Carbon accounting method and device based on artificial intelligence and medium

Abstract

The invention discloses a carbon accounting method and device based on artificial intelligence and a medium, and the method comprises the steps: collecting original data through multi-data source collection, and carrying out the data cleaning and standardization processing of the original data, and obtaining standardized data; extracting carbon emission related features from the standardized data, and screening the carbon emission related features through a correlation analysis or information gain method to generate a feature combination; and training an artificial intelligence model by using the feature combination, inputting the preprocessed data of the to-be-checked object into the trained model, and outputting a carbon emission estimated value. According to the method, the dependence of a traditional method on a preset formula is broken through, the capturing capability of the model on complex nonlinear correlation is remarkably enhanced, the accounting error rate is reduced, and the prediction accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a carbon accounting method and device based on artificial intelligence and a medium. BACKGROUND

[0002] With the acceleration of global carbon emission reduction, carbon accounting has become a key link for enterprises and regions to implement environmental management. Traditional carbon accounting methods generally rely on manual collection of scattered data such as energy consumption and production activities, and then perform static calculation through preset emission coefficients. Such methods require a large amount of manpower to check multi-source heterogeneous data, and are prone to calculation errors due to data omissions, format conflicts or record deviations; especially when dealing with dynamic factors such as geographical climate differences and industry production characteristics, it is difficult for fixed calculation formulas to accurately quantify the actual impact of environmental variables on carbon emissions, resulting in accounting results deviating from the real scene.

[0003] In the prior art, there is a significant efficiency bottleneck in the data integration process. Due to the differences in data structure between energy monitoring sensors, production management systems and public databases, traditional methods require manual execution of repeated cleaning and format conversion, which not only consumes time but also cannot guarantee data integrity. More importantly, carbon emissions and production intensity, climate conditions and other factors have complex nonlinear correlations, and existing models only handle such correlations through simple linear formulas, which cannot automatically identify key features (such as unit production energy consumption ratio and air temperature correction coefficient), resulting in insufficient adaptability for accounting of distributed energy systems or small and micro enterprises.

[0004] Current carbon accounting systems have serious limitations in dynamic updating and result application. On the one hand, traditional static models cannot continuously optimize parameters based on new data, and need to rebuild the calculation framework when facing energy structure upgrading or policy adjustment; on the other hand, the accounting results are mostly presented in static reports, lacking spatial dimension positioning ability of high-emission areas and time dimension trend inflection point warning mechanism, making it difficult to provide real-time visual support for emission reduction decision-making. SUMMARY

[0005] The embodiments of the present application provide a carbon accounting method, device and medium based on artificial intelligence to solve the above technical problems.

[0006] In one aspect, the embodiments of the present application provide a carbon accounting method based on artificial intelligence, comprising:

[0007] Collecting raw data from multiple data sources, and performing data cleaning and standardization processing on the raw data to obtain standardized data;

[0008] Extracting carbon emission related features from the standardized data, and filtering the carbon emission related features through a correlation analysis method to generate a feature combination;

[0009] The artificial intelligence model is trained by using the feature combination, and preprocessed data of a to-be-accounted object is input into the trained model to output a carbon emission estimation value.

[0010] In an implementation of the present application, raw data is collected through multiple data sources, and the raw data is subjected to data cleaning and standardization processing to obtain standardized data, specifically including:

[0011] Energy consumption data is collected through sensors, and production activity data is obtained from an enterprise production management system; the energy consumption data includes real-time usage of electricity, coal and natural gas, and the production activity data includes equipment operation parameters and production output;

[0012] Geographical information data and industry category data are extracted from a public database; the geographical information data includes climate conditions and altitude;

[0013] The raw data is subjected to data cleaning to remove duplicate data, error data and missing data in the raw data, and for missing data, the mean value of adjacent time points is used for interpolation filling;

[0014] The cleaned multi-source data is subjected to normalization processing, and the normalized multi-source data is subjected to time format unification to obtain standardized data.

