Method for calculating carbon emissions of engineering project on basis of financial expenditure data

By employing natural language processing and cross-validation techniques, combined with a multi-level carbon emission factor library and a dynamic unit cost database, the accuracy and efficiency issues of carbon emission parameter extraction from financial data have been resolved, enabling efficient and accurate carbon emission calculation for engineering projects.

WO2026233469A1PCT designated stage Publication Date: 2026-11-12HEBEI CONSTRUCTION ENGINEERING GROUP CO LIMITID
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2026/103204
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-12-26
Filing Date
2026-06-12
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Existing technologies for automatically extracting carbon emission-related technical parameters from financial records suffer from low accuracy and reliance on manual review. Furthermore, they lack effective cross-system data verification and dynamic cost matching mechanisms, resulting in inaccurate and inefficient carbon emission calculations.

Method used

Natural language processing technology is used to analyze financial expenditure data. Combined with confidence scoring and cross-validation processes, validated parameters are obtained from other systems. Through a multi-level carbon emission factor library and a dynamic unit cost database, the parameters are automatically optimized and accurately matched.

Benefits of technology

It improves the accuracy and efficiency of extracting carbon emission parameters from non-standard financial documents, reduces the need for manual review, ensures the reliability and accuracy of carbon emission calculations, and provides a self-optimizing closed-loop process.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present application belongs to the technical fields of carbon emission calculation and engineering project management. Disclosed is a method for calculating the carbon emissions of an engineering project on the basis of financial expenditure data, which aims to solve the problems of low confidence in extraction results and poor parameter accuracy when a natural language processing technique is used to automatically extract carbon-emission-related technical attribute parameters from item description text. The method comprises: parsing item description text in a financial expenditure record by means of a natural language processing technique to extract technical attribute parameters and generate a confidence score; and if the confidence score is lower than a preset threshold, triggering a cross-validation process, acquiring, on the basis of data association mapping rules, validated technical attribute parameters from other systems of an engineering project for comparison, determining finally adopted parameters on the basis of the credibility of a data source, and feeding a result back to a model and a professional dictionary, so as to realize closed-loop self-optimization. The method is mainly used for carbon footprint accounting, carbon emission monitoring and emission reduction management for engineering projects, and provides reliable data support for low-carbon decision-making.
Need to check novelty before this filing date? Find Prior Art

Description

Methods for calculating carbon emissions from engineering projects based on financial expenditure data

[0001] This patent application claims priority to Chinese Patent Application No. CN 202511987216.6, filed on December 26, 2025. The disclosure of the earlier application is incorporated herein by reference in its entirety. Technical Field

[0002] This application belongs to the field of carbon emission calculation and engineering project management technology, specifically involving a method for calculating carbon emissions of engineering projects based on financial expenditure data. Background Technology

[0003] In the full lifecycle management of engineering projects, accurate carbon emission accounting is a fundamental task for achieving low-carbon transformation and green construction. Current technologies primarily rely on two types of data inputs: firstly, activity data based on on-site measurements or energy and material consumption inventories; and secondly, theoretical calculations based on design drawings or bills of materials. However, both of these mainstream approaches face significant challenges in practical applications.

[0004] The primary challenge lies in reliably extracting the essential technical attributes necessary for carbon emission calculations from concise and non-standardized descriptions of goods in financial records. Descriptions in financial systems are often based on accounting subjects or internal conventions, frequently using abbreviations, aliases, or non-technical terms, such as referring to "C30 ready-mixed concrete" as "commercial concrete" or "ready-mixed concrete." Accurate matching of the carbon emission factor database highly depends on the specific technical specifications of materials or processes, such as concrete strength grade and steel type. When directly using natural language processing to parse such text, the technical parameters identified by the model often contain uncertainties or errors due to non-standard terminology and lack of context. Currently, when the confidence level of automated parsing results is low, the common approach is to rely on manual judgment and correction. However, for large projects with tens of thousands of financial records, manual review is inefficient, costly, and impractical for large-scale application. How to effectively improve the accuracy of automated extraction of key parameters from non-standardized text without relying entirely on manual methods remains a long-standing bottleneck.

[0005] Secondly, after extracting the technical parameters, ensuring their accuracy presents a verification challenge. Due to the lack of independent, comparable, and standardized data sources, the system often struggles to self-check the correctness of its own identification results. A deviation in the initial analysis can lead to subsequent errors in carbon emission factor matching, systematically impacting the final accounting results. Previous attempts to improve this by introducing structured data from other project management systems, such as bills of materials or building information model (BIM) data, have proven ineffective in practice. The relationships between financial records and data from other systems are often unclear or difficult to automatically correlate. Inconsistencies in identifiers and classification systems across different systems complicate the automatic retrieval and comparison of relevant technical parameters across systems. Furthermore, conflicting information from multiple data sources and the lack of objective rules for determining which data source is more reliable make it difficult to reliably establish an automated cross-validation mechanism.

[0006] Another key issue lies in the initial stage of data transformation: how to reasonably convert financial expenditures into the physical consumption of goods. The currently common method is to perform a division conversion based on a predefined, static mapping between item categories and unit costs. However, the unit costs of engineering materials and services are significantly affected by market price fluctuations, regional differences, procurement channels, and specific specifications. Using a fixed, averaged unit cost for conversion introduces significant errors, leading to discrepancies between the estimated physical quantities and the actual consumption of the project. While building a dynamic cost database is an obvious direction for improvement, in practice, obtaining high-quality, timely cost data sources and achieving accurate matching presents difficulties. Internal project procurement data is the most representative, but its format may be inconsistent and may not cover all expenditure categories; external authoritative market data is timely, but may differ from the actual procurement costs of a specific project. How to design priorities and matching rules to ensure that the unit costs used for conversion are both close to the project's reality and that data availability is guaranteed is a real problem that needs to be solved.

[0007] In summary, the core obstacles to using financial data for carbon emission calculations lie in two key areas: first, the automated and highly reliable conversion and verification from non-standardized text to precise technical parameters; and second, the highly accurate dynamic conversion from monetary amounts to physical quantities. These problems stem from the contradiction between the low technical information density of financial data and the high precision requirements of carbon emission accounting, as well as the complexity of cross-system data correlation and fusion. These difficulties have long prevented carbon accounting methods based on financial data from meeting the demands of refined management in terms of accuracy, efficiency, and automation. Summary of the Invention

[0008] The purpose of this application is to solve at least one of the aforementioned problems. To this end, this application provides a method for calculating carbon emissions from engineering projects based on financial expenditure data, the method comprising:

[0009] Obtain financial expenditure data for the engineering project. The financial expenditure data includes multiple expenditure records, and each expenditure record includes at least the item category, expenditure amount, and item description text.

