intra-cluster similarity Inter-cluster discrimination Set threshold =0.75, if ≥ and ≥ Then proceed to the clustering verification unit; otherwise, adjust... Use the initial center or value to re-cluster, repeating up to 3 times.

[0094] 2. Clustering Validation Unit

[0095] The similarity between the current cluster and historical clusters is calculated using the Jaccard coefficient formula as shown in Example 1, and processed into a clustering rationality assessment result. A threshold is set. =0.7, if similarity ≥ If the aggregation result is valid, it will be entered into the carbon trading model construction module; otherwise, it will be returned to the clustering algorithm unit and re-aggregated with reference to historical parameters.

[0096] V. Carbon Trading Model Construction Module

[0097] This module constructs a carbon trading model based on aggregated data, including cost functions and constraints, and optimizes model parameters by combining historical carbon price trends. It includes a model construction unit and a model optimization unit.

[0098] 1. Model building unit

[0099] Establish a carbon trading cost function as shown in Example 1, incorporating electricity demand and carbon quota constraints, and process it into a mathematical model containing an objective function and constraints. Verify the model's feasibility using a solver. If a feasible solution exists, proceed to the model optimization unit; otherwise, adjust the constraint range and reconstruct the model, repeating this process a maximum of two times.

[0100] 2. Model Optimization Unit

[0101] By utilizing seasonal decomposition methods based on historical carbon price data (such as STL decomposition in Example 1), the range of carbon price fluctuations is predicted, and model parameters are optimized to create a parameter-optimized carbon trading model. Experts are invited to review the model's rationality and set feasibility indicators. If the value is ≥0.8, proceed to the strategy optimization module; otherwise, return to the model building unit to adjust the cost function weights.

[0102] VI. Carbon Trading Strategy Optimization Module

[0103] This module solves the carbon trading model based on a genetic algorithm, optimizes the trading strategy by combining a historical strategy library, and verifies the feasibility of the strategy through simulation. It includes a genetic algorithm unit and a strategy verification unit.

[0104] 1. Genetic Algorithm Unit

[0105] Define strategy encoding The fitness function, as shown in Example 1, is used to evaluate the effectiveness of the strategy. The strategy is then optimized by combining it with strategies from similar historical scenarios, resulting in an optimized carbon trading strategy. Calculate the policy fitness. If it reaches 90% of the historical best fitness, proceed to the policy validation unit; otherwise, adjust the crossover / mutation probabilities and re-optimize, with a maximum of 50 iterations.

[0106] 2. Strategy Verification Unit

[0107] Multi-scenario data is generated through Monte Carlo simulation. The effectiveness of the strategy is verified using the payoff ratio formula as shown in Example 1, and then processed into a strategy effectiveness evaluation value. Set a threshold. =0.9, if ≥ If the strategy is successful, it will be entered into the transaction execution module; otherwise, it will be returned to the genetic algorithm unit, and the search space will be expanded before re-optimization.

[0108] VII. Transaction Execution and Feedback Module

[0109] This module executes carbon trading based on optimization strategies, monitors trading data in real time, compares historical results to generate feedback, and updates the historical database and model parameters. It includes a trading execution unit and a result evaluation and feedback unit.

[0110] 1. Transaction Execution Unit

[0111] Utilizing blockchain technology to execute transactions, the system monitors carbon trading volume, prices, and other data in real time, processing them into transaction execution records and real-time monitoring data. A 5% deviation warning threshold is set; if the monitored data deviation exceeds the threshold, an alert is triggered, prompting manual intervention; otherwise, the transaction is executed normally until completion, at which point it enters the result evaluation unit.

[0112] 2. Results Evaluation and Feedback Unit

[0113] By comparing actual transaction results with historical cases, and using evaluation indicators such as those in Example 1... Analyze the reasons for the discrepancies and process them into transaction evaluation reports and model update parameters. Set thresholds. =0.85, if ≥ If the transaction is successful, the process ends; otherwise, the system returns to the strategy optimization module, where the strategy is re-optimized based on the feedback results, and the historical database is updated.

[0114] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for multi-type load carbon trading and data aggregation, characterized in that, Includes the following steps: S1. Multi-type load data collection: Deploy smart meters and carbon emission monitoring sensors at load terminals, and collect current data including timestamps, load types, electricity consumption and carbon emissions through 5G and NB-IoT technologies. Simultaneously extract historical data, calculate the Euclidean distance between current data and historical data to assess similarity, and judge the completeness and accuracy of the collected data based on similarity. S2. Data preprocessing: Clean the collected data to remove outliers and duplicate data, convert the cleaned data to JSON format, standardize the data using a normalization formula, and analyze the fluctuation trend by combining the moving average benchmark of historical data. S3. Feature extraction based on preprocessed data: Time series features are extracted using the sliding window method, carbon emission trends are analyzed using the least squares method, and the ARIMA model is trained based on historical feature sets to predict current features and assess the correlation between features and carbon trading. S4. Aggregate multi-type load data based on the data after feature extraction: Use the K-means clustering algorithm, combine the elbow rule and historical clustering results to determine the number of clusters, calculate the distance between data points and cluster centers and perform clustering, and evaluate the rationality of the current clustering by referring to historical clustering results; S5. Construct a carbon trading model based on the aggregation results: Establish a cost function that includes carbon price and carbon emissions based on the aggregated data, add electricity demand and carbon quota constraints, optimize the model parameters using the seasonal decomposition of historical carbon prices, and construct a carbon trading model based on the optimized model parameters. S6. Optimize carbon trading strategies based on carbon trading models: Solve carbon trading models using genetic algorithms, evaluate the quality of strategies through fitness functions, optimize strategies by combining historical strategy libraries, and verify the effectiveness of strategies using Monte Carlo simulations. S7. Execute carbon trading strategies and provide data feedback: Utilize blockchain technology to execute transactions, monitor data in real time and compare it with expectations, store the data in the historical database after the transaction is completed, compare the results with historical cases to evaluate the results and update the model parameters.

