Green low-carbon information recognition method and device based on recognition model and storage medium

Abstract

The application discloses a green low-carbon information identification method and device based on an identification model and a storage medium. The method comprises the following steps: firstly, historical material data in target products of distribution network type materials is collected, a data collection standard system is constructed, initial information is obtained, the initial information is preprocessed and subjected to semantic analysis, and a multi-source heterogeneous data knowledge base is built; secondly, key features of information in the multi-source heterogeneous data knowledge base are extracted and corrected to construct a bottom intelligent engine; thirdly, the bottom intelligent engine, a green low-carbon attribute index system and a retrieval database are integrated to generate a low-carbon information identification model; finally, target information is collected according to the data collection standard system and is input into the low-carbon information identification model for identification, so that the technical problem of low green low-carbon information identification accuracy is solved.

Description

Technical Field

[0001] This application relates to the interdisciplinary field of artificial intelligence and environmental protection technology for power systems, and in particular to a green and low-carbon information identification method, device and storage medium based on an identification model. Background Technology

[0002] The energy and manufacturing industries face significant challenges in their green and low-carbon transformation. As the core of the national energy system, the power industry, especially in the construction of power distribution networks, involves a large number of typical materials (such as cables, conductors, switches, transformers, etc.). These materials consume energy, materials, and carbon emissions throughout their entire life cycle, including production, transportation, use, and recycling.

[0003] Currently, research on carbon emissions and energy consumption identification of power materials in the industry mainly focuses on data statistics for single links or the construction of carbon accounting models. Some enterprises and research institutions manually collect energy and material consumption data provided by production enterprises and combine them with life cycle assessment methods to quantitatively analyze the carbon footprint of materials. At the same time, some existing technologies are attempting to establish carbon emission databases using information platforms to achieve centralized management of carbon data. In the field of green supply chains, some systems have introduced big data platforms to track suppliers' environmental qualifications and carbon reduction performance. In addition, some existing technologies are attempting to use natural language processing technology to perform semantic analysis on enterprise reports, standard texts, and other materials to extract relevant information on energy consumption or carbon emissions. However, due to the shortcomings of the above technologies, such as inconsistent data sources and collection standards, insufficient intelligent parsing capabilities for unstructured data, and the inability to openly search, complete, and authoritatively trace data, the technical problem of low accuracy in identifying green and low-carbon information has emerged. Summary of the Invention

[0004] In view of this, embodiments of this application provide a green and low-carbon information identification method, apparatus and storage medium based on an identification model, so as to at least solve the technical problem of low accuracy in green and low-carbon information identification.

[0005] According to one aspect of this application, a green and low-carbon information identification method based on an identification model is provided. The method includes: collecting historical material data in the target products of power distribution network material manufacturing, and constructing a data collection standard system by referring to the collected historical material data; acquiring initial data information related to power distribution network material manufacturing, and preprocessing and semantically analyzing the initial data information to construct a multi-source heterogeneous data knowledge base; extracting key features from the information in the multi-source heterogeneous data knowledge base, and correcting the extracted key features' extraction structure to construct a bottom-level intelligent engine using the corrected key features; constructing a green and low-carbon attribute index system, integrating the bottom-level intelligent engine and the green and low-carbon attribute index system, and accessing a retrieval database to obtain a low-carbon information identification model; and when identifying green and low-carbon information, collecting target information to be identified by referring to the data collection standard system, and inputting the target information into the low-carbon information identification model for identification.

[0006] Optionally, the extracted key features are subjected to extraction structure correction to construct an underlying intelligent engine using the corrected key features. This includes: setting an input flow set for the target product based on the extracted key features, wherein the input flow set includes at least the original input quantities of multiple input flows; mapping the original input quantities of multiple input flows based on the material and energy database of the target product to obtain equivalent quantities of multiple input flows, wherein the equivalent quantities are uniform quantities of the original input quantities under a preset dimension; constructing a constraint loss function based on the equivalent quantities of multiple input flows and the preset equivalent quantities; constructing a target optimization function based on the constraint loss function and the original input quantities of multiple input flows; and using the constraint loss function and the target optimization function to perform extraction structure correction on the extracted key features to construct an underlying intelligent engine using the corrected key features.

[0007] Optionally, a constrained loss function is constructed based on the equivalent quantities of multiple input flows and a preset equivalent quantity, using the following formula:

[0008] in, The constraint loss function is used to represent the degree of deviation between the weighted sum of the equivalent quantities of multiple input flows and the preset equivalent quantity; N is used to represent the number of input flows. Used to indicate the first The equivalent quantity of each input flow; Used to indicate the first The weighting coefficients of each input flow are used to reflect the relative contribution of different raw materials in the manufacturing process of the target product; Used to represent a preset equivalent quantity Used to represent the 2-norm.

[0009] Optionally, the objective optimization function is constructed based on the constrained loss function and the original input amounts of multiple input streams using the following formula:

[0010] in, Used to represent the first after being constrained by the constraint loss function. The amount of input into each input stream, Used to indicate the first The original input amount of each input stream, Used to represent the trade-off coefficient.

[0011] Optionally, a green and low-carbon attribute indicator system is constructed, including: acquiring multiple green and low-carbon attribute indicators related to the manufacturing of power distribution network materials; performing hierarchical processing on the multiple green and low-carbon attribute indicators to obtain primary attribute indicators and secondary attribute indicators, wherein the granularity of the primary attribute indicators is greater than that of the secondary attribute indicators; and constructing a green and low-carbon attribute indicator system based on the primary and secondary attribute indicators.

[0012] Optionally, the initial data information is preprocessed and semantically analyzed to construct a multi-source heterogeneous data knowledge base, including: cleaning the initial data information to obtain cleaned target data information, wherein the data noise of the target data information is less than that of the initial data information; obtaining the target file storing the target data information and the file type of the target file, and splitting the target file based on the file type to obtain multiple split files; and performing vectorization processing and semantic analysis on the multiple split files to construct a multi-source heterogeneous data knowledge base.

[0013] Optionally, key features are extracted from the information in the multi-source heterogeneous data knowledge base, including: obtaining the task name field, category field, usage field, and unit field in the target product of power distribution network material manufacturing, and setting an identifier field related to the task name field; constructing a prompt word template based on the task name field, category field, usage field, unit field, and identifier field; extracting key features from the information in the multi-source heterogeneous data knowledge base to obtain the extracted key features, and outputting the extracted key features according to the prompt word template.

[0014] According to another aspect of this application, a green and low-carbon information identification device based on an identification model is provided. The device includes: a data acquisition unit for acquiring historical material data in the target products of power distribution network material manufacturing, and constructing a data acquisition standard system based on the acquired historical material data; a first construction unit for acquiring initial data information related to power distribution network material manufacturing, and preprocessing and semantically analyzing the initial data information to construct a multi-source heterogeneous data knowledge base; a second construction unit for extracting key features from the information in the multi-source heterogeneous data knowledge base, and correcting the extraction structure of the extracted key features to construct a bottom-level intelligent engine using the corrected key features; an acquisition unit for constructing a green and low-carbon attribute indicator system, integrating the bottom-level intelligent engine and the green and low-carbon attribute indicator system, and accessing a retrieval database to obtain a low-carbon information identification model; and an identification unit for acquiring target information to be identified based on the data acquisition standard system during green and low-carbon information identification, and inputting the target information into the low-carbon information identification model for identification.