[0010] Natural language processing technology is used to parse the description text of the item, identify and extract technical attribute parameters related to carbon emissions, and generate confidence scores for the technical attribute parameters.

[0011] When the confidence score is lower than a preset threshold, a cross-validation process is triggered. Based on the item category and key identifier of the current expenditure record, and according to preset data association mapping rules, at least one structured data source and corresponding verified technical attribute parameters are obtained from other systems of the project besides the financial system. The preset data association mapping rules include validating the verified technical attribute parameters. During the validation, the key features in the item description text and the update time of the data source are combined to filter out verified technical attribute parameters and corresponding structured data sources whose update time is within a set time limit and whose key features match.

[0012] The technical attribute parameters extracted using natural language processing technology are compared with the verified technical attribute parameters, and the final adopted technical attribute parameters are determined according to the preset data source credibility evaluation rules.

[0013] The final adopted technical attribute parameters and the corresponding item description text are used as training samples and fed back to the model corresponding to the natural language processing technology and the preset engineering materials and process professional dictionary to achieve iterative optimization of the model and dictionary;

[0014] Based on the predefined mapping relationship between item categories and unit costs, the expenditure amount of each expenditure record is converted into the physical quantity of the item;

[0015] Obtain the carbon emission factor corresponding to each expenditure record;

[0016] For each expenditure record, the physical quantity of the item is multiplied by the corresponding carbon emission factor to calculate the carbon emissions of that expenditure record.

[0017] The total carbon emissions of the project are obtained by summing the carbon emissions from all expenditure records. Beneficial effects

[0018] Compared with existing technologies, the beneficial technical effects of the calculation method provided in this application are as follows. By introducing a cross-validation and closed-loop self-optimization mechanism based on confidence scores, this application can proactively obtain verified parameters from other reliable data sources for comparison and confirmation when the uncertainty of the natural language processing model's recognition results is high, thereby effectively improving the accuracy and reliability of technical attribute parameter extraction. This mechanism reduces the risk of parameter misjudgment directly caused by non-standard text descriptions, and by feeding back verified correct samples to the model and dictionary, it achieves continuous learning and self-improvement of the system, reducing long-term reliance on manual review. It can solve the problem of low confidence in extraction results due to non-standard text descriptions or model recognition limitations when automatically extracting carbon emission-related technical attribute parameters from item description texts in financial expenditure records using natural language processing technology, thus affecting parameter accuracy. Embodiments of the present invention

[0019] To further illustrate the technical means and effects adopted in this application, the following detailed description is provided in conjunction with specific embodiments. The specific embodiments described herein are merely for explaining this application and are not intended to limit it.

[0020] According to one embodiment of this application, a method for calculating carbon emissions from engineering projects based on financial expenditure data includes the following steps:

[0021] S10. Obtain the financial expenditure data of the project, wherein the financial expenditure data includes multiple expenditure records, and each expenditure record includes at least the item category, expenditure amount and item description text;

[0022] S11. Use natural language processing technology to parse the item description text, identify and extract technical attribute parameters related to carbon emissions, and generate confidence scores for the technical attribute parameters.

[0023] S12. When the confidence score is lower than a preset threshold, a cross-validation process is triggered. Based on the item category and key identifier of the current expenditure record, and based on preset data association mapping rules, at least one structured data source and corresponding verified technical attribute parameters are obtained from other systems of the project other than the financial system. The preset data association mapping rules include validating the verified technical attribute parameters. During the validation, the key features in the item description text and the update time of the data source are combined to filter out verified technical attribute parameters and corresponding structured data sources whose update time is within the set time limit and whose key features match.

[0024] S13. The technical attribute parameters extracted using natural language processing technology are compared with the verified technical attribute parameters, and the final adopted technical attribute parameters are determined according to the preset data source credibility assessment rules.

[0025] S14. The final adopted technical attribute parameters and the corresponding item description text are used as training samples and fed back to the model corresponding to the natural language processing technology and the preset engineering materials and process professional dictionary to achieve iterative optimization of the model and dictionary.

[0026] S20. Based on the predefined mapping relationship between item categories and unit costs, convert the expenditure amount of each expenditure record into the physical quantity of the item.

[0027] S30. Obtain the carbon emission factor corresponding to each expenditure record;

[0028] S40. For each expenditure record, multiply the physical quantity of the item by the corresponding carbon emission factor to calculate the carbon emissions of the expenditure record.

[0029] S50. Sum the carbon emissions of all expenditure records to obtain the total carbon emissions of the project.

[0030] The calculation method provided in this embodiment can systematically convert financial expenditure data into carbon emissions. The calculation first requires acquiring the financial expenditure data of the engineering project. This data contains multiple expenditure records, each with key information such as item category, expenditure amount, and item description text. This information serves as input for subsequent conversion and calculation processes. The first step in the conversion and calculation process is to convert the expenditure amount into the physical quantity of the item according to a predefined mapping relationship. This step relies on a static correspondence table between item categories and unit costs; for example, mapping the category of ready-mixed concrete to a fixed unit cost such as 400 yuan / cubic meter.

[0031] The second step is to obtain the carbon emission factor corresponding to each expenditure record. Multiplying this carbon emission factor by the physical quantity of the item obtained in the first step yields the carbon emissions generated by that expenditure. This calculation is performed for each expenditure. Finally, the system sums the carbon emissions calculated from all expenditure records to obtain the total carbon emissions for the project.

[0032] In this application, after calculating the carbon emission data corresponding to each detailed financial expenditure, project management can be conducted to minimize carbon emissions. Specifically, one or multiple expected emission caps can be set for different categories of carbon-emitting activities. After obtaining a carbon emission data point, it can be determined whether the data exceeds the emission cap. When the emission cap is exceeded, the corresponding detailed financial expenditure data is written into a list. Finally, after calculating the carbon emissions corresponding to all detailed financial expenditure data, a complete list is obtained. This list can be sent to a monitor for display. Project managers can view which financial data exceeds the carbon emission limit (exceeding the emission cap) from this list, and then perform corresponding management work, such as seeking lower-carbon-emission alternatives for activities exceeding the carbon emission limit, and using these alternative methods to replace the original carbon-emitting activities, thereby reducing carbon emissions.