2. The multi-type load carbon trading and data aggregation method as described in claim 1, characterized in that, In step S1, when judging the completeness and accuracy of the collected data, the completeness threshold is... Set to 0.9, accuracy threshold Set to 0.95, if the integrity ratio ≥ And accuracy ratio ≥ If the data is found to be qualified, proceed to step S2; if the conditions are not met, the system will automatically re-execute step S1. If the data still does not meet the standards after three re-collections, the data collection operation will be stopped, and technical personnel will be notified in a timely manner to inspect the relevant equipment and investigate the data source.

3. The multi-type load carbon trading and data aggregation method as described in claim 1, characterized in that, In step S2, the quartile method is used. Identify outliers in electricity consumption data; use normalization formulas Electricity consumption data Processing is carried out, among which, and These are the minimum and maximum values ​​in this set of electricity consumption data. The data is standardized; and quality assessment indicators are calculated for the preprocessed data. Set quality assessment thresholds =0.85, when ≥ If the data preprocessing is deemed satisfactory, the system can proceed to step S3; otherwise, it will return to step S1 to re-collect data and strengthen the data verification rules during the re-collection process.

4. The multi-type load carbon trading and data aggregation method as described in claim 1, characterized in that, In step S3, the sliding window size is set to 1 hour, i.e., 60 minutes of data points are used to extract features, and the average value within the window is calculated. Predict features and calculate dissimilarity using the ARIMA model. ,in, To actually extract the feature set, Set the feature set length; set the feature validity threshold. The predicted difference threshold is 0.

7. When it is 0.15, ≥ and ≤ hour, If the Pearson correlation coefficient is obtained, proceed to step S4; if the conditions are not met, the system will return to step S2 to re-preprocess the data and attempt to adjust the preprocessing parameters.

5. The method for multi-type load carbon trading and data aggregation as described in claim 1, characterized in that, In step S4, the K-means clustering algorithm is used to calculate the distance between data points and the centers of each cluster. ,in, For the first The first data point 1 eigenvalue, For the first The first data point One feature value; similarity coefficient ,in This is the current clustering result. For historical clustering results; set a threshold for the reasonableness of aggregation. The similarity threshold for historical clustering is 0.

75. Set to 0.7, when the intra-cluster similarity ≥ Inter-cluster discrimination ≥ and ≥ If the aggregation result is deemed reasonable, the system can proceed to step S5; otherwise, the system will return to step S4 to readjust the clustering parameters.

6. The method for multi-type load carbon trading and data aggregation as described in claim 1, characterized in that, In step S5, a carbon trading cost function is established. ,in, For the first The carbon valence corresponding to each load cluster For the first Carbon emissions of each load cluster; setting feasibility assessment indicators. Feasibility threshold Set to 0.8, when ≥ If the model is deemed feasible, proceed to step S6; if not, the system will return to step S5 to readjust the model parameters. If the model remains infeasible after multiple adjustments, it is necessary to further check whether there are any deviations in the data aggregation results, and if necessary, return to step S4 to re-aggregate the data.

7. The method for multi-type load carbon trading and data aggregation as described in claim 1, characterized in that, In step S6, when using Monte Carlo simulation to verify the effectiveness of the strategy, a validity verification threshold is set. Given a value of 0.9, calculate the ratio of actual return to expected return. ,when ≥ If the optimization result is deemed valid, proceed to step S7. If the conditions are not met, the system will return to step S6 to re-optimize the strategy. If the results are still unsatisfactory after multiple optimizations (maximum of 3 optimizations), the carbon trading model needs to be re-examined, and the system will return to step S5 to adjust the model.

8. The method for multi-type load carbon trading and data aggregation as described in claim 1, characterized in that, In step S7, when comparing the evaluation results with historical cases, evaluation indicators are set. Set evaluation threshold It is 0.85, when ≥ If the transaction is successful, the process ends; if it fails to meet the criteria, the system returns to step S6 to re-optimize the trading strategy.

9. A multi-type load carbon trading and data aggregation system implementing the multi-type load carbon trading and data aggregation method according to any one of claims 1 to 8, characterized in that, The system includes: The data acquisition module is used to collect real-time load data of various types, including industrial, commercial, and residential loads, by deploying terminal equipment, and to compare and verify the data with historical data. It includes a data acquisition unit and a historical data comparison unit. The data preprocessing module is used to clean, convert, and standardize the collected data, and to identify anomalies by combining historical data fluctuation patterns. It includes a data cleaning and conversion unit and a data standardization unit. The data feature extraction module is used to extract load features related to carbon trading from preprocessed data and verify them by combining historical feature model predictions. It includes a feature extraction unit and a feature prediction verification unit. The data aggregation module is used to cluster and aggregate multiple types of load data based on load characteristics and historical clustering results using the K-means algorithm, and to evaluate the rationality of the aggregation. It includes a clustering algorithm unit and a clustering verification unit. The carbon trading model building module is used to build a carbon trading model containing cost functions and constraints based on aggregated data, and to optimize model parameters by combining historical carbon price trends. It includes a model building unit and a model optimization unit. The carbon trading strategy optimization module is used to solve the carbon trading model based on the genetic algorithm, optimize the trading strategy by combining the historical strategy library, and verify the feasibility of the strategy through simulation. It includes a genetic algorithm unit and a strategy verification unit. The transaction execution and feedback module is used to execute carbon transactions based on optimized strategies, monitor transaction data in real time, compare historical results to generate feedback, and update historical databases and model parameters. It includes a transaction execution unit and a result evaluation and feedback unit.

Images