[0015] According to another aspect of this application, a storage medium is provided on which a computer program is stored, which, when executed by a processor, implements the above-described green and low-carbon information identification method based on the identification model.

[0016] According to another aspect of this application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described green and low-carbon information identification method based on the identification model.

[0017] Based on the above technical solutions, this application provides a green and low-carbon information identification method, device, and storage medium based on an identification model. The method includes: first, collecting historical material data in the target products of power distribution network materials manufacturing to construct a data collection standard system and obtain initial data information; preprocessing and semantically analyzing the relevant initial data information to build a multi-source heterogeneous data knowledge base; second, extracting and correcting key features from the information in the obtained multi-source heterogeneous data knowledge base to construct an underlying intelligent engine; third, integrating the underlying intelligent engine, the green and low-carbon attribute indicator system, and the retrieval database to generate a low-carbon information identification model; and finally, collecting target information according to the data collection standard system and inputting it into the low-carbon information identification model for identification. By constructing a data collection standard system to unify data specifications, building a multi-source heterogeneous data knowledge base to achieve semantic parsing of unstructured information, and introducing an extraction structure correction mechanism to eliminate model illusions and logical errors, a low-carbon information recognition model with dynamic recognition capabilities is finally generated. This low-carbon information recognition model is used to identify the target information to be identified in order to obtain green and low-carbon information, thereby solving the technical problem of low accuracy in green and low-carbon information recognition and achieving the technical effect of improving the accuracy of green and low-carbon information recognition.

[0018] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a green and low-carbon information identification method based on an identification model provided in an embodiment of this application is shown. Figure 2 This illustration shows a schematic diagram of the design structure of a prompt word template design method provided in an embodiment of this application; Figure 3 This illustration shows an output diagram of a structured output method provided in an embodiment of this application. Figure 4 This illustration shows a schematic diagram of a green and low-carbon information identification device based on an identification model, according to an embodiment of this application. Figure 5 A schematic diagram of the device structure of a computer device provided in an embodiment of this application is shown. Detailed Implementation

[0020] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0021] First, some of the nouns or terms that appear in the description of the embodiments of this application are explained as follows: A flow refers to an "object" that moves between process units within the boundaries of a product's lifecycle system, encompassing materials, energy, emissions, and waste. Examples include iron ore, copper ingots, electricity, polyethylene particles, and CO2 emissions. Each flow has a unique identifier (UUID), category, and unit of measurement in the database.

[0022] Process: refers to the operational unit that performs a specific function (such as raw material extraction, smelting, polymerization, etc.), and is the carrier of input and output of a flow. For example, "iron ore mining", "copper electrolytic refining", and "polyethylene polymerization" are typical processes.

[0023] The relationship between flow and process: Each process consists of a set of input flows (raw materials, energy, auxiliary materials, etc.) and output flows (main products, co-products, emissions, waste, etc.). The process defines the quantitative and qualitative transformation relationships between flows, constructing the causal chain of the life cycle.

[0024] In this embodiment, Figure 1 This illustration shows a flowchart of a green and low-carbon information identification method based on an identification model provided in an embodiment of this application. Figure 1 As shown, the method includes: Step S101: Collect historical material data in the target products of power distribution network material manufacturing, and construct a data collection standard system by referring to the collected historical material data.

[0025] In the technical solution provided in step S101 of this application, historical material data in the target products of power distribution network material manufacturing can be collected, and a data collection standard system can be constructed based on the obtained historical material data. The data collection standard system includes at least data field definitions, data format constraints, and unit unification rules.

[0026] Optionally, historical material data is used to characterize the basic attribute information, energy and material consumption types, and consumption data of upstream bulk materials (such as coal, oil, natural gas, electricity, iron, copper, lithium, plastics, etc.) and power distribution network products (such as transformers, distribution boxes, wires, cables, energy storage devices, etc.) in the manufacturing process of distribution network materials throughout their entire life cycle.

[0027] Table 1. Data Collection Fields Table

[0028] Table 2. Data Acquisition Fields for Process Data

[0029] Optionally, first define what the collected historical material information should include, and store this information in a structured and standardized format. Table 1 is the collection field table for flow data, and Table 2 is the collection field table for process data. Flow and process are the two core data types. The collection fields for flow data are shown in Table 1. The collection fields for flow data include: field name, element / attribute name, and field description. That is, when the field name is a unique identifier code, the corresponding element / attribute name is UUID, and the corresponding field description is a unique identifier code; when the field name is a flow classification, the corresponding element / attribute name is classification, and the corresponding field description is to specify the classification code and name; when the field name is a flow type, the corresponding element / attribute name is typeOfDataSet, and the corresponding field description is a single selection: product flow, waste flow, and basic flow; ...; when the field name is a flow attribute value, the corresponding element / attribute name is meanValue, and the corresponding field description is filled in as the baseline flow attribute, and other flow attributes are filled in with the conversion value of the baseline flow attribute. The collection fields for other flow data are shown in Table 1 and will not be listed here.

[0030] Furthermore, the data collection fields for process data are shown in Table 2. When the field name is flow value, the corresponding element / attribute name is meanValue. The field description can be the average flow value of the input or output, and only valid numbers should be specified. The data collection fields for other process data are shown in Table 2 and will not be listed here.

[0031] Table 3 Standard Product Flow Table

[0032] For example, when collecting information on major upstream materials related to typical power distribution network products, this material information includes, but is not limited to, transformers, distribution boxes, conductors, cables, and energy storage devices. Table 3 is a standard product flow table, specifying the energy and raw material products to be collected for bulk material categories such as coal, oil, natural gas, electricity, iron, copper, aluminum, and plastics. As shown in Table 3, when the serial number is 1 and the UUID is 9ff1d63b-2eab-4f82-969a-71dd1474f0f1, the corresponding name is "hard coal, anthracite; washing and processing; production blending, at a coal washing plant; 27.631 MJ / kg net calorific value," and the corresponding flow attribute is "mass, unit: kilogram (kg)." It should be noted that the other energy and raw material information required for this product is shown in Table 3 and will not be listed here.

[0033] Step S102: Obtain initial data information related to the manufacturing of distribution network materials, and perform preprocessing and semantic analysis on the initial data information to construct a multi-source heterogeneous data knowledge base.

[0034] In the technical solution provided by step S102 of this application, initial data information related to the manufacturing of distribution network materials can be obtained, and then the initial data information can be preprocessed and semantically analyzed to achieve the purpose of constructing a multi-source heterogeneous data knowledge base.

[0035] Optionally, after obtaining the initial data information, a target tool is used to preprocess the initial data information, and a preset embedding model (BERT) is used to perform semantic analysis on the preprocessed initial data information to convert unstructured data into structured data containing a unified field structure, thereby achieving the goal of building a multi-source heterogeneous data knowledge base.