[0015] In an implementation of the present application, carbon emission related features are extracted from the standardized data, specifically including:

[0016] Based on the energy consumption data, the proportion of each type of energy consumption in total energy consumption is calculated, and the proportion is taken as an energy structure feature;

[0017] Based on the production activity data, the ratio of output to equipment operation time per unit time is calculated, and the ratio is taken as a production efficiency feature;

[0018] Based on the geographical information data, the influence degree of climate conditions on energy consumption is calculated to generate a geographical environment influence coefficient feature.

[0019] In an implementation of the present application, based on the geographical information data, the influence degree of climate conditions on energy consumption is calculated to generate a geographical environment influence coefficient feature, specifically including:

[0020] The climate condition data and energy consumption data of the location of the enterprise are associated; the climate conditions include temperature, humidity or altitude;

[0021] A mapping relationship between climate conditions and energy consumption change rate is constructed, and based on the mapping relationship, a correction coefficient of geographical environment on energy consumption is calculated.

[0022] In an implementation of the present application, the pre-processed data of the object to be calculated is input into the trained model, and the carbon emission estimation value is output. After that, the method further comprises:

[0023] The carbon emission estimation value is grouped by time series, the emission difference of each group is calculated, and the difference is converted into a gradient color depth by calling a chart engine;

[0024] The preset carbon emission threshold is matched based on the industry category data, and the coordinates of the accounting objects exceeding the threshold are superimposed on the digital map to generate a heat map layer;

[0025] The extreme points of the carbon emission change rate in the adjacent time period are identified, warning marks are inserted in the preset display mode, and abnormal production activity data records are associated.

[0026] In an implementation of the present application, the pre-processed data of the object to be calculated is input into the trained model, and the carbon emission estimation value is output. Specifically, it comprises:

[0027] A carbon emission digital report is generated according to the time dimension, and the emission estimation values corresponding to different accounting objects are marked in the carbon emission digital report;

[0028] The emission estimation value is associated with the corresponding industry category data, and the high-carbon emission hotspot area is marked in the carbon accounting area;

[0029] The carbon emission change trend is displayed through a preset display mode; the preset display mode includes a column chart or a line chart.

[0030] In an implementation of the present application, it further comprises:

[0031] When the amount of newly added multi-source data reaches a preset period, the newly added multi-source data is cleaned and standardized to generate an incremental data set;

[0032] Based on the incremental data set, the correlation weight of the feature and the carbon emission is recalculated, and the features with a weight lower than a preset threshold are excluded;

[0033] An incremental learning algorithm is used to load the historical training model, and the weight matrix is updated based on the incremental data set. After the loss function converges, the updated model version is determined.

[0034] In an implementation of the present application, the artificial intelligence model is trained by using the feature combination, specifically comprising:

[0035] The feature combination is divided into a training set, a validation set and a test set. Based on the training set, the error between the output value and the true carbon emission value is minimized by adjusting the model weight and the threshold;

[0036] Based on the evaluation set, the model performance is evaluated, and in the case of continuous decline in model performance, an early stopping mechanism is triggered, and the model hyperparameters are optimized.

[0037] During the model training process, the accuracy, recall rate and mean square error of the model are calculated through the test set to verify the generalization ability of the model.

[0038] In another aspect, the embodiments of the present application also provide a carbon accounting device based on artificial intelligence, which comprises:

[0039] at least one processor;

[0040] and a memory in communication connection with the at least one processor;

[0041] wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above-mentioned carbon accounting method based on artificial intelligence.

[0042] In another aspect, the embodiments of the present application also provide a non-volatile computer storage medium storing computer executable instructions, which, when executed, implement the above-mentioned carbon accounting method based on artificial intelligence.