[0033] In this embodiment, while calculating carbon emissions, an automated multi-source dynamic cross-validation and closed-loop self-optimization mechanism is introduced through steps S11 to S14. The process shown in steps S11 to S14 is illustrated below with an example. First, financial expenditure data for a specified time period is obtained from the project's financial management system. This data contains multiple expenditure records, each with at least three fields: item category, expenditure amount, and item description text. Then, a pre-trained natural language processing model is used to parse the item description text, aiming to identify and extract key technical attribute parameters related to carbon emission calculation, such as material specifications, model, or strength grade. While outputting the extracted parameters, the model generates a confidence score indicating the reliability of the identification result. The system presets a confidence threshold, for example, 80%. When the confidence score output by the model is lower than this threshold, it indicates that the uncertainty of the automated parsing result is high, and the system will automatically trigger the cross-validation process.

[0034] The cross-validation process relies on pre-defined data association mapping rules. These rules define the association paths between key identifiers in financial expenditure records (such as purchase order numbers, contract numbers, or supplier names) and key identifiers of related components or bills of materials in other project management systems (such as material requirements planning systems or building information models). Based on the item category and key identifiers of the current low-confidence expenditure record, the system proactively retrieves one or more structured data sources from relevant systems according to these mapping rules, and extracts validated technical attribute parameters related to the current record. For example, it extracts the accurate technical specifications of materials from the purchase order details of the material requirements planning system.

[0035] After acquiring validated parameters from other systems, the system compares the parameters initially extracted by the natural language processing model with these parameters. This comparison process is not a simple equivalence judgment, but rather determines the final adopted parameter based on a set of preset data source credibility assessment rules. These rules comprehensively consider the reliability history of the data source itself, such as the consistency ratio between data provided by a certain system and human review results in past cross-validation, as well as the matching degree of the current association. If the parameters provided by multiple data sources are consistent, then that parameter is adopted; if there is a conflict, the parameter from the data source with higher historical credibility is adopted.

[0036] Ultimately, the selected technical attribute parameters, along with the corresponding original item description text, constitute a high-quality training sample, which is then fed back to the natural language processing model and a pre-defined engineering materials and processes dictionary. This dictionary stores the standardized names, aliases, and corresponding standard technical parameters of items within the engineering field. Through this feedback, the model can be further optimized on specific domain texts, and the coverage and accuracy of the dictionary are enhanced, thus achieving closed-loop self-optimization of the entire technical attribute parameter extraction process. Furthermore, the data association mapping rules also include a step to validate the validity of verified parameters obtained from other systems. During validation, key features in the item description text and the update time of the data source are considered to ensure that the adopted parameter information matches the key features and is within the set validity period. The following describes the calculation method provided in this implementation based on a specific financial expenditure record of a construction project. The financial expenditure record shows that the item category is commercial concrete, the expenditure amount is 500,000 yuan, and the item description text is commercial concrete. The system uses a natural language processing model to parse this description text, attempting to extract the concrete strength grade parameter. The model output parameter was C30, but the generated confidence score was 60%, lower than the system's preset threshold of 80%. The system then automatically triggered a cross-validation process. According to preset rules, the system retrieved the corresponding purchase order from the project's material requirements planning (MRP) system by associating it with the purchase contract number. The order details clearly stated that the concrete technical parameter was C30. The system extracted this validated parameter and compared it with the C30 extracted by the model; they matched. The data source credibility assessment rules determined that the MRP system had high historical credibility, therefore the system ultimately adopted C30 as the technical attribute parameter for this record. Simultaneously, the correspondence between ready-mixed concrete and C30 was used as a training sample and fed back to the natural language processing model and a specialized dictionary for optimization. After this, the record did not need to be pushed to the manual review interface.

[0037] Existing technologies generally use the same natural language processing model to parse financial expenditure records. For the same record describing an item as ready-mixed concrete, the model will also output the parameter C30 with a confidence score of 60%. However, existing technologies lack a subsequent automated cross-validation mechanism. There are typically two approaches: First, setting a low confidence threshold and forcibly adopting the low-confidence parameter. This may lead to the adoption of incorrect parameters due to textual ambiguity; for example, ready-mixed concrete might refer to C25 or C35, thus introducing errors in subsequent carbon emission factor matching. Second, pushing all low-confidence records to a manual review interface for manual interpretation and input of the correct parameters. For large projects with massive amounts of financial records, this approach creates a huge burden of manual review, is inefficient, and makes it difficult to ensure consistency in judgment standards among different reviewers.

[0038] The computational method provided in this embodiment introduces an automated multi-source dynamic cross-validation and closed-loop self-optimization mechanism. Existing technologies, when the confidence level of a natural language processing model is insufficient, face a dilemma of sacrificing either accuracy or efficiency. This application, however, intelligently connects to structured data sources from other management systems within the project through preset rules, using the verified technical parameters to validate and confirm the model's initial results. This process not only automatically resolves low confidence issues in most cases, reducing reliance on manual review and improving overall processing efficiency, but more importantly, by adopting parameters validated by other systems, it significantly improves the accuracy of the final technical attribute parameters. Furthermore, feeding back validated correct samples to the model and dictionary enables continuous learning and evolution of the system, allowing the entire method to continuously optimize with increased usage, resulting in enhanced adaptability and reliability.

[0039] This embodiment effectively solves the problem of automatically extracting accurate technical attribute parameters from non-standardized financial texts using the claimed method. When the reliability of automated identification results is insufficient, it does not simply rely on manual intervention or forced adoption, but actively activates cross-system data verification capabilities, utilizing existing reliable data sources within the project for cross-checking. This significantly reduces the need for manual intervention while ensuring the accuracy of parameter acquisition. This method constructs a closed-loop process from identification and verification to self-optimization, not only improving the reliability of single parameter extraction but also enabling the system to adapt and evolve through continuous feedback learning. This provides a solid and high-quality data foundation for subsequent accurate matching of carbon emission factors based on financial data and reliable carbon emission accounting, making efficient and accurate carbon management using existing financial system data possible.

[0040] In one implementation, step S30, which involves obtaining the carbon emission factor corresponding to each expenditure record, includes the following steps:

[0041] S31. Pre-construct a multi-level carbon emission factor library, wherein at least one item category is associated with multiple sub-category carbon emission factors, and each sub-category carbon emission factor corresponds to one or more technical attribute parameters.