[0036] For example, target tools are used to systematically clean and analyze the initial data information related to the manufacturing of typical power distribution network materials. Through data cleaning steps such as noise removal and format unification, the original messy data in the initial data information is effectively transformed into a structured format. Then, the BERT model is used to perform semantic analysis on the transformed initial data information to build a multi-source heterogeneous data knowledge base.

[0037] Step S103: Extract key features from information in the multi-source heterogeneous data knowledge base, and perform extraction structure correction on the extracted key features in order to build an underlying intelligent engine using the corrected key features.

[0038] In the technical solution provided in step S103 of this application, key features are extracted from the information in the obtained multi-source heterogeneous data knowledge base, and then the extracted key features are structurally corrected to utilize the corrected key features and a large language model to build an underlying intelligent engine. The key features can be referred to as structured feature data, and at least include a task name field, a category field, a usage field, and a unit field.

[0039] For example, using historical data and domain-specific technical literature as the basic corpus, and leveraging the semantic analysis capabilities of a large language model, key features are extracted from the original text of a multi-source heterogeneous data knowledge base to obtain the extraction results.

[0040] It should be noted that the above process can not only accurately identify the type, quantity, and attributes of energy and raw materials such as coal, oil, natural gas, electricity, iron, copper, zinc, and plastics, but also extract fine-grained information such as relevant process paths, product categories, and consumption links by combining contextual understanding. Driven by engineering and high-quality corpora, the large model can automatically complete tasks such as terminology standardization, unit conversion, and attribute normalization, greatly improving the efficiency and accuracy of structured data collection.

[0041] Furthermore, the key features in the above extraction results are subjected to extraction structure correction in order to utilize the corrected key features to achieve the goal of building the underlying intelligent engine.

[0042] Step S104: Construct a green and low-carbon attribute indicator system, integrate the underlying intelligent engine and the green and low-carbon attribute indicator system, and connect to the retrieval database to obtain a low-carbon information identification model.

[0043] In the technical solution provided in step S104 of this application, a green and low-carbon attribute indicator system can be constructed, and then the underlying intelligent engine and the green and low-carbon attribute indicator system obtained above can be integrated and connected to the retrieval database to realize the acquisition of a low-carbon information identification model.

[0044] Optionally, multiple green and low-carbon attribute indicators related to the manufacturing of power distribution network materials can be obtained; based on these multiple green and low-carbon attribute indicators, a green and low-carbon attribute indicator system can be constructed. Among them, the multiple green and low-carbon attribute indicators include at least: energy consumption, carbon footprint, resource utilization efficiency, environmental impact indicators, and product recyclability / reusability, etc.

[0045] Optionally, Energy Consumption: This covers the total energy consumption (electricity, gas, steam, etc.) and unit energy consumption during the product's production, transportation, operation, and recycling stages. Carbon Footprint: Based on the Life Cycle Assessment (LCA) method, this quantifies the total greenhouse gas emissions (in CO2-eq) per unit of product throughout its entire life cycle. Resource Efficiency: This includes dimensions such as raw material utilization rate, proportion of recycled materials, unit product energy consumption, and carbon emission intensity. Environmental Impact: This includes solid waste generation, water resource consumption, and the control status of hazardous substances in the environment. Product Recyclability / Reusability: This includes indicators such as the recyclability of packaging materials and the recyclable proportion of main materials.

[0046] Optionally, a green and low-carbon attribute indicator system is first constructed to establish standardized identification dimensions. Then, an underlying intelligent engine with data extraction and correction capabilities is integrated with this indicator system and introduced into a retrieval database. Finally, the three work together to form a complete low-carbon information identification model. The model can automatically map the discrete data extracted by the underlying engine to the standardized indicator system, realizing the transformation of unstructured text into structured data. At the same time, with the access of the retrieval database, the model has the ability to actively retrieve and complete missing key data, thereby ensuring the completeness, standardization, and accuracy of the final identification results.

[0047] Step S105: When identifying green and low-carbon information, collect the target information to be identified in accordance with the data collection standard system, and input the target information into the low-carbon information identification model for identification.

[0048] In the technical solution provided in step S105 of this application, target information to be identified is collected according to the data acquisition standard system, and then the target information is input into the low-carbon information identification model for analysis and identification, so as to achieve the purpose of obtaining green and low-carbon information. The low-carbon information identification model can be called a low-carbon attribute intelligent identification model.

[0049] Optionally, based on the data collection standard system, low-carbon information identification model, and retrieval database, an identification system (AI identification system) is generated to identify green and low-carbon information in the target information. For example, by using a generative AI identification system that integrates a large language model (semantic parsing model), a retrieval database, a green attribute indicator system, and a web retrieval module, a generative AI identification system can be obtained to intelligently identify the green and low-carbon information of input distribution network materials.

[0050] Optionally, the AI ​​recognition system possesses powerful capabilities for fusing multi-source data across documents, formats, and lifecycle stages. The model supports intelligent merging, deduplication, and consistency comparison of green and low-carbon attribute indicators for the same product at different stages, from different companies, and from different sources. It automatically detects outliers and potential data conflicts, improving the quality of collected data and the scientific rigor of industry comparative analysis.

[0051] Optionally, the low-carbon information identification model deeply integrates technologies such as big data analysis, automatic web retrieval, and natural language processing (NLP) to achieve batch collection, intelligent analysis, and efficient aggregation of multi-source heterogeneous data throughout the entire lifecycle of distribution network materials. The system integrates data sources covering enterprise annual reports, industry databases, technical standard documents, and third-party evaluation reports, significantly improving the completeness, accuracy, and real-time nature of green and low-carbon attribute data.

[0052] Optionally, the model enhances information retrieval capabilities based on open networks. Through the target interface, it can automatically access and retrieve external resources such as public databases, official corporate websites, and industry information platforms, automatically supplementing local data with green and low-carbon attribute indicators not yet covered. For each retrieved piece of information, the system simultaneously labels the authoritative data source (such as standard document numbers, industry reports, third-party certification numbers, etc.) to ensure information compliance and traceability. The function calling in this process supports on-demand customization of the search scope and field mapping rules, possessing excellent scalability and flexibility.

[0053] Optionally, the low-carbon attribute intelligent identification model employs NLP semantic analysis and deep learning algorithms to uniformly parse structured data (such as tabular databases (Excel databases), statistical tables) and unstructured data (such as reports in target formats (PDF reports), web page text, scanned documents, etc.) uploaded by enterprises. The AI ​​identification system can not only automatically identify core green attributes such as energy consumption (electricity, gas, coal), greenhouse gas emissions, raw material sources and their renewability, and material recycling rates, but also automatically extract related information such as product models, functional units, and evaluation stages, and standardize, archive, number, and store all attributes. Function calls for related prompt word templates support one-click batch collection and automatic completion of historical data.

[0054] For example, firstly, referring to the data collection standard system, the energy consumption and raw material consumption data of its production process are collected as target information. This information is then input into a low-carbon information identification model. The model uses its underlying engine to automatically identify and extract key data such as "electricity," "copper," and "polyethylene," standardizing them into a UUID|name|unit|usage format. Simultaneously, it combines an indicator system to collect attributes such as "carbon footprint," and uses online searches to complete missing information, ultimately outputting structured green and low-carbon attribute identification results, i.e., green and low-carbon information.

