Product carbon footprint assessment method, apparatus, device, and storage medium

Abstract

The application relates to the technical field of carbon footprint assessment, and provides a product carbon footprint assessment method, device, equipment and storage medium, the method comprising the following steps: in response to a carbon footprint assessment request for a target product, dividing a life cycle stage of the target product, and acquiring activity data, real-time influence factor data and a target region label corresponding to each stage; calling a pre-constructed staged AI assessment model; the staged AI assessment model is configured to include independent sub-models for different life cycle stages, and the different independent sub-models establish a mapping relationship between influence factor data and carbon footprint factors by learning regional and staged historical data; based on the real-time influence factor data, the target region label and the independent sub-models for different life cycle stages, dynamic carbon footprint factors of each stage are obtained; and based on the dynamic carbon footprint factors of each stage and the activity data, the total carbon footprint of the target product is determined.

Description

Technical Field

[0001] This application relates to the field of carbon footprint assessment technology, and in particular to a method, apparatus, equipment and storage medium for assessing the carbon footprint of a product. Background Technology

[0002] Carbon footprint refers to an indicator used to quantify and assess the greenhouse gas emissions directly or indirectly generated by a product, service, or organization throughout its entire life cycle through a systematic approach. Life Cycle Assessment (LCA) is a commonly used method to assess emissions throughout a product's entire life cycle, and its general calculation formula is typically as follows: Where CFP stands for product carbon footprint, AD for activity data at each stage, and EF for the carbon footprint factor of the activity data. Activity data typically refers to the consumption at each stage (such as electricity consumption and fuel consumption), while the carbon footprint factor is a coefficient that converts activity data into carbon dioxide emission equivalents.

[0003] However, in existing carbon footprint accounting processes, the carbon footprint factor data used are usually derived from publicly available static databases or literature. These factor data are often statistical averages or estimates for large regions (such as the whole country or the whole province), and are typically fixed, static values. This traditional accounting method ignores the geographical differences in specific production areas and the dynamic environmental changes during actual production. Using such poorly applicable and coarse-grained static factors makes it difficult to accurately reflect the true carbon emission levels of products under specific spatiotemporal conditions, resulting in insufficient consistency and reliability of the final carbon footprint assessment results. Summary of the Invention

[0004] This application provides a product carbon footprint assessment method, apparatus, equipment, and storage medium to address the technical problem of insufficient consistency and reliability of carbon footprint assessment results in the prior art.

[0005] Firstly, this application provides a method for assessing the carbon footprint of a product, including: In response to a request for a carbon footprint assessment of a target product, the product's lifecycle stages are divided, and activity data, real-time influencing factor data, and target area labels are obtained for each stage. The pre-built phased AI assessment model is invoked; the phased AI assessment model is configured to include independent sub-models for different life cycle stages, and the different independent sub-models establish the mapping relationship between influencing factor data and carbon footprint factors by learning historical data by region and stage. Based on real-time influencing factor data, target area labels, and independent sub-models for different life cycle stages, dynamic carbon footprint factors for each stage are obtained. Based on the dynamic carbon footprint factor and activity data at each stage, the total carbon footprint of the target product is determined.

[0006] Secondly, this application also provides a product carbon footprint assessment device, comprising: The first acquisition module is used to respond to the carbon footprint assessment request for the target product, divide the life cycle stages of the target product, and acquire the activity data, real-time influencing factor data and target area labels corresponding to each stage. The module is used to invoke a pre-built phased AI assessment model. The phased AI assessment model is configured to include independent sub-models for different life cycle stages. The different independent sub-models establish a mapping relationship between influencing factor data and carbon footprint factors by learning historical data by region and stage. The second acquisition module is used to obtain the dynamic carbon footprint factor for each stage based on real-time influencing factor data, target area labels, and independent sub-models for different life cycle stages. The determination module is used to determine the total carbon footprint of a target product based on dynamic carbon footprint factors and activity data at each stage.

[0007] Thirdly, this application also provides an electronic device, including a processor and a memory storing a computer program, wherein the processor executes the program to implement the product carbon footprint assessment method of the first aspect.

[0008] Fourthly, this application also provides a processor-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the product carbon footprint assessment method of the first aspect.

[0009] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the product carbon footprint assessment method of the first aspect.