[0042] S32. Parse the item description text in the expenditure record, and identify and extract technical attribute parameters related to carbon emissions through natural language processing technology;

[0043] S33. Based on the item category and the identified technical attribute parameters, query the multi-level carbon emission factor library, match and obtain a corresponding sub-category carbon emission factor;

[0044] In this embodiment, step S30, which involves obtaining the carbon emission factor corresponding to each expenditure record, is achieved through multiple sub-steps. This embodiment constructs a multi-level carbon emission factor library. This library is organized according to item categories, with each category further subdivided into multiple subcategories. Each subcategory corresponds to a specific set of technical attribute parameters and is associated with a specific carbon emission factor value. For example, under the item category of ready-mixed concrete, multiple subcategories such as C25, C30, and C35 can be established, each corresponding to different concrete strength grade parameters and carbon emission factors. This multi-level carbon emission factor library design ensures a detailed characterization of the differences in carbon emissions for items with different technical specifications, overcoming the coarseness of using a single average factor.

[0045] By constructing a complete calculation chain from financial expenditure amount to physical quantity of goods, and then to matching carbon emission factors based on a multi-level factor library, this application provides a feasible path for systematically utilizing financial data for carbon accounting. This method avoids the cost of dedicated carbon activity data collection and utilizes engineering-specific dictionaries and natural language processing technology to improve the automation and standardization of extracting key parameters from non-standard text, laying the foundation for subsequent accurate calculations and making carbon accounting results based on financial data more practically valuable.

[0046] In one implementation, the engineering materials and processes dictionary stores the standardized names, common aliases, abbreviations, and corresponding standard technical attribute parameters of items in the engineering field; step S32, namely parsing the item description text in the expenditure record, and identifying and extracting the technical attribute parameters related to carbon emissions through natural language processing technology, specifically includes:

[0047] S321. Use the engineering materials and processes professional dictionary to preprocess the text and standardize the terminology, mapping the non-standard expressions in the text into standardized technical terms.

[0048] S322. Based on the standardized text (i.e. the technical terms of the standardization), perform a natural language processing model to identify and extract technical attribute parameters, and generate a confidence score for each extracted parameter;

[0049] S323. If the confidence score is lower than a preset threshold, the expenditure record and its item description text are pushed to the manual review interface, and the correct technical attribute parameters are received after manual review. The parameters and their corresponding relationships are fed back to the engineering materials and processes professional dictionary to optimize the dictionary and natural language processing model.

[0050] The technical solution provided in this embodiment can solve the problem of how to effectively convert financial expenditures into physical quantities of goods and accurately match the corresponding carbon emission factors, so as to complete the systematic calculation of carbon emissions from financial data.

[0051] In this embodiment, an engineering materials and processes dictionary is used to parse the item description text in expenditure records to extract key technical attribute parameters. This dictionary stores the standardized names, common aliases, and abbreviations of various items in the engineering field, and maps them to standard technical attribute parameters. When parsing the text, the dictionary is first used for preprocessing and terminology standardization; for example, the term "ready-mixed concrete" in the text is automatically mapped to the standardized term "commercial concrete." Based on the standardized text, a natural language processing model is executed to identify and extract specific technical attribute parameters, such as concrete grade C30. Simultaneously, the model generates a confidence score for each extracted parameter to quantify the reliability of the identification.

[0052] If the confidence score is lower than a preset threshold, such as 80%, in this embodiment, the system pushes the expenditure record and its item description text to a dedicated manual review interface, waiting for the reviewer to input the correct technical attribute parameters as determined. The correct parameters input by the reviewer and their correspondence with the original description text are captured by the system and fed back to the engineering materials and processes dictionary and the natural language processing model to optimize dictionary content and model performance, achieving further closed-loop learning.

[0053] After obtaining the technical attribute parameters determined through the above process, the system queries the aforementioned multi-level carbon emission factor library according to the item category and specific parameters to which the expenditure record belongs, accurately matches and obtains a corresponding sub-category carbon emission factor, for example, the factor corresponding to C30 concrete is 280 kgCO2e / m³.

[0054] For each expenditure record, the carbon emissions generated by that expenditure can be obtained by multiplying the calculated physical quantity of the item by the carbon emission factor matched to that record.

[0055] A financial record for a certain engineering project shows the item category as steel bars, the expenditure amount as 1,000,000 yuan, and the item description as grade III rebar. The system, based on the mapping relationship, uses a static unit cost of 5,000 yuan / ton to convert it into a physical quantity of 200 tons. Grade III rebar is standardized using a professional dictionary, and a natural language processing model extracts the technical parameter as HRB400 with a confidence level of 90%, which is above the threshold. Based on this, the system queries the HRB400 subclass of steel bars in a multi-level carbon emission factor database, obtaining a carbon emission factor of 2300 kgCO2e / ton. The carbon emission of this record is calculated to be 460,000 kgCO2e. Finally, all records for the project are calculated and summed in this way to obtain the total emission.

[0056] For the same financial data, existing technology first requires manual intervention based on experience or simple queries to determine and fill in an approximate unit cost for each record and perform physical quantity conversion. This process is time-consuming and prone to errors due to price fluctuations or differences in individual experience. Subsequently, when matching carbon emission factors, due to the lack of a structured multi-level factor library and professional text parsing tools, two simplified methods are usually adopted: one is for manual reading of item descriptions and searching for matches in a limited list of general factors based on personal knowledge, which is labor-intensive and inconsistent; the other is to assign a single average emission factor to the entire item category, such as all steel bars, ignoring the differences in carbon emissions between different specifications and models. Although this method can complete the calculation, its efficiency and accuracy are highly dependent on human experience, making it difficult to scale. Furthermore, due to coarse conversion and inaccurate factor matching, the final accounting results have high uncertainty.

[0057] The computational method provided in this embodiment systematically replaces the traditional, manually-driven, discrete, and coarse-grained calculation process by constructing a standardized and automated complete computational chain. This application achieves rapid conversion through predefined mapping, automates and standardizes text parsing by introducing engineering professional dictionaries and natural language processing technology, and achieves precise parameter matching through a structured multi-level factor library. This transforms multiple steps that previously relied on human experience into a repeatable and optimizable automated process. The manual review and feedback mechanism at low confidence levels ensures the accuracy of key nodes while also providing fuel for continuous system optimization.

[0058] This implementation constructs an end-to-end systematic calculation method for carbon emissions, integrating previously cumbersome and error-prone fragmented processing steps into a coherent, standardized, and automated process. This method fully utilizes readily available financial data from projects, avoiding the additional costs of collecting activity data specifically for carbon accounting. By introducing an engineering lexicon and natural language processing, the ability to understand non-standard descriptive text and the efficiency of parameter extraction are significantly improved, laying a reliable foundation for subsequent accurate calculations. The multi-level carbon emission factor library design ensures a detailed characterization of the differences in carbon emissions across different technical specifications, overcoming the coarsening of using a single average factor. This entire approach enables relatively accurate, efficient, and scalable carbon emission accounting based on financial expenditure data, providing a practical quantitative tool for carbon management in engineering projects.