[0055] For example, the identification system, based on a function call mechanism, can automatically retrieve and aggregate green and low-carbon attribute information for "cable" products. After a request is sent, the system retrieves, standardizes, and structures the various green attribute indicators of the target product based on publicly available online information, industry standards, and authoritative third-party reports. All results are derived from authoritative data sources and presented in clear and intuitive tabular form, facilitating direct use by enterprises in practical scenarios such as green procurement and compliance assessments. Furthermore, the system can continuously improve its ability to identify and aggregate green attributes by accumulating high-confidence samples and automatically labeling results. In addition, the identification system can adjust data mapping relationships and prompt word templates according to the latest policies, regulations, standards, and advancements in green and low-carbon technologies, enhancing the applicability and scalability of the evaluation system.

[0056] In the technical solutions provided by steps S101 to S105 of this application, firstly, historical material data in the target products of power distribution network materials are collected to construct a data collection standard system and obtain initial data information. Preprocessing and semantic analysis of the relevant initial data information are then performed to build a multi-source heterogeneous data knowledge base. Secondly, key features are extracted and corrected from the information in the obtained multi-source heterogeneous data knowledge base to construct an underlying intelligent engine. Thirdly, the underlying intelligent engine, the green and low-carbon attribute indicator system, and the retrieval database are integrated to generate a low-carbon information identification model. Finally, target information is collected according to the data collection standard system and input into the low-carbon information identification model for identification. By constructing a data collection standard system to unify data specifications, building a multi-source heterogeneous data knowledge base to achieve semantic parsing of unstructured information, and introducing an extraction structure correction mechanism to eliminate model illusions and logical errors, a low-carbon information recognition model with dynamic recognition capabilities is finally generated. This low-carbon information recognition model is used to identify the target information to be identified in order to obtain green and low-carbon information, thereby solving the technical problem of low accuracy in green and low-carbon information recognition and achieving the technical effect of improving the accuracy of green and low-carbon information recognition.

[0057] The method described in this embodiment will be further described below.

[0058] As an optional implementation, the extracted key features are subjected to extraction structure correction to construct an underlying intelligent engine using the corrected key features. This includes: setting an input flow set for the target product based on the extracted key features, wherein the input flow set includes at least the original input quantities of multiple input flows; mapping the original input quantities of multiple input flows based on the target product's material and energy database to obtain equivalent quantities of multiple input flows, wherein the equivalent quantities are uniform quantities of the original input quantities under a preset dimension; constructing a constraint loss function based on the equivalent quantities of multiple input flows and the preset equivalent quantities; constructing a target optimization function based on the constraint loss function and the original input quantities of multiple input flows; and using the constraint loss function and the target optimization function to perform extraction structure correction on the extracted key features to construct an underlying intelligent engine using the corrected key features.

[0059] In this embodiment, based on the extracted key features, a set of input flows for the target product is first established. Then, based on the material and energy database of the target product, the original input quantities of multiple input flows are mapped to obtain equivalent quantities for multiple input flows. Next, a constraint loss function is constructed based on the equivalent quantities of multiple input flows and a preset equivalent quantity. Finally, a target optimization function is constructed based on the constraint loss function and the original input quantities of multiple input flows. Finally, using the constraint loss function and the target optimization function, the extracted key features can be structurally corrected, enabling the construction of a lower-level intelligent engine using the corrected key features. The constraint loss function can be called a consistency constraint loss function. The target optimization function can be called a joint optimization objective function. The input flow set also includes the units of measurement corresponding to the original input quantities of multiple input flows.

[0060] Optionally, after completing the structured extraction of energy consumption and material consumption information based on the large language model, in order to avoid problems such as abnormal usage, cross-document inconsistency, or physical quantity inconsistency caused by relying solely on semantic recognition, a consistency constraint mechanism is introduced to perform secondary correction on the extraction results (key features after extraction).

[0061] Optionally, for the extracted key features, let the set of input flows obtained by the model in identifying the target product be represented as: ,in, This is used to represent the set of input flows identified for the target product within the current functional unit. Used to indicate the first The index identifier of the input stream, Used to represent the usage value of the corresponding input stream. The unit of measurement used to represent the amount used includes energy inputs and raw material inputs, and both use the same functional unit as the statistical benchmark.

[0062] Optionally, based on the target product's material and energy database, an equivalent mapping function is used to map the original input quantities of multiple input streams, obtaining equivalent quantities for each input stream. The material and energy database can be referred to as a pre-built knowledge base of material and energy attributes. For example, to achieve comparability and unified representation between different types of input streams, based on the pre-built knowledge base of material and energy attributes, equivalent benchmark quantity mapping is performed on each input stream to obtain equivalent quantities under a unified dimension. This process can be represented by the following formula:

[0063] in, Used to indicate the first The equivalent quantity of an input flow under a unified benchmark dimension; Used to represent equivalent mapping functions, which map input streams to uniform energy equivalents, mass equivalents, or carbon equivalents based on their quantity, unit, and attribute parameters. This is used to represent the set of attribute parameters corresponding to the input flow, and the attribute parameters in this set of attribute parameters include, but are not limited to, lower heating value, carbon emission factor, material density, energy conversion factor or unit conversion factor.

[0064] As an optional implementation method, a constraint loss function is constructed based on the equivalent quantities of multiple input flows and a preset equivalent quantity using the following formula, including:

[0065] in, The constraint loss function is used to represent the degree of deviation between the weighted sum of the equivalent quantities of multiple input flows and the preset equivalent quantity; N is used to represent the number of input flows. Used to indicate the first The equivalent quantity of each input flow; Used to indicate the first The weighting coefficients of each input flow are used to reflect the relative contribution of different raw materials in the manufacturing process of the target product; Used to represent a preset equivalent quantity Used to represent the 2-norm.

[0066] In this embodiment, a constraint loss function is constructed using the equivalent quantities of multiple input flows and a preset equivalent quantity, and the constraint loss function can be expressed by the following formula:

[0067] in, It is used to represent a preset equivalent quantity, or it can be used to represent the theoretical reference value of the target product under the current functional unit.

[0068] Optionally, the equivalent quantities of multiple input flows under a unified benchmark dimension are weighted and summed, and the deviation between the weighted summation result and the preset equivalent quantity is used as a consistency constraint index to characterize the degree of deviation of the material or energy balance of the target product under a functional unit.

[0069] Optionally, before constructing the constraint loss function using the preset equivalent quantity, an original preset equivalent quantity is determined based on multiple historical preset equivalent quantities; in response to the original preset equivalent quantity being greater than the first preset equivalent quantity and less than the second preset equivalent quantity, the original preset equivalent quantity is determined as the preset equivalent quantity, wherein the first preset equivalent quantity is less than the second preset equivalent quantity.