[0043] The embodiments of the present application provide a carbon accounting method, device and medium based on artificial intelligence, which at least have the following beneficial effects:

[0044] Through automatic collection and standardized processing, multi-source heterogeneous data such as sensors, enterprise systems and public databases are directly connected, eliminating omissions and format conflicts caused by manual collection; dynamic cleaning mechanism is adopted to automatically exclude duplicate or incorrect data, and missing values are intelligently filled, improving data integration efficiency, while ensuring the integrity and reliability of input data; based on the feature dynamic extraction and screening mechanism, the energy structure, production efficiency and geographical environmental impact coefficient are automatically quantified from the standardized data, breaking through the dependence on preset formulas in traditional methods; through correlation weight analysis and screening of key feature combinations, the model's ability to capture complex nonlinear correlations is significantly enhanced, reducing the error rate of accounting; using the training-validation-test closed-loop architecture, the weight threshold is dynamically adjusted during training and the early stopping mechanism is triggered, effectively preventing overfitting; combined with incremental learning driven model updating, continuously absorbing new data to optimize feature weights and parameter matrices, making the model adapt to dynamic scenarios such as energy structure upgrading and policy adjustment, improving the prediction accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0045] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0046] Figure 1 A schematic diagram of a process flow of an artificial intelligence-based carbon accounting method provided in an embodiment of the present application;

[0047] Figure 2 A schematic diagram of the internal structure of an artificial intelligence-based carbon accounting device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0049] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0050] Figure 1 A flowchart of an artificial intelligence-based carbon accounting method provided in an embodiment of the present application.

[0051] The analysis method involved in the embodiments of the present application can be implemented by a terminal device or a server, and the present application does not impose any special restrictions on this. For ease of understanding and description, the following embodiments are described in detail using a server as an example.

[0052] It should be noted that the server can be a single device or a system composed of multiple devices, that is, a distributed server, and this application does not make any specific restrictions on this.

[0053] like Figure 1 As shown, an artificial intelligence-based carbon accounting method provided in an embodiment of the present application includes:

[0054] Step 101: collect raw data from multiple data sources, and perform data cleaning and standardization on the raw data to obtain standardized data.

[0055] In this embodiment, the power consumption pulse signal is captured in real time by the current sensor deployed in the production equipment, while the equipment running status code and production counter reading are extracted from the work order database of the Enterprise Resource Planning (ERP) system, thereby constructing a dynamic monitoring network of energy consumption and production activities. For example, the natural gas consumption is quantified by the frequency signal generated by the gas meter electromechanical conversion module, and the equipment running time is obtained by jointly analyzing the internal clock of the controller and the start-stop flag bit. It should be noted that the geographic information data is obtained by calling the latitude and longitude coordinate query interface of the public API of the meteorological bureau, and the industry category data is automatically matched by associating the industry coding library of the National Bureau of Statistics.

[0056] Specifically, three layers of filtering are performed on the original data. The deduplication mechanism detects and deletes duplicate uploaded data packets based on timestamp hash values, such as data at the same time point caused by sensor network retransmission. The error correction engine identifies abnormal values by device threshold, such as current peaks exceeding the rated load of the transformer, and replaces them with the moving average of the previous and subsequent valid data. The missing data filling module uses linear interpolation of adjacent two sampling points to calculate the supplementary value for data gaps caused by communication interruption. It can be understood that this cleaning mechanism ensures data integrity and removes duplicate, error and missing data.

[0057] It should be noted that the cleaned data enters the normalization processor: different dimension parameters such as power consumption value (unit: kilowatt-hour) and natural gas volume (unit: cubic meter) are mapped to the [0, 1] interval by the maximum-minimum scaling method. At the same time, all time data are converted to ISO 8601 standard format by the format converter, such as converting "2025 / 6 / 27 14:30" to "2025-06-27T14:30:00Z". For example, this process realizes the technical features of normalization processing and time format unification, providing consistent input for subsequent feature engineering.

[0058] Step 102, extracting carbon emission related features from the standardized data, and selecting carbon emission related features by correlation analysis or information gain method to generate feature combinations.

[0059] In this embodiment, the feature extraction process is realized by a multi-dimensional carbon emission factor quantization engine, which specifically includes dynamic generation mechanisms for three types of features: energy structure, production efficiency and geographical environment. It should be noted that this process is directly related to the construction method of the geographical environment influence coefficient, forming a key bridge from data to decision-making.