[0059] In one implementation, the cross-validation process in step S12 includes:

[0060] S1201. Based on the item category of the current expenditure record, obtain the associated structured data from other data systems of the project besides the financial system, including the material requirements planning system or building information model.

[0061] S1202, Extract verified technical attribute parameters related to the current expenditure record from the associated structured data;

[0062] S1203. Compare the technical attribute parameters identified by natural language processing technology with the technical attribute parameters extracted from the associated structured data;

[0063] S1204. If the comparison is consistent, the technical attribute parameter will be automatically adopted and used as a training sample along with the corresponding item description text to update the natural language processing model. The record will no longer be pushed to the manual review interface.

[0064] In this embodiment, when the confidence score of the identified technical attribute parameter is lower than a preset threshold, the aforementioned cross-validation process is automatically triggered. The technical solution provided by this embodiment addresses the problem of how to reduce reliance on manual review and improve the efficiency and accuracy of automated system processing when the confidence score of technical parameters extracted by automated text parsing technology is low, by introducing other trusted data sources for cross-validation.

[0065] For example, for the description text of ready-mixed concrete, the parameters extracted by natural language processing have low confidence, triggering cross-validation. The system searches for relevant records in the material requirements planning system and extracts the validated parameter, concrete grade C30. If it matches the parameter extracted by natural language processing, the system automatically adopts C30 and uses the correspondence between ready-mixed concrete and C30 as training samples to update the model, avoiding manual review.

[0066] Through an automated cross-validation mechanism, when the system's confidence level is insufficient, it can proactively retrieve validated data from other business systems for comparison. This significantly reduces the number of records that need to be pushed to manual review and reduces reliance on manual review. While ensuring data accuracy, it significantly improves the automation level and processing efficiency of the entire calculation process. When the comparison structure is consistent, it is automatically adopted and the model is updated, providing high-quality training samples for model optimization. This enhances the system's automation level and learning ability, further reducing the need for manual intervention.

[0067] In one implementation, step S1201, which involves obtaining relevant structured data from data systems other than the financial system for the project based on the item category of the current expenditure record, includes:

[0068] S1201a. Construct a data association mapping table, which defines one or more potential association paths between key identifiers in financial expenditure records and key identifiers of components or bills of materials in the material requirements planning system and building information model;

[0069] S1201b: Based on the data association mapping table, attempt to match one or more associated structured data sources for the current expenditure record, and generate an association confidence score for each matched data source;

[0070] S1201c, Predefine and dynamically maintain a source credibility score for each potential data source, the source credibility score being calculated and updated based on the consistency ratio between the data source and the manual review results in historical cross-validation;

[0071] Step S1203 involves comparing the technical attribute parameters identified by natural language processing technology with the technical attribute parameters extracted from the associated structured data, including:

[0072] S1203a. When making comparisons, prioritize the technical attribute parameters provided by data sources whose association confidence and source confidence scores are both higher than their respective thresholds for comparison.

[0073] S1203b: If there are multiple data sources that meet the conditions and provide the same technical attribute parameters, then the parameter shall be adopted; if there are conflicts in the parameters provided by multiple data sources, then the parameter from the data source with the highest source credibility score shall be adopted.

[0074] The technical solution provided in this embodiment can solve the problems of how to intelligently associate and select the most reliable technical attribute parameters from multiple systems in an engineering project during cross-validation, and how to make decisions when there are conflicts in multi-source data, so as to ensure the accuracy of parameter acquisition.

[0075] For example, for expenditure records, the system uses a data association mapping table to match the Material Requirements Planning (MRP) system and the Building Information Model (BIM) system. Matched MRP records have an association confidence of 90% and a source confidence score of 95%, while BIM records have an association confidence of 85% and a source confidence score of 90%. Both provide parameter C30, and the system adopts C30. If the MRP system provides C30 while the BIM system provides C25, the parameter C30 from the MRP system with the higher source confidence score is adopted.

[0076] This embodiment constructs an intelligent data association mapping and credibility assessment mechanism, enabling the selection of technical attribute parameters from multiple data sources, effectively identifying and utilizing the most reliable data source. Furthermore, in the event of data conflicts, this mechanism provides a clear decision-making basis based on source credibility scores. The aforementioned mechanism provided by this embodiment improves the accuracy and reliability of cross-validation, enhancing the robustness and adaptability of the calculation method in this application.

[0077] In one implementation, step S20, which involves converting the expenditure amount of each expenditure record into the physical quantity of the item based on a predefined mapping relationship between item categories and unit costs, specifically includes:

[0078] S21. Construct and maintain a dynamic unit cost database that records the market unit cost of different item categories in different time intervals and / or different regions.

[0079] S22. Extract transaction time information and project location information from pending expenditure records;

[0080] S23. Based on the item category of the expenditure record, and in conjunction with the transaction time information and project location information, query the dynamic unit cost database to match and obtain a dynamic unit cost corresponding to the time and space.

[0081] S24. Replace the predefined static unit cost with the dynamic unit cost, and divide the expenditure amount by the dynamic unit cost to obtain the calibrated physical quantity of the item.

[0082] The technical solution provided in this embodiment can solve the problem of how to overcome the conversion error caused by market fluctuations and regional differences in static unit cost when converting financial expenditure amount into physical quantity of goods, thereby improving the accuracy of physical quantity calculation.

[0083] For example, a financial expenditure record for a project in East China in 2023 shows that the item category was steel bars and the expenditure amount was 1,000,000 yuan. The system extracts the transaction time as the third quarter of 2023 and the project location as East China. A query of the dynamic unit cost database shows that the dynamic unit cost of steel bars in that region at that time was 4,000 yuan per ton. Dividing the expenditure amount of 1,000,000 yuan by 4,000 yuan per ton yields a physical quantity of 250 tons of steel bars.

[0084] By employing a dynamic unit cost database, the conversion from monetary value to physical quantity reflects actual fluctuations in market prices and regional differences. This effectively overcomes the systematic biases introduced by using static average costs, significantly improving the accuracy and reliability of physical quantity estimations and providing more reliable input for subsequent carbon emission calculations.

[0085] In one implementation, the dynamic unit cost database integrates multiple data sources, including at least one authoritative market quotation platform data source and at least one actual procurement data source from within the project.