[0070] Furthermore, the first presupposed equivalent quantity can be called the minimum reasonable equivalent input quantity, through... This is expressed as follows. The second preset equivalent quantity can be called the maximum reasonable equivalent input quantity, which is expressed through... This can be represented. The original preset equivalent quantity can be expressed through... To express.

[0071] For example, after obtaining the equivalent quantities of each input stream, a reference constraint interval is constructed for the original preset equivalent quantity of the target product under the current functional unit. When the original preset equivalent quantity Within the reference constraint interval, the original preset equivalent quantity will be... Determined as the preset equivalent quantity .in, This is used to represent the minimum reasonable equivalent input for the target product under the corresponding manufacturing process path. This is used to represent the maximum reasonable equivalent input for the target product under the corresponding manufacturing process path. The reference constraint range can be determined based on historical sample data, industry statistics, standards and specifications, or knowledge base rules.

[0072] As an optional implementation method, the objective optimization function is constructed based on the constraint loss function and the original input amounts of multiple input streams using the following formula, including:

[0073] in, Used to represent the first after being constrained by the constraint loss function. The amount of input into each input stream, Used to indicate the first The original input amount of each input stream, Used to represent the trade-off coefficient.

[0074] In this embodiment, the objective optimization function is constructed based on the obtained constraint loss function and the original input amounts of multiple input flows. The constraint loss function, also known as the consistency constraint function, characterizes the material or energy balance relationship of each input flow within the target product functional unit.

[0075] Optionally, while keeping the original input amount of the input stream relatively stable, a consistency constraint function is introduced to optimize and adjust the input amount of the input stream in order to achieve a balance between the semantic recognition result and the material conservation or energy conservation constraint.

[0076] For example, to maintain the stability of semantic recognition results of large language models while satisfying physical consistency constraints under functional units, a joint optimization objective function can be constructed. By solving the joint optimization objective function, the input flow usage after consistency constraint correction can be obtained, which can then be used to adaptively correct the extraction results, resulting in the extraction structure correction result, the expression of which is:

[0077] in, Used to represent the first after correction by the consistency constraint loss function Each input flow usage, Used to indicate the first The original input amount of each input stream has two terms: the first term constrains the deviation between the correction result and the original extraction result to avoid excessive deviation from the semantic recognition result, and the second term introduces a physical consistency constraint under the functional unit. It is used to represent the trade-off coefficient, which is used to adjust the balance between semantic stability and physical consistency.

[0078] In this embodiment, based on the functional unit corresponding to the target product, the extracted input streams are mapped to equivalent benchmark quantities, and a consistency constraint loss function is constructed based on a preset theoretical reference interval. A joint optimization objective is constructed to minimize the deviation between the extraction result and the original semantic recognition result and the consistency constraint function. The extracted energy consumption and material consumption are adaptively corrected to obtain a structured recognition result that satisfies the consistency constraint under the functional unit.

[0079] As an optional implementation method, a green and low-carbon attribute indicator system is constructed, including: acquiring multiple green and low-carbon attribute indicators related to the manufacturing of power distribution network materials; performing hierarchical processing on the multiple green and low-carbon attribute indicators to obtain primary attribute indicators and secondary attribute indicators, wherein the granularity of the primary attribute indicators is larger than that of the secondary attribute indicators; and constructing a green and low-carbon attribute indicator system based on the primary attribute indicators and the secondary attribute indicators.

[0080] In this embodiment, after obtaining multiple green and low-carbon attribute indicators related to the manufacturing of power distribution network materials, the indicators can be hierarchically processed using their semantic categories and evaluation objects to obtain primary and secondary attribute indicators. Then, a green and low-carbon attribute indicator system is constructed based on these primary and secondary indicator systems. The primary attribute indicators can be referred to as primary attributes, and the secondary attribute indicators as secondary attributes.

[0081] For example, after obtaining multiple green and low-carbon attribute indicators, these indicators were stratified and clearly divided into primary and secondary attributes to improve the systematicness and application flexibility of the indicator system. Table 4 is the classification table of green and low-carbon attribute indicators. As shown in Table 4, primary attributes include: environmental protection, health and safety, energy conservation and consumption reduction, and resource conservation. Secondary attributes include: packaging and packaging materials, electromagnetic radiation, power consumption, raw material conservation, solid waste, environmental risk, recycling, carbon footprint declaration and labeling, minimum requirements for recyclable content, recycling rate and material recycling efficiency, traceability of key material sources, limits for hazardous substances, noise, health and safety (health exposure, occupational safety, etc.), limits for hazardous substances, other energy-saving measures, energy efficiency and energy density, raw material conservation, recycling, minimum requirements for recyclable content, recycling rate and material recycling efficiency, traceability of key material sources, and product durability and performance.

[0082] Furthermore, this indicator system, based on the core principles of quantification and standardization, relies on actual data tables and typical parameters to evaluate the green and low-carbon attributes of materials throughout the entire process of production, circulation, use, and disposal. This evaluation framework not only facilitates data collection and horizontal comparison but also provides a solid scientific foundation for the subsequent construction of data labels for artificial intelligence models and the training of recognition algorithms, thereby ensuring that the intelligent identification and evaluation process of green and low-carbon attributes is highly standardized and operable.

[0083] Table 4. Classification of Green and Low-Carbon Attribute Indicators

[0084] As an optional implementation method, the initial data information is preprocessed and semantically analyzed to construct a multi-source heterogeneous data knowledge base, including: cleaning the initial data information to obtain cleaned target data information, wherein the data noise of the target data information is less than that of the initial data information; obtaining the target file storing the target data information and the file type of the target file, and splitting the target file based on the file type to obtain multiple split files; and performing vectorization processing and semantic analysis on the multiple split files to construct a multi-source heterogeneous data knowledge base.

[0085] In this embodiment, the initial data information is first cleaned to obtain the cleaned target data information. Then, the target file storing the target data information and the file type of the target file are obtained. Then, the target file is split according to the file type of the target file to obtain multiple split files. Finally, the multiple split files can be vectorized and semantically analyzed to achieve the purpose of constructing a multi-source heterogeneous data knowledge base.

[0086] Optionally, the file types include at least: word processing document (Word document) format and portable document (PDF document) format. The target data information includes at least: target text information, target table information, and image text information.

[0087] Optionally, noise removal is performed on the text information in the initial data to obtain the first text information; the first text information is then processed to unify its format to obtain the target text information. For example, through steps such as noise removal and format unification, the original messy data in the initial data is transformed into a structured format.

[0088] Optionally, the target parser is used to denoise the tabular data in the initial data information to obtain the target table information, wherein the table data is precisely processed by the target parser (HTML parser).

[0089] Optionally, optical character recognition (OCR) technology can be used to extract image text information from the image information in the initial data information. For example, text content in an image can be extracted using OCR technology.

[0090] Optionally, a document segmentation tool (partition_docx) is used to extract document information from the text processing document, and a document chunking tool (chunk_by_title) is used to segment the document information into chunks according to the title information of the text processing document, resulting in multiple first split files to ensure that the title and content are correctly associated. These multiple first split files are saved as first format (.pkl format) files and second format (.txt format) files. A first preset embedding model is used to vectorize the first and second format files, resulting in first and second vector files. Third and fourth vector files, with the target file type being Portable Documents format, are obtained. Simultaneously, the first, second, third, and fourth vector files are uploaded to a vector database. A second preset embedding model (Transformer model) is used to perform semantic analysis on the first, second, third, and fourth vector files to extract key indicators related to material consumption and energy consumption. These key indicators are then used to construct a multi-source heterogeneous data knowledge base.