[0010] This application provides a product carbon footprint assessment method, apparatus, equipment, and storage medium. By introducing real-time influencing factors and regional labels, the model can output dynamic carbon footprint factors that conform to the current actual operating conditions, solving the problem that traditional static factors cannot reflect environmental changes and regional differences, and improving the accuracy of carbon footprint assessment. Users only need to provide easily accessible activity data and influencing factor data, without the need to deploy high-precision physical metering equipment on a large scale in all stages to measure emissions in real time. High-precision emission factors can be fitted by AI models, reducing assessment costs. At the same time, differentiated modeling is performed for the characteristics of different life cycle stages, effectively avoiding the problem of insufficient fitting ability of general models, and realizing refined carbon footprint assessment. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic flowchart of the product carbon footprint assessment method provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the product carbon footprint assessment device provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] Figure 1 This is a flowchart illustrating the product carbon footprint assessment method provided in the embodiments of this application, such as... Figure 1 As shown, this method uses artificial intelligence technology to address the issues of insufficient consistency and reliability in carbon footprint assessment results caused by static factors through software evaluation. The method includes the following steps: S101. In response to a request for a carbon footprint assessment of the target product, divide the life cycle stages of the target product and obtain activity data, real-time influencing factor data and target area labels for each stage.

[0015] Specifically, users can initiate carbon footprint assessment requests for target products through the software interface. To achieve a refined assessment, it is first necessary to break down the entire lifecycle of the target product (including product production or distribution processes) into multiple discrete lifecycle stages, such as the raw material acquisition stage, the manufacturing stage, and the transportation and distribution stage.

[0016] In addition, it is necessary to obtain activity data, real-time influencing factor data, and target area labels corresponding to each life cycle stage.

[0017] Activity data refers to directly quantifiable business data at each stage, such as electricity consumption, weight of materials used, and transportation distance. This data is usually readily available from a company's resource planning system or production reports.

[0018] Real-time influencing factor data refers to variables that dynamically affect carbon emissions at each stage. This application's embodiments introduce these dynamic variables, such as environmental factors like temperature and humidity (affecting the energy consumption of air conditioning or refrigeration equipment), or technical factors like process parameters and equipment operating status, which differs from traditional static calculations.

[0019] Target area labels refer to specific geographical locations (e.g., "XX Province, XX City") at each stage. By introducing area labels, it is possible to subsequently utilize environmental or technological benchmarks unique to that region, thereby reflecting regional differences.

[0020] S102, invoke the pre-built phased AI evaluation model.

[0021] The phased AI assessment model is configured to include independent sub-models for different life cycle stages. These independent sub-models learn from historical data from different regions and stages to establish a mapping relationship between influencing factor data and carbon footprint factors.

[0022] Specifically, since different stages of a product lifecycle have completely different physical characteristics and carbon emission mechanisms, it is difficult to fit all scenarios simultaneously using a single model. Therefore, in this embodiment, a phased artificial intelligence (AI) evaluation model containing multiple independent sub-models is pre-constructed, with each independent sub-model specifically designed and trained for a particular lifecycle stage.

[0023] During the training phase, these independent sub-models learn by utilizing a large amount of historical data from different regions and stages, and capture the complex nonlinear mapping relationship between historical influencing factor data and the true value of carbon footprint factor through algorithms.

[0024] S103. Based on real-time influencing factor data, target area labels, and independent sub-models for different life cycle stages, obtain the dynamic carbon footprint factor for each stage.

[0025] Specifically, the real-time influencing factor data and target area labels obtained in S101 are input into the corresponding independent sub-models in S102. The model will combine the input area labels (area features) and real-time influencing factor data (dynamic features) to perform inference calculations. The dynamic carbon footprint factor obtained at this time is a modified carbon footprint factor customized for the current region, current environmental conditions and current process status, rather than a fixed static value obtained by looking up tables, and has higher accuracy.

[0026] S104. Based on the dynamic carbon footprint factors and activity data at each stage, determine the total carbon footprint of the target product.

[0027] Specifically, using the general formula for calculating the carbon footprint of a product using LCA, the dynamic carbon footprint factors obtained in S103 at each stage are multiplied and summed with the activity data at each stage to obtain the total carbon footprint of the target product.

[0028] The product carbon footprint assessment method provided in this application introduces real-time influencing factors and regional labels. The model can output dynamic carbon footprint factors that conform to the current actual operating conditions, solving the problem that traditional static factors cannot reflect environmental changes and regional differences, and improving the accuracy of carbon footprint assessment. Users only need to provide easily accessible activity data and influencing factor data, without the need to deploy high-precision physical metering equipment on a large scale in all stages to measure emissions in real time. The AI ​​model can fit high-precision emission factors, reducing assessment costs. At the same time, differentiated modeling is performed for the characteristics of different life cycle stages, effectively avoiding the problem of insufficient fitting ability of general models, and realizing refined carbon footprint assessment.

[0029] In some embodiments, the lifecycle stages include one or more of the following: raw material stage, production stage, warehousing and transportation stage, use stage, and disposal stage.

[0030] In some embodiments, S101 includes dividing the lifecycle stages of the target product, including: The life cycle stages of the target product are determined based on the Life Cycle Assessment (LCA) method.