[0060] Specifically, the energy structure feature is generated based on the real-time collected standardized energy consumption data. The power consumption data is obtained by converting the pulse frequency signal of the smart meter, and the coal / natural gas consumption is obtained by the volume-mass correspondence table generated by the high-precision mass flow meter. The single energy consumption is divided by the total energy consumption to generate the energy proportion vector. For example, in the manufacturing industry scenario, the proportion of coal consumption in the total energy consumption (power + coal + natural gas) is calculated, which is used as the fossil energy dependence index. It can be understood that this feature directly reflects the progress of enterprise energy clean transformation. When the proportion continues to decrease, it indicates that the carbon emission reduction effect is significant, and the calculation of the proportion of various energy consumptions as the energy structure feature provides a basic energy portrait for the model.

[0061] It should be noted that the construction of the production efficiency feature deeply integrates the production activity data. The device running state code (such as 0-stop / 1-run) is grabbed from the manufacturing execution system (MES), and the yield counter reading is combined with the work order database to accurately calculate the effective production time. The yield reading is divided by the device running time (unit: hour) to generate the unit time production intensity. For example, in the automobile production line scenario, a welding robot produces 80 parts in 2 hours, and the production intensity is 40 pieces / hour. It can be understood that this feature reveals the unit time carbon emission density, and the emission surge phenomenon accompanied by high intensity production can be accurately captured by the model, which breaks through the limitation of the traditional calculation of the linear correlation between production capacity and emission.

[0062] In this embodiment, when dynamically modeling the geographical environmental influence coefficient, the GPS coordinates of the enterprise registration address are called, the temperature and humidity time series data of the 5-kilometer grid are extracted from the meteorological database, and the historical energy consumption curve of the location is associated. For example, the geographical coordinates of a chemical plant are associated with the synchronous fluctuation of the peak value of the summer daily average temperature and the power consumption of the cooling system.

[0063] It should be noted that the least square fitting algorithm is used to fit the sensitivity slope parameter β of the temperature change to the energy consumption with the temperature as the independent variable and the air conditioning system energy consumption as the dependent variable. The altitude is converted into the air pressure correction coefficient α by every hundred meters, and finally the geographical environment correction factor γ = α × β is generated. For example, a data center in the Yunnan-Guizhou Plateau (altitude 1800 meters) has a higher γ value than the same type of facility along the coast due to the decrease in heat dissipation efficiency caused by low air pressure.

[0064] It should be noted that the bidirectional correlation screen is used to calculate the Pearson correlation coefficient of each feature and the true value of carbon emission in the positive direction, and to retain the features with an absolute value greater than a preset threshold, such as the production efficiency feature which usually presents strong negative correlation; in the reverse direction, the variance inflation factor (VIF) is evaluated, and when there is multicollinearity among multiple features, such as altitude and atmospheric pressure, only the feature with the lowest VIF value is retained to ensure that the combination of input models is both concise and effective.

[0065] It can be understood that this coefficient makes the implicit influence of geographical climate explicit, enabling the model to automatically correct. The reasonable increase in energy consumption for winter heating in high-latitude areas, the additional emissions of refrigeration equipment in hot and humid regions, and the indirect emissions caused by equipment efficiency reduction in high-altitude areas. The core innovation of generating geographical environmental impact coefficient features solves the defect of ignoring regional differences in traditional methods.

[0066] Step 103, training an artificial intelligence model using the feature combination, and inputting the preprocessed data of the object to be calculated into the trained model to output the carbon emission estimation value.

[0067] In this embodiment, it should be noted that the feature combination is divided into a training set, a validation set, and a test set according to a dynamic proportion. In the training phase, the adaptive moment estimation (Adam) optimizer is used to adjust the neural network weights, and the back propagation algorithm is used to minimize the mean square error between the predicted carbon emission value and the true value. The validation monitors the loss function curve in real time, and triggers the early stopping mechanism when the validation loss increases for consecutive multiple training periods, and freezes the current optimal model parameters. Robustness testing injects random disturbance data into the test set, and calculates the model's anti-interference capability indicators such as accuracy fluctuation range. Exemplarily, this process realizes a technical closed loop of evaluating performance and optimizing hyperparameters through the validation set.