[0091] For example, the elements in the `.docx` file are first extracted using `partition_docx`, and then these elements are processed in chunks by title using `chunk_by_title`. The extracted text chunks are merged by title to generate a series of text strings, while compound elements and table content are processed simultaneously. By merging consecutive title chunks, the correct association between the title and its corresponding content is ensured, and the processed text is saved as a string. The processed text files are stored in `.pkl` and `.txt` formats in the first directory (`results_pickle`) and the second directory (`results_txt`), respectively. Parallel processing of multiple files is achieved using the parallel processing function (`ProcessPoolExecutor`) in the software, and performance is improved by using multiple Central Processing Units (CPUs) to effectively process a large number of `.docx` files.

[0092] For another example, the code extracts all records from a relational database and filters out those matching the `.docx` filename. For each file to be processed, the code loads the corresponding `.pkl` file data and performs text processing through a series of functions, including splitting tables, fixing encoding issues in computer encoding (UTF-8), and generating embedding vectors. The processed vector data is then uploaded to a vector database. To improve the program's reliability and fault tolerance, the code utilizes a retry library (tenacity library) in multiple steps to implement a retry mechanism, ensuring automatic retries in case of network fluctuations or other anomalies, preventing program interruption. The code also uses a standard logging module to record detailed information for each step, including success and error handling, facilitating debugging and maintenance.

[0093] Optionally, when the target file is in Portable Document format, the content in the Portable Document is parsed using a parsing function (sci_chunk function) to obtain a parsed file; the parsed file is then divided into chunks according to the title information of the Portable Document to obtain multiple second split files; the multiple second split files are mapped using a parallel processing mechanism, and the mapped multiple second split files are stored as a third format file and a fourth format file; the third format file and the fourth format file are vectorized using a first preset embedding model to obtain a third vector file and a fourth vector file.

[0094] For example, a series of helper functions are defined, including checking for special text, calculating the number of tokens in a string, and fixing UTF-8 encoding issues. The `sci_chunk` function parses the PDF file content by calling the document element splitting function (the `partition` function), filters out the necessary elements, and then splits it into chunks by title. Each chunk undergoes cleanup processing, and optional visual completion is performed on the images. The processed text chunks are ultimately stored as title-content pairs. Next, the code defines a conversion function (the `split_chunks` function) to convert the chunked content into a dictionary list, as well as functions for processing web page (HTML) tables and merging chunk lists. The logic processing function (the `process_pdf` function) is responsible for the specific PDF file processing logic, calling the `sci_chunk` function to extract and process the text, and then saving the results as `.pkl` and `.txt` files respectively. The processed files are stored in the `results_pickle` and `results_txt` directories for later use and management. Finally, the code uses ProcessPoolExecutor to process multiple PDF files in parallel, leveraging multiple CPU cores to improve performance and effectively handle a large number of PDF files. After processing, the program prints a message indicating successful data insertion.

[0095] Furthermore, the program extracts records from a relational database, processes the corresponding 'PDF' files, and stores the results in a vector database. By loading environment variables to obtain the crucial interface key and database path (URL), and initializing the necessary clients and indexes, the program can efficiently process and embed document data. First, the code extracts all records from the relational database and filters out those not yet embedded. For each file to be processed, the code loads the corresponding .pkl file data and performs text processing through a series of functions, including table splitting, fixing UTF-8 encoding issues, and generating embedding vectors. The processed vector data is finally uploaded to the vector database. To improve the program's reliability and fault tolerance, the code uses the tenacity library in several steps to implement a retry mechanism. For example, in the two critical steps of embedding vector generation and vector data upload, the code automatically retryes in case of network fluctuations or other anomalies by setting the number of retries and a fixed waiting time to avoid program interruption. In addition, the code uses the logging module to record detailed information for each step, including success and error handling, facilitating debugging and maintenance. The log file records the start and end times of each processing step, the file identifier, and any potential error messages. Depending on the filename, the code differentiates between cases containing page numbers and those without. For .pdf files containing page numbers, the code extracts the page numbers from each text block and retains this information when generating the embedding vector for more precise text content location during retrieval. For files without page numbers, the code processes the text blocks directly. During text processing, the code performs necessary text cleaning and table segmentation to ensure the accuracy and consistency of the generated embedding vectors. The processed vector data, including text content and relevant metadata such as file identifiers and course information, is then uploaded to the vector database for subsequent retrieval and analysis.

[0096] As an optional implementation method, key feature extraction is performed on the information in the multi-source heterogeneous data knowledge base, including: obtaining the task name field, category field, usage field, and unit field in the target product of power distribution network material manufacturing, and setting an identifier field related to the task name field; constructing a prompt word template based on the task name field, category field, usage field, unit field, and identifier field; extracting key features from the information in the multi-source heterogeneous data knowledge base to obtain the extracted key features, and outputting the extracted key features according to the prompt word template.

[0097] In this embodiment, the task name, category, usage, and unit fields of the target product in the distribution network material manufacturing process are first obtained. Simultaneously, an identifier field related to the task name field is set. Then, based on the obtained task name, category, usage, unit, and identifier fields, a prompt word template is constructed. Finally, key features are extracted from the information in the multi-source heterogeneous data knowledge base to obtain the extracted key features. The extracted key features are then output according to the aforementioned prompt word template. Here, the task name field is abbreviated as "name," the category field as "category," the usage field as "usage," and the unit field as "unit." The prompt word template is used to constrain the order and meaning of the output fields.

[0098] Optionally, based on the product carbon footprint report text, automatically extract the main energy and raw materials consumed in the production process of the target product. The output should be structured, including the name, category (energy / raw material), unit, and quantity for each item, ignoring other information. The output structure should be: UUID|Name|Unit|Quantity. For example, taking the production process of "1 kilometer (km) of power cable" as an example... Figure 2 This illustration shows a schematic diagram of the design structure of a prompt word template design method provided in an embodiment of this application. Figure 2 The image shows the Prompt template input to the large language model. Tables 1, 2, and 3 mentioned above can be imported into the large language model through the Application Programming Interface (API) or an automated form mapping tool to achieve automatic synchronization and updates of the content.

[0099] Furthermore, at the level of model integration and functional implementation, based on the collaborative support of semantic parsing models, prompt word templates, and domain knowledge bases, the system can automatically identify and classify energy consumption and material consumption information categories. The system utilizes structured prompt word templates, combined with a product carbon footprint data dictionary and classification tags, to parse carbon footprint report text, standardizing and tagging the extraction and aggregation of energy and raw material consumption information.