[0086] Furthermore, in step S23, the rule for matching and obtaining the dynamic unit cost is as follows:

[0087] S231. Prioritize querying the actual procurement data source within the project. If there is an actual procurement unit price in the actual procurement data source that matches the item category, transaction time, and project region of the current expenditure record, then use the actual procurement unit price as the dynamic unit cost.

[0088] S232. If no corresponding actual purchase unit price is found in the actual procurement data source, the authoritative market quotation platform data source is queried to obtain the matching market unit cost, and the market unit cost is used as the dynamic unit cost.

[0089] The technical solution provided in this embodiment can solve the problem of data source selection and priority in dynamic unit cost database, so as to ensure that the obtained unit cost is timely, authoritative and highly consistent with the actual cost of the project.

[0090] For example, the system processes a steel rebar procurement record from East China in 2023. First, it queries the project's internal procurement database and finds a record for steel rebar procurement in the same period and region with an actual unit price of 4100 yuan per ton; this unit price is then directly adopted. If no matching record is found in the internal database, it queries an authoritative market quotation platform to obtain a market reference unit price of 3950 yuan per ton for that period and adopts this price.

[0091] By setting priority rules for internal and external data sources, the system ensures that the unit costs used are both highly representative and closely reflect the actual procurement situation of the project, while also being supplemented by authoritative market data when internal data is lacking, thus guaranteeing data availability. This strategy balances data representativeness and availability, ensuring the practicality and reliability of the dynamic unit cost database.

[0092] In one implementation, after step S50 (i.e., summing the carbon emissions of all expenditure records to obtain the total carbon emissions of the project), the method further includes:

[0093] S60. Calculate the uncertainty assessment index of the total carbon emissions of the project.

[0094] Optionally, step S60 includes:

[0095] S61. Define an error distribution for the key input parameters in the calculation chain, wherein the key input parameters include at least the unit cost and / or the carbon emission factor;

[0096] S62. Based on the Monte Carlo simulation method, the calculation process of all expenditure records of the project is repeated multiple times. In each calculation, the key input parameters are randomly sampled according to their respective error distributions to obtain a series of simulated values ​​of total carbon emissions.

[0097] S63. Based on the simulated values ​​of the series of total carbon emissions, determine the confidence interval and probability distribution of the total carbon emissions;

[0098] S64. When outputting the total carbon emissions of the project, output its confidence interval in conjunction with the output.

[0099] The technical solution provided by this embodiment can address the uncertainty of how to quantify the final total carbon emissions calculation results, thereby providing decision-makers with assessment information on the reliability of the calculation results, rather than just a single valuation issue.

[0100] For example, the system calculates the total carbon emissions of a project based on a baseline of 10,000 tons of CO2 equivalent. An uncertainty assessment is then performed to define a normal error distribution for the unit cost and carbon emission factor. 10,000 Monte Carlo simulations are run to obtain 10,000 simulated total carbon emissions values. Based on these simulations, a 95% confidence interval is determined to be between 9,500 and 10,500 tons of CO2 equivalent. The final output is a total carbon emissions figure of 10,000 tons of CO2 equivalent, along with its confidence interval.

[0101] By introducing uncertainty assessment based on Monte Carlo simulation, the total carbon emissions are expanded from a fixed point estimate to a probability distribution with a confidence interval. This provides policymakers with an intuitive quantitative indicator of the reliability of the calculation results, helping to more scientifically understand and utilize carbon accounting results for risk assessment and decision-making, and avoiding over-reliance on a single numerical value.

[0102] In one implementation, step S62, which involves repeatedly calculating all expenditure records for the project based on Monte Carlo simulation, is specifically optimized as follows:

[0103] S62a. Before performing a full Monte Carlo simulation, perform a global sensitivity analysis to determine the contribution of each expenditure record to the uncertainty of total carbon emissions.

[0104] S62b. Based on the aforementioned contribution level, all expenditure records are divided into a critical record set and a non-critical record set;

[0105] S62c. In subsequent Monte Carlo simulations, the key record set is simulated with full precision, that is, the key input parameters are randomly sampled in each simulation.

[0106] S62d. For the non-critical record set, an approximation processing strategy is adopted. The strategy is to calculate a baseline carbon emission value for it at the beginning of the simulation, and in subsequent simulations, multiply the baseline value by a global micro-perturbation factor to represent its fluctuation. The micro-perturbation factor is generated based on the overall fluctuation trend of the critical record set in the current simulation.

[0107] The technical solution provided in this embodiment can solve the problem of excessive consumption of computing resources and low efficiency caused by the large number of expenditure records in the full-project Monte Carlo simulation, and how to significantly improve computing efficiency while ensuring the accuracy of the evaluation.

[0108] For example, an engineering project has 1000 expenditure records. The system first performs a global sensitivity analysis to determine that 50 records contribute more than 90% to the uncertainty of total carbon emissions; these are classified as the critical record set, while the remaining 950 are classified as the non-critical record set. In the subsequent 10,000 Monte Carlo simulations, the system performs full-precision random sampling simulations on the 50 critical records each time. For the 950 non-critical records, a baseline carbon emission of 2000 tons of CO2 equivalent is calculated in the first simulation. In subsequent simulations, a micro-perturbation factor is generated based on the overall fluctuation of the critical record set, for example, 0.95 to 1.05, and the baseline value of 2000 tons is multiplied by this factor to represent its fluctuation.

[0109] By combining global sensitivity analysis with a differentiated simulation strategy, the computational load of Monte Carlo simulations is significantly reduced while maintaining the core accuracy of uncertainty assessment. This approach addresses the problem of inefficiency in large-scale project simulations, enabling rapid and frequent uncertainty assessments for complex engineering projects.

[0110] In one implementation, after step S50 (i.e., summing the carbon emissions of all expenditure records to obtain the total carbon emissions of the project), the method further includes:

[0111] S70. Attribution analysis of carbon emissions from engineering projects based on the responsibility dimension of configuration.

[0112] Optionally, step S70 includes:

[0113] S71. Predefine multiple responsibility dimensions, which include at least: responsibility subject dimension, engineering structure dimension, and time phase dimension;

[0114] S72. Construct a dimension mapping rule base, wherein the rule base defines the mapping relationship between the feature information in the expenditure record and each responsibility dimension;

[0115] S73. For each expenditure record, based on the feature information it contains, determine its attribution under each responsibility dimension through the dimension mapping rule base;

[0116] S74. Based on the aforementioned attribution relationship, allocate the carbon emissions recorded in the expenditure records to the corresponding nodes under each responsibility dimension;

[0117] S75. Aggregate carbon emissions by responsibility dimension to generate a multi-dimensional carbon emission attribution analysis report.