[0068] Specifically, for the carbon emission estimation value sequence, a monthly emission digital report is generated in the time dimension, and the emission is labeled according to the workshop / production line. The geographic information system (GIS) engine is called to superimpose the coordinates of emission points exceeding the industry threshold as a heat layer, such as a red radiation area in a petrochemical park, to realize heat layer generation. Trend warning identifies extreme points of emission change rate in adjacent periods, automatically associates device fault logs, and inserts warning icons in the line chart, such as triggering an alarm when a device anomaly causes a sharp increase in weekly comparison, to realize warning mark association.

[0069] It can be understood that the system is built-in incremental learning trigger. When the amount of new data reaches the quarterly threshold, the standardized pipeline processes the incremental data set. The feature weight reevaluation module calculates the importance score of the features in the new data, and eliminates features with low contribution, such as old energy coefficients that are no longer valid due to process upgrades. An elastic weight solidification algorithm is used to load the historical model infrastructure, only fine-tuning the top fully connected layer weights sensitive to new data, and switching to the new model version after the validation set converges.

[0070] In this embodiment, the model training mechanism is implemented through an adaptive learning framework, which includes three layers of dynamic optimization processes. It should be noted that this framework directly references hyperparameter optimization and early stopping mechanisms to form a closed-loop model evolution system.

[0071] Specifically, for a feature combination data set, the data is sorted by the data collection timestamp, and the earliest 60% of the period data is used as the training set, the middle 20% of the period data is used as the validation set, and the latest 20% of the period data is used as the test set. For example, the data from January to June 2024 of a manufacturing enterprise is used for training, the data from July to August is used for validation, and the data from September to October is used for testing. It can be understood that this design avoids future data leakage and realizes the technical basis of "dividing the training set, the validation set, and the test set" in claim 8. High volatility industries (such as power production) use rolling window division, and stable industries (such as food processing) use random stratified division. This mechanism ensures that the data set represents the actual scene distribution.

[0072] It should be noted that the training process performs weight adjustment logic. Specifically, an adaptive moment estimation (Adam) optimizer is used to initialize all weight matrices and bias vectors of the neural network, forward calculate the predicted carbon emission value, calculate the loss through the mean square error function, and update the weight matrix and bias vector along the gradient back propagation. For example: in the multi-process coupling scenario of a steel mill, the model increases the weight of the smelting process to 2.3 times that of the rolling process through 300 iterations.

[0073] It can be understood that when the validation loss enters a plateau, the learning rate is exponentially decayed, such as halving every 50 steps from an initial value of 0.01, and the sawtooth period is scheduled to periodically increase the learning rate to escape local optima.

[0074] In this embodiment, when performing enterprise carbon emission accounting, for a large manufacturing enterprise, energy consumption data such as electricity and natural gas is collected through energy sensors installed in its production workshops, office buildings, etc., production activity data such as yield and equipment running time is obtained from the enterprise's production management system, and geographic information data such as climate conditions and elevation of the enterprise's location is obtained from a public geographic information database. The collected data is cleaned to remove duplicate, incorrect and missing data, and then standardized, such as normalizing energy consumption data to the [0, 1] interval.

[0075] Features are extracted from the pre-processed data, such as the proportion of different energy consumption in total energy consumption as energy consumption features, the ratio of output to equipment running time per unit time as production activity intensity features, and the influence coefficient of annual average temperature of the enterprise location on energy consumption as geographical environment factors, etc. The most influential feature combination for enterprise carbon emission accounting is selected through correlation analysis.

[0076] An artificial neural network model is selected as the accounting model. The data set after feature engineering is divided into training set, validation set and test set according to the ratio of 7:2:1. The artificial neural network model is trained using the training set. In the training process, the weights and thresholds of the network are adjusted to make the model output as close as possible to the known true value of enterprise carbon emission. The true value of enterprise carbon emission is obtained by traditional accounting method. The model performance in the training process is evaluated and optimized using the validation set to prevent overfitting. Finally, the trained model is tested using the test set to calculate its accuracy, recall rate and other indicators to ensure that the model has good generalization ability.

[0077] The relevant data of the enterprise to be accounted for the period are pre-processed and feature-engineered and input into the trained artificial neural network model. The model calculates the estimated value of carbon emission of the enterprise in the period. The result is output in the form of a digital report to show the carbon emission of different workshops and different months of the enterprise, and visual charts such as bar charts can also be generated to compare the carbon emission differences of different workshops.