[0100] For example, taking the extraction of data from the "1km power cable" production process as an example, based on a semantic parsing model, prompt word templates, and a knowledge base, the model automatically identifies and extracts major inputs such as "electricity," "steam," "aluminum rods," "copper," "polyethylene," "steel strips," and "polyvinyl chloride (PVC)," assigning a unique identifier (UUID) to each data item. The model can automatically determine the input attributes based on the context, accurately distinguishing input / output, units of measurement (such as kg, kilowatt-hours (kWh), m³), ​​and corresponding production stages. For instance, the model automatically identifies "electricity consumption: 362 kWh / km" as "electricity (energy)," classifies it in the energy list using "kWh," and associates it with the corresponding UUID. Furthermore, for ambiguous descriptions within the enterprise such as "AC / DC power" and "refined copper / copper rods," the model can automatically standardize names and complete attributes (such as form, use, purity, etc.) based on the context and knowledge base entries, effectively eliminating terminological ambiguity.

[0101] In the embodiments of this application, Figure 3 This illustration shows an output diagram of a structured output method provided in an embodiment of this application, such as... Figure 3 As shown, all identified inputs are output in a structured table, including information such as UUID, name, category, usage, and unit, facilitating subsequent database access and traceability. For raw materials whose UUIDs cannot be found in the knowledge base (such as "steel strip" or "low-smoke halogen-free sheathing material"), the model will automatically mark them as "to be completed" or "manually processed" to ensure data consistency and integrity.

[0102] By applying the technical solution of this embodiment, firstly, historical material data in the manufacturing target products of power distribution network materials is collected to construct a data collection standard system and obtain initial data information. This initial data information is then preprocessed and semantically analyzed to build a multi-source heterogeneous data knowledge base. Secondly, key features are extracted and corrected from the information in the obtained multi-source heterogeneous data knowledge base to construct an underlying intelligent engine. Thirdly, the underlying intelligent engine, green and low-carbon attribute indicator system, and retrieval database are integrated to generate a low-carbon information recognition model. Finally, target information is collected according to the data collection standard system and input into the low-carbon information recognition model for recognition. Considering the need to unify data specifications by constructing a data collection standard system, build a multi-source heterogeneous data knowledge base to achieve semantic parsing of unstructured information, and introduce an extraction structure correction mechanism to eliminate model illusions and logical errors, a low-carbon information recognition model with dynamic recognition capabilities is ultimately generated. This model is used to identify the target information to be identified, thereby obtaining green and low-carbon information. This solves the technical problem of low accuracy in green and low-carbon information recognition and achieves the technical effect of improving the accuracy of green and low-carbon information recognition.

[0103] Furthermore, as Figure 1 In the specific implementation of the method, in the embodiments of this application, Figure 4 This illustration shows a structural schematic diagram of a green and low-carbon information identification device based on an identification model, as provided in an embodiment of this application. Figure 4 As shown, the green and low-carbon information identification device 400 based on the identification model includes: a collection unit 401, a first construction unit 402, a second construction unit 403, an acquisition unit 404, and an identification unit 405.

[0104] The data acquisition unit 401 is used to collect historical material data in the target products of power distribution network material manufacturing, and to construct a data acquisition standard system by referring to the collected historical material data.

[0105] The first construction unit 402 is used to acquire initial data information related to the manufacturing of distribution network materials, and to preprocess and semantically analyze the initial data information in order to construct a multi-source heterogeneous data knowledge base.

[0106] The second building unit 403 is used to extract key features from information in a multi-source heterogeneous data knowledge base and to perform extraction structure correction on the extracted key features in order to build an underlying intelligent engine using the corrected key features.

[0107] The acquisition unit 404 is used to construct a green and low-carbon attribute indicator system, integrate the underlying intelligent engine and the green and low-carbon attribute indicator system, and access the retrieval database to obtain a low-carbon information identification model.

[0108] The identification unit 405 is used to collect target information to be identified in accordance with the data collection standard system when identifying green and low-carbon information, and to input the target information into the low-carbon information identification model for identification.

[0109] Optionally, the device is further configured to: perform extraction structure correction on the extracted key features to construct an underlying intelligent engine using the corrected key features, including: setting an input flow set for the target product based on the extracted key features, wherein the input flow set includes at least the original input quantities of multiple input flows; mapping the original input quantities of multiple input flows based on the material and energy database of the target product to obtain equivalent quantities of multiple input flows, wherein the equivalent quantities are uniform quantities of the original input quantities under a preset dimension; constructing a constraint loss function based on the equivalent quantities of multiple input flows and the preset equivalent quantities; constructing a target optimization function based on the constraint loss function and the original input quantities of multiple input flows; and performing extraction structure correction on the extracted key features using the constraint loss function and the target optimization function to construct an underlying intelligent engine using the corrected key features.

[0110] Optionally, the device is also used to: construct a constrained loss function based on the equivalent quantities of multiple input flows and a preset equivalent quantity using the following formula, including:

[0111] in, The constraint loss function is used to represent the degree of deviation between the weighted sum of the equivalent quantities of multiple input flows and the preset equivalent quantity; N is used to represent the number of input flows. Used to indicate the first The equivalent quantity of each input flow; Used to indicate the first The weighting coefficients of each input flow are used to reflect the relative contribution of different raw materials in the manufacturing process of the target product; Used to represent a preset equivalent quantity Used to represent the 2-norm.

[0112] Optionally, the device is also used to: construct an objective optimization function based on the constraint loss function and the original input amounts of multiple input streams using the following formula, including:

[0113] in, Used to represent the first after being constrained by the constraint loss function. The amount of input into each input stream, Used to indicate the first The original input amount of each input stream, Used to represent the trade-off coefficient.

[0114] Optionally, the device is also used to: construct a green and low-carbon attribute indicator system, including: acquiring multiple green and low-carbon attribute indicators related to the manufacturing of power distribution network materials; performing hierarchical processing on the multiple green and low-carbon attribute indicators to obtain primary attribute indicators and secondary attribute indicators, wherein the granularity of the primary attribute indicators is greater than that of the secondary attribute indicators; and constructing a green and low-carbon attribute indicator system based on the primary attribute indicators and the secondary attribute indicators.

[0115] Optionally, the device is also used to: preprocess and semantically analyze the initial data information to construct a multi-source heterogeneous data knowledge base, including: cleaning the initial data information to obtain cleaned target data information, wherein the data noise of the target data information is less than that of the initial data information; acquiring the target file storing the target data information and the file type of the target file, and splitting the target file based on the file type to obtain multiple split files; and performing vectorization processing and semantic analysis on the multiple split files to construct a multi-source heterogeneous data knowledge base.

[0116] Optionally, the device is also used to: extract key features from information in a multi-source heterogeneous data knowledge base, including: obtaining the task name field, category field, usage field, and unit field in the target product of power distribution network material manufacturing, and setting an identifier field related to the task name field; constructing a prompt word template based on the task name field, category field, usage field, unit field, and identifier field; extracting key features from information in the multi-source heterogeneous data knowledge base to obtain the extracted key features, and outputting the extracted key features according to the prompt word template.