[0118] The technical solution provided in this embodiment can solve the problem of how to decompose and attribute the total carbon emissions of a project according to different management needs (such as responsible entities, project locations, and time phases) to support accurate carbon management and responsibility allocation.

[0119] Optionally, the system predefined responsible entity dimensions include the construction unit and the contractor; the engineering structure dimension includes the foundation and main structure; and the time stage dimension includes the preparation stage and the construction stage. For a steel reinforcement procurement expenditure record, it is mapped to the contractor based on its supplier information, to the main structure based on its usage location information, and to the construction stage based on its transaction time. After allocating the carbon emissions of this record to the above-mentioned dimension nodes, the system can aggregate and generate reports displaying the carbon emission percentage of the contractor, the carbon emission percentage of the main structure, and the carbon emission percentage of the construction stage.

[0120] By employing multi-dimensional carbon emission attribution analysis, total carbon emissions are decomposed according to management needs, enabling precise positioning of carbon responsibility. This helps managers clearly identify carbon emission hotspots and responsible parties, providing direct data support for developing targeted emission reduction measures and implementing refined carbon management. In one implementation, the dimension mapping rule base in step S72 supports dynamic configuration;

[0121] Step S73, that is, for each expenditure record, based on its contained feature information and through the dimensional mapping rule base, determines its attribution under each responsibility dimension, including:

[0122] S73a. When a single expenditure record has multiple possible attributions under a certain responsibility dimension, a weighted allocation strategy is adopted to split the carbon emissions of the expenditure record into multiple attribution nodes according to a preset weight ratio.

[0123] S73b provides a rule conflict detection mechanism, which prompts the user to optimize the rules when a conflict is detected in the mapping rules.

[0124] Step S74, which involves allocating the carbon emissions recorded in the expenditure records to the corresponding nodes under each responsibility dimension based on the attribution relationship, includes:

[0125] S74a supports source tracing queries based on attribution results, enabling reverse lookup of all original expenditure records and their detailed calculation processes that constitute the aggregated carbon emissions from any dimension node.

[0126] The technical solution provided in this embodiment can solve the problems of how to reasonably allocate carbon emissions when a single expenditure record can be attributed to multiple dimension nodes in the carbon emission attribution process, and how to ensure the flexibility of attribution rules and the traceability of results.

[0127] For example, an expenditure record shows the purchase of general-purpose cement for both main structure construction and temporary road paving. At the engineering structure level, this record has two possible classifications. The system uses a weighted allocation strategy to split the carbon emissions of this record into two nodes: main structure and temporary road, in a 7:3 ratio. The system detects that two mapping rules are attempting to match the same type of expenditure record and prompts the user to optimize the rules. The user can reverse-engineer all the original expenditure records that constitute this result, including this split cement purchase record and its calculation process, from the aggregated carbon emissions of the main structure node.

[0128] By supporting weight allocation, rule conflict detection, and result traceability, the multidimensional attribution analysis process becomes more flexible, transparent, and auditable. This enhances the ability to handle complex attribution relationships, ensures the rationality and consistency of attribution rules, and facilitates users in tracing and verifying aggregation results, thereby improving the credibility and practical value of attribution analysis reports.

[0129] The above embodiments are used to explain the inventive concept of this application and are not intended to limit the scope of protection of this application. Technical solutions that apply the inventive concept of this application, as well as technical solutions that have made obvious modifications and changes to the technical solutions, are all within the scope of protection of this application.

Claims

1. A method of calculating carbon emissions for an engineering project based on financial expenditure data, characterised in that, Includes the following steps: Obtain financial expenditure data for engineering projects. The financial expenditure data includes multiple expenditure records, and each expenditure record includes at least the item category, expenditure amount, and item description text. Natural language processing technology is used to parse the description text of the item, identify and extract technical attribute parameters related to carbon emissions, and generate confidence scores for the technical attribute parameters. When the confidence score is lower than a preset threshold, a cross-validation process is triggered. Based on the item category and key identifier of the current expenditure record, and according to preset data association mapping rules, at least one structured data source and corresponding verified technical attribute parameters are obtained from other systems of the project besides the financial system. The preset data association mapping rules include validating the verified technical attribute parameters. During the validation, the key features in the item description text and the update time of the data source are combined to filter out verified technical attribute parameters and corresponding structured data sources whose update time is within a set time limit and whose key features match. The technical attribute parameters extracted using natural language processing technology are compared with the verified technical attribute parameters, and the final adopted technical attribute parameters are determined according to the preset data source credibility evaluation rules. The final adopted technical attribute parameters and the corresponding item description text are used as training samples and fed back to the model corresponding to the natural language processing technology and the preset engineering materials and process professional dictionary to achieve iterative optimization of the model and dictionary; Based on the predefined mapping relationship between item categories and unit costs, the expenditure amount of each expenditure record is converted into the physical quantity of the item; Obtain the carbon emission factor corresponding to each expenditure record; For each expenditure record, the physical quantity of the item is multiplied by the corresponding carbon emission factor to calculate the carbon emissions of that expenditure record. The total carbon emissions of the project are obtained by summing the carbon emissions from all expenditure records.

2. The method of claim 1, wherein, The process of obtaining the carbon emission factor corresponding to each expenditure record includes: A multi-level carbon emission factor library is pre-built, in which at least one item category is associated with multiple sub-category carbon emission factors, and each sub-category carbon emission factor corresponds to one or more technical attribute parameters. The text describing items in expenditure records is analyzed, and technical attribute parameters related to carbon emissions are identified and extracted using natural language processing technology. Based on the item category and the identified technical attribute parameters, query the multi-level carbon emission factor library, match and obtain a corresponding sub-category carbon emission factor; 3. The method for calculating carbon emissions from engineering projects based on financial expenditure data as described in claim 2, characterized in that: The engineering materials and processes dictionary stores the standard names, common aliases, abbreviations, and corresponding standard technical attribute parameters of items in the engineering field; The item description text in the parsed expenditure records is used to identify and extract technical attribute parameters related to carbon emissions through natural language processing technology, including: The text is preprocessed and terminology is standardized using a professional dictionary of engineering materials and processes, mapping non-standard expressions in the text into standardized technical terms. Based on the technical terminology of the aforementioned standard, a natural language processing model is executed to identify and extract technical attribute parameters, and a confidence score is generated for each extracted parameter; When the confidence score is lower than the preset threshold, the expenditure record and its item description text are pushed to the manual review interface. The correct technical attribute parameters are received after manual review, and the parameters and corresponding relationships are fed back to the engineering materials and processes professional dictionary to optimize the dictionary and natural language processing model.