[0078] In this embodiment, when performing regional carbon emission accounting, for a specific region, multi-source data such as energy consumption data reported by enterprises in the region, power supply data obtained from power companies, and climate data obtained from meteorological departments are integrated. These data are cleaned and standardized to ensure the integrity and accuracy of the data.

[0079] Features are extracted such as the proportion of different industries in the region, the relationship between average temperature in the region and energy consumption change rate, etc. The most influential key feature combination for regional carbon emission accounting is selected by information gain and other methods.

[0080] A support vector machine model is selected as the accounting model. The data set after feature engineering is divided according to the same data set division method (such as 6:3:1). The support vector machine model is trained using the training set, the model parameters are optimized, the model performance is evaluated using the validation set, and the model performance is tested using the test set to ensure that the model has good generalization ability.

[0081] The relevant data of the region in the time period to be calculated is input into the trained support vector machine model, and the estimated value of carbon emission of the region in the time period is calculated. The result is output in the form of a line chart to show the change trend of carbon emission of the region in different years.

[0082] As can be seen from the above embodiments, the carbon accounting method based on artificial intelligence can effectively overcome the disadvantages of the traditional carbon accounting method, realize more accurate and efficient carbon accounting, and provide strong technical support for coping with climate change and formulating emission reduction strategies.

[0083] The above is the method embodiment of the present application. Based on the same inventive concept, the present application embodiment also provides an artificial intelligence-based carbon accounting device, the structure of which is shown in Figure 2 .

[0084] Figure 2 An internal structure diagram of an artificial intelligence-based carbon accounting device provided by the present application embodiment is shown in Figure 2 .

[0085] at least one processor;

[0086] and a memory in communication connection with the at least one processor;

[0087] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0088] Collect raw data through multiple data sources, and perform data cleaning and standardization processing on the raw data to obtain standardized data;

[0089] Extract carbon emission related features from the standardized data, and filter the carbon emission related features through correlation analysis or information gain method to generate a feature combination;

[0090] Train an artificial intelligence model using the feature combination, and input the preprocessed data of the object to be calculated into the trained model to output an estimated value of carbon emission.

[0091] The present application embodiment also provides a non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions can be executed to:

[0092] Collect raw data through multiple data sources, and perform data cleaning and standardization processing on the raw data to obtain standardized data;

[0093] Extract carbon emission related features from the standardized data, and filter the carbon emission related features through correlation analysis or information gain method to generate a feature combination;

[0094] The artificial intelligence model is trained by using feature combinations, and preprocessed data of a to-be-accounted object is input into the trained model to output a carbon emission estimation value.

[0095] The embodiments in the present application are described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, the device and medium embodiments are described simply because they are basically similar to the method embodiments, and the relevant parts can be referred to the part of the method embodiments.

[0096] The device and medium provided by the embodiments of the present application are one-to-one corresponding to the method, and therefore, the device and medium also have similar beneficial technical effects to the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device and medium will not be described here.

[0097] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0098] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device implemented in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for performing the function specified by one or more blocks.

[0099] These computer program instructions can also be stored in a computer-readable memory that can guide the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction means, which implements the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0100] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 Figure 1

[0101] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0102] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory, non-volatile memory, such as read-only memory (ROM), EPROM, and / or flash memory, etc. The memory is an example of computer readable media.

[0103] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0104] It should also be noted that the terms "comprising," "including," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements recited, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without limitation, an element preceded by "comprises a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0105] ​​The above merely provides an example of the present application, and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall fall into the scope of claims of the present application.

Claims

1. A carbon accounting method based on artificial intelligence, characterized in that: The method comprises: Collecting raw data from multiple data sources, and performing data cleaning and standardization on the raw data to obtain standardized data; Extracting carbon emission-related features from the standardized data, and screening the carbon emission-related features through a correlation analysis method to generate a feature combination; The feature combination is used to train the artificial intelligence model, and the pre-processed data of the object to be accounted for is input into the trained model to output the estimated value of carbon emissions.