[0117] In this embodiment, firstly, historical material data from the target products manufactured using distribution network materials is collected to construct a data collection standard system and obtain initial data information. This initial data information is then preprocessed and semantically analyzed to build a multi-source heterogeneous data knowledge base. Secondly, key features are extracted and corrected from the information in the obtained multi-source heterogeneous data knowledge base to construct an underlying intelligent engine. Thirdly, the underlying intelligent engine, green and low-carbon attribute indicator system, and retrieval database are integrated to generate a low-carbon information recognition model. Finally, target information is collected according to the data collection standard system and input into the low-carbon information recognition model for identification. By constructing a data collection standard system to unify data specifications, building a multi-source heterogeneous data knowledge base to achieve semantic parsing of unstructured information, and introducing an extraction structure correction mechanism to eliminate model illusions and logical errors, a low-carbon information recognition model with dynamic recognition capabilities is ultimately generated. This model is used to identify the target information to obtain green and low-carbon information, thus solving the technical problem of low accuracy in green and low-carbon information recognition and achieving the technical effect of improving the accuracy of green and low-carbon information recognition.

[0118] It should be noted that other corresponding descriptions of the functional units involved in the power metadata fusion device provided in this application embodiment can be found by referring to... Figure 1 The corresponding descriptions in [the document] will not be repeated here.

[0119] This application also provides a computer device, which may specifically be a personal computer, a server, a network device, etc. Figure 5 This application provides a schematic diagram of the device structure of a computer device according to an embodiment of the present application. Figure 5As shown, the computer device includes a bus, a processor, memory, and a communication interface, and may also include an input / output interface and a display device. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores location information. The network interface allows communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the various method embodiments.

[0120] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0121] In one embodiment, a computer-readable storage medium is provided, which may be non-volatile or volatile, having stored thereon a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0122] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0123] It should be noted that the user personal information involved in the embodiments of this application is all authorized (with the knowledge and consent) by the relevant parties or fully authorized by all parties, and the executing entity can obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with the relevant laws and regulations of the relevant countries and regions, and do not violate public order and good morals. It should be noted that if any software tools or components other than those of this company appear in the embodiments of this application, they are merely illustrative examples and do not represent actual use.

[0124] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0125] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0126] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A green and low-carbon information identification method based on an identification model, characterized in that, The method includes: Collect historical material data in the target products of power distribution network material manufacturing, and construct a data collection standard system by referring to the collected historical material data; Acquire initial data information related to the manufacturing of the power distribution network materials, and preprocess and perform semantic analysis on the initial data information to construct a multi-source heterogeneous data knowledge base; Key features are extracted from the information in the multi-source heterogeneous data knowledge base, and the extracted key features are corrected in terms of extraction structure, so as to build an underlying intelligent engine using the corrected key features. A green and low-carbon attribute indicator system is constructed, the underlying intelligent engine and the green and low-carbon attribute indicator system are integrated, and the system is accessed to a retrieval database to obtain a low-carbon information identification model. When identifying green and low-carbon information, the target information to be identified is collected in accordance with the data collection standard system, and the target information is input into the low-carbon information identification model for identification.

2. The method according to claim 1, characterized in that, The extracted key features are subjected to extraction structure correction in order to build an underlying intelligent engine using the corrected key features, including: Based on the extracted key features, an input flow set for the target product is set, wherein the input flow set includes at least the original input amounts of multiple input flows; Based on the material and energy database of the target product, the original input quantities of multiple input streams are mapped to obtain the equivalent quantities of multiple input streams, wherein the equivalent quantities are the uniform quantities of the original input quantities under a preset dimension. Based on the equivalent quantities of multiple input flows and the preset equivalent quantities, a constrained loss function is constructed; Based on the constraint loss function and the original input amounts of the multiple input streams, a target optimization function is constructed; The extracted key features are corrected using the constraint loss function and the objective optimization function, so as to construct the underlying intelligent engine using the corrected key features.

3. The method according to claim 2, characterized in that, A constrained loss function is constructed based on the equivalent quantities of multiple input flows and a preset equivalent quantity using the following formula, including: in, The constraint loss function is used to represent the degree of deviation between the weighted sum of the equivalent quantities of the multiple input flows and the preset equivalent quantity; N is used to represent the number of input flows. Used to indicate the first The equivalent quantity of the input flow; Used to indicate the first The weighting coefficients of each input flow are used to reflect the relative contribution of different raw materials in the manufacturing process of the target product; Used to represent the preset equivalent quantity. Used to represent the 2-norm.

4. The method according to claim 2, characterized in that, Based on the constrained loss function and the original input amounts of the multiple input streams, the objective optimization function is constructed using the following formula, including: in, Used to represent the first after being constrained by the constraint loss function. The amount of input into each input stream, Used to indicate the first The original input amount of each input stream, Used to represent the trade-off coefficient.

5. The method according to claim 1, characterized in that, Construct a green and low-carbon attribute indicator system, including: Obtain multiple green and low-carbon attribute indicators related to the manufacturing of the aforementioned power distribution network materials; The green and low-carbon attribute indicators are processed in a hierarchical manner to obtain primary attribute indicators and secondary attribute indicators, wherein the granularity of the primary attribute indicators is larger than that of the secondary attribute indicators. Based on the primary attribute indicators and the secondary attribute indicators, the green and low-carbon attribute indicator system is constructed.

6. The method according to claim 1, characterized in that, The initial data information is preprocessed and semantically analyzed to construct a multi-source heterogeneous data knowledge base, including: The initial data information is cleaned to obtain cleaned target data information, wherein the data noise of the target data information is less than the data noise of the initial data information; Obtain the target file storing the target data information and the file type of the target file, and split the target file based on the file type to obtain multiple split files; Vectorization and semantic analysis are performed on multiple split files to construct the multi-source heterogeneous data knowledge base.

7. The method according to claim 1, characterized in that, Key feature extraction is performed on the information in the multi-source heterogeneous data knowledge base, including: Obtain the task name field, category field, usage field, and unit field from the target product manufactured from distribution network materials, and set an identifier field related to the task name field; Based on the task name field, the category field, the usage field, the unit field, and the identifier field, a prompt word template is constructed; Key features are extracted from the information in the multi-source heterogeneous data knowledge base to obtain the extracted key features, and the extracted key features are output according to the prompt word template.

8. A green and low-carbon information identification device based on a recognition model, characterized in that, The device includes: The data acquisition unit is used to collect historical material data in the target products of power distribution network material manufacturing, and to construct a data acquisition standard system by referring to the collected historical material data. The first construction unit is used to acquire initial data information related to the manufacturing of the distribution network materials, and to preprocess and semantically analyze the initial data information to construct a multi-source heterogeneous data knowledge base. The second construction unit is used to extract key features from the information in the multi-source heterogeneous data knowledge base, and to perform extraction structure correction on the extracted key features, so as to build an underlying intelligent engine using the corrected key features. The acquisition unit is used to construct a green and low-carbon attribute indicator system, integrate the underlying intelligent engine and the green and low-carbon attribute indicator system, and access the retrieval database to obtain a low-carbon information identification model. The identification unit is used to collect target information to be identified in accordance with the data collection standard system when identifying green and low-carbon information, and to input the target information into the low-carbon information identification model for identification.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

10. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images