4. The method for calculating carbon emissions from engineering projects based on financial expenditure data as described in claim 1, characterized in that, The other systems in the project besides the financial system include a material requirements system and a building information model; the process of obtaining at least one structured data source and corresponding verified technical attribute parameters from the other systems in the project besides the financial system includes: Extract verified technical attribute parameters related to the current expenditure record from the associated structured data. The step of comparing the technical attribute parameters extracted using natural language processing technology with the verified technical attribute parameters, and determining the final adopted technical attribute parameters according to preset data source credibility assessment rules, includes: If the comparison is successful, the technical attribute parameter will be automatically adopted and used as a training sample along with the corresponding item description text to update the natural language processing model. The record will no longer be pushed to the manual review interface.

5. The method for calculating carbon emissions from engineering projects based on financial expenditure data as described in claim 4, characterized in that, The step of obtaining at least one structured data source from systems other than the financial system in the project includes: Construct a data association mapping table that defines one or more potential association paths between key identifiers in financial expenditure records and key identifiers in the material requirements planning system, building information model, or bill of materials. Based on the data association mapping table, attempt to match one or more associated structured data sources for the current expenditure record, and generate an association confidence score for each matched data source; A source credibility score is predefined and dynamically maintained for each potential data source. The source credibility score is calculated and updated based on the consistency ratio between the data source and the human review results in historical cross-validation. The step of comparing the technical attribute parameters extracted using natural language processing technology with the verified technical attribute parameters includes: When making comparisons, priority is given to comparing the technical attribute parameters provided by data sources whose association confidence and source confidence scores are both higher than their respective thresholds; If multiple data sources meet the criteria and provide the same technical attribute parameters, then the parameter is adopted; if the parameters provided by multiple data sources conflict, then the parameter from the data source with the highest source credibility score is adopted.

6. The method for calculating carbon emissions from engineering projects based on financial expenditure data as described in claim 1, characterized in that, The process of converting the expenditure amount of each expenditure record into the physical quantity of the item based on the predefined mapping relationship between item categories and unit costs includes: Build and maintain a dynamic unit cost database that records the market unit cost of different item categories in different time intervals and / or different regions; Extract transaction time information and project location information from pending expenditure records; Based on the item category in the expenditure record, and combined with transaction time information and project location information, query the dynamic unit cost database to match and obtain a dynamic unit cost corresponding to a specific time and space. The calibrated physical quantity of the item is obtained by replacing the predefined static unit cost with the dynamic unit cost and dividing the expenditure amount by the dynamic unit cost.

7. The method for calculating carbon emissions from engineering projects based on financial expenditure data as described in claim 6, characterized in that, The dynamic unit cost database includes at least one authoritative market quotation platform data source and at least one actual procurement data source from within the project; Furthermore, the matching process for obtaining a spatiotemporally corresponding dynamic unit cost includes: Prioritize querying the actual procurement data source within the project. If an actual procurement unit price exists in the actual procurement data source that matches the item category, transaction time, and project region of the current expenditure record, then use that actual procurement unit price as the dynamic unit cost. If no corresponding actual purchase unit price is found in the actual procurement data source, the data source of the authoritative market quotation platform is queried to obtain the matching market unit cost, and the market unit cost is used as the dynamic unit cost.

8. The method for calculating carbon emissions of engineering projects based on financial expenditure data as described in claim 1, characterized in that, After summing the carbon emissions from all expenditure records to obtain the total carbon emissions of the project, the process also includes: Define an error distribution for the key input parameters in the computation chain, which shall include at least unit cost and / or carbon emission factor; Based on the Monte Carlo simulation method, the calculation process of all expenditure records of the project is repeated multiple times. In each calculation, key input parameters are randomly sampled according to their respective error distributions to obtain a series of simulated values ​​of total carbon emissions. Based on a series of simulated values ​​of total carbon emissions, the confidence interval and probability distribution of total carbon emissions are determined; When outputting the total carbon emissions of an engineering project, its confidence interval is also output.

9. The method for calculating carbon emissions of engineering projects based on financial expenditure data as described in claim 8, characterized in that, The calculation process for all expenditure records of the engineering project based on the Monte Carlo simulation method involves repeated calculations, including: Before performing a full Monte Carlo simulation, a global sensitivity analysis is performed to determine the contribution of each expenditure record to the uncertainty of total carbon emissions. Based on contribution, all expenditure records are divided into critical record sets and non-critical record sets; In subsequent Monte Carlo simulations, full-precision simulations are performed on the key record set, meaning that key input parameters are randomly sampled for each simulation. For non-critical record sets, an approximation strategy is adopted: a baseline carbon emission is calculated at the beginning of the simulation, and in subsequent simulations, this baseline value is multiplied by a global micro-perturbation factor to represent its fluctuation. The micro-perturbation factor is generated based on the overall fluctuation trend of the critical record set in the current simulation.

10. The method for calculating carbon emissions of engineering projects based on financial expenditure data as described in claim 1, characterized in that, After summing the carbon emissions from all expenditure records to obtain the total carbon emissions of the project, the process also includes: Multiple responsibility dimensions are predefined, and the responsibility dimensions include at least: the responsible party dimension, the engineering structure dimension, and the time phase dimension; A dimensional mapping rule base is constructed, which defines the mapping relationship between the feature information in expenditure records and each responsibility dimension; For each expenditure record, its attribution under each responsibility dimension is determined based on the feature information it contains through a dimensional mapping rule base. Based on attribution, the carbon emissions recorded in expenditures are allocated to the corresponding nodes under each responsibility dimension; Carbon emissions are aggregated according to responsibility dimensions to generate a multi-dimensional carbon emission attribution analysis report.

11. The method for calculating carbon emissions of engineering projects based on financial expenditure data as described in claim 10, characterized in that, The dimension mapping rule base supports dynamic configuration; for each expenditure record, based on its contained feature information, the dimension mapping rule base is used to determine its attribution under each responsibility dimension, including: When a single expenditure record has multiple possible attributions under a certain responsibility dimension, a weighted allocation strategy is adopted to split the carbon emissions of the expenditure record into multiple attribution nodes according to a preset weight ratio. Provide a rule conflict detection mechanism; when a conflict is detected in the mapping rules, prompt the user to optimize the rules. The allocation of carbon emissions from expenditure records to corresponding nodes under each responsibility dimension based on attribution relationships includes: It supports source tracing queries based on attribution results, enabling reverse lookup of all original expenditure records and their detailed calculation processes that constitute the aggregated carbon emissions from any dimension node.