2. The carbon accounting method based on artificial intelligence according to claim 1, characterized in that: Collecting raw data from multiple data sources, and performing data cleaning and standardization on the raw data to obtain standardized data, specifically including: Collect energy consumption data through sensors and obtain production activity data from the enterprise production management system; the energy consumption data includes real-time usage of electricity, coal, and natural gas, and the production activity data includes equipment operating parameters and production output; Extracting geographic information data and industry category data from a public database; the geographic information data includes climate conditions and altitude; Cleaning the original data to remove duplicate data, erroneous data, and missing data, and interpolating and filling missing data using the mean of adjacent time points; The cleaned multi-source data are normalized, and the time format of the normalized multi-source data is unified to obtain standardized data.

3. The carbon accounting method based on artificial intelligence according to claim 1, characterized in that: Extracting carbon emission-related features from the standardized data specifically includes: Based on energy consumption data, calculate the proportion of each type of energy consumption in total energy consumption, and use the proportion as the energy structure characteristic; Based on the production activity data, the ratio of output per unit time to equipment operating time is calculated, and the ratio is used as a production efficiency feature; Based on geographic information data, the impact of climate conditions on energy consumption is calculated and the characteristics of the geographic environment impact coefficient are generated.

4. The carbon accounting method based on artificial intelligence according to claim 3 is characterized in that: Based on geographic information data, the impact of climate conditions on energy consumption is calculated to generate geographic environment impact coefficient characteristics, including: Correlating climate condition data of the enterprise's location with energy consumption data; the climate condition may include temperature, humidity, or altitude; A mapping relationship between climate conditions and energy consumption change rate is constructed, and based on the mapping relationship, a correction coefficient of the geographical environment on energy consumption is calculated.

5. The carbon accounting method based on artificial intelligence according to claim 1, characterized in that: After inputting the pre-processed data of the object to be accounted for into the trained model and outputting the estimated carbon emissions value, the method further includes: Group the carbon emission estimates by time series, calculate the emission differences between each group, and call the chart engine to convert the differences into gradient color depths; Matching preset carbon emission thresholds based on industry category data and overlaying the coordinates of accounting objects exceeding the threshold on a digital map to generate a thermal layer; Identify the extreme points of carbon emission change rate in adjacent time periods, insert warning marks in the preset display method, and associate abnormal records of production activity data.

6. The carbon accounting method based on artificial intelligence according to claim 1, characterized in that: Input the pre-processed data of the object to be accounted for into the trained model, and output the estimated carbon emissions, including: Generate a carbon emissions digital report based on the time dimension, and mark the emission estimates corresponding to different accounting objects in the carbon emissions digital report; Correlate the emission estimates with the corresponding industry category data and mark high carbon emission hotspots in the carbon accounting area; The carbon emission change trend is displayed through a preset display method; the preset display method includes a bar chart or a line chart.

7. The carbon accounting method based on artificial intelligence according to claim 1, characterized in that: The method further comprises: When the amount of newly added multi-source data reaches a preset period, the newly added multi-source data is cleaned and standardized to generate an incremental data set; Based on the incremental data set, recalculate the correlation weights between the features and carbon emissions, and filter out features whose weights are lower than a preset threshold; An incremental learning algorithm is used to load the historical training model, and the weight matrix is ​​updated based on the incremental data set. After verifying that the loss function converges, the updated model version is determined.

8. The carbon accounting method based on artificial intelligence according to claim 1, characterized in that: The artificial intelligence model is trained using the feature combination, specifically including: Divide the feature set into a training set, a validation set, and a test set, and minimize the error between the output value and the true carbon emission value based on the training set and by adjusting the model weights and thresholds; Evaluate model performance based on the validation set. If model performance continues to decline, trigger an early stopping mechanism and optimize model hyperparameters. During the model training process, the accuracy, recall and mean square error of the model are calculated using the test set to verify the generalization ability of the model.

9. An artificial intelligence-based carbon accounting device, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the artificial intelligence-based carbon accounting method as described in any one of claims 1 to 8.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: When the computer-executable instructions are executed, an artificial intelligence-based carbon accounting method as described in any one of claims 1 to 8 is implemented.

Images