[0031] Specifically, this embodiment follows the general LCA methodology to divide the target product's lifecycle into stages. Specifically, the system discretizes the entire lifecycle into five core stages: Raw Material Stage: covering ore mining, raw material processing, and upstream transportation; Production Stage: covering manufacturing, assembly, and in-plant logistics within the factory; Warehousing and Transportation Stage: covering distribution and transportation of finished products after they leave the factory and intermediate warehousing; Use Stage: covering energy consumption during consumer use of the product; and Disposal Stage: covering product recycling, dismantling, landfill, or regeneration. This division provides clear physical boundaries for subsequent stage-by-stage modeling.

[0032] In some embodiments, the model structure of the independent sub-models for different lifecycle stages satisfies one or more of the following: If the lifecycle stage includes the raw material stage, the independent sub-model of the raw material stage is configured as a Long Short-Term Memory (LSTM) network structure. If the lifecycle stage includes a production stage, the independent sub-model of the production stage is configured as a combination of a convolutional neural network and a long short-term memory network in a CNN-LSTM structure. If the lifecycle stage includes the warehousing and transportation stage, the independent sub-model of the warehousing and transportation stage is configured as a bidirectional long short-term memory network Bi-LSTM structure. If the lifecycle phase includes a usage phase, the independent sub-model of the usage phase is configured as a gated recurrent unit (GRU) structure. If the lifecycle phase includes a discard phase, the independent sub-model of the discard phase is configured as a Transformer structure.

[0033] Specifically, in practical applications, the lifecycle stages of a product's carbon footprint typically include one or more of the following: raw material stage, production stage, warehousing and transportation stage, usage stage, and disposal stage. Since these stages each have distinctly different physical properties and data characteristics, in order to extract effective features to the greatest extent and improve prediction accuracy, specific deep neural network topologies are configured for the independent sub-models of different stages in this application embodiment.

[0034] Optionally, if the lifecycle includes a raw material stage, the independent sub-model for the raw material stage is configured as a Long Short-Term Memory (LSTM) network structure. Specifically, raw material acquisition often involves complex supply chain networks, and its carbon emissions are affected by the cumulative effects of upstream mining, initial processing, and multi-level transportation over time. For example, the delivery time of raw materials and seasonal supply fluctuations exhibit significant time-series characteristics. Therefore, for the raw material stage, the independent sub-model is configured as an LSTM structure. Through its unique gating mechanism (forget gate, input gate, output gate), the LSTM network can effectively capture long-distance time dependencies and accurately predict the carbon footprint factor of the raw material acquisition process, which changes dynamically over time.

[0035] Optionally, if the lifecycle stage includes a production stage, the independent sub-model for the production stage is configured as a CNN-LSTM structure combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) network. Specifically, the manufacturing process has both spatial and temporal characteristics. On the one hand, different production processes (such as assembly line layout and equipment combinations) have spatial topological features; on the other hand, production energy consumption fluctuates continuously with processing time. Therefore, for the production stage, the independent sub-model is configured as a CNN-LSTM hybrid structure. The CNN part utilizes convolutional layers to extract the spatial distribution features and local correlation features of process parameters (such as equipment model and production line layout). The LSTM part receives the feature vectors extracted by the CNN and further processes the temporal energy consumption data during the production process (such as power load fluctuations in continuous production). Through this hybrid architecture, the carbon footprint factor of the production stage can be accurately fitted simultaneously from both the dimensions of process complexity and temporal continuity.

[0036] If the lifecycle includes a warehousing and transportation phase, the independent sub-model for this phase is configured as a Bidirectional Long Short Term Memory (Bi-LSTM) network structure. Specifically, the warehousing and transportation process not only involves the outbound journey from the origin to the destination but also often includes the return journey from the destination or multi-point loops. Road conditions, load, and driving habits are interrelated in their impact on fuel consumption during the round trip. Therefore, for the warehousing and transportation phase, the independent sub-model is configured as a Bi-LSTM structure. Bi-LSTM contains two LSTM layers, forward and backward, which can simultaneously utilize information from past moments (such as outbound road conditions) and future moments (such as return journey planning), thereby more comprehensively extracting bidirectional logistics characteristics and improving the estimation accuracy of carbon footprint factors in the transportation process.

[0037] If the product lifecycle includes a usage phase, the independent sub-models for the usage phase are configured as Gated Recurrent Unit (GRU) structures. The usage phase (e.g., server operation, appliance use) typically lasts for a long time, and energy consumption data exhibits periodic fluctuations. Compared to LSTM, the GRU structure has fewer parameters, higher computational efficiency, and the same ability to handle long-sequence data. Considering that the data volume during the usage phase is usually large and requires rapid response, configuring the independent sub-models with GRU structures for the usage phase can reduce the computational complexity of the model while maintaining accuracy, making it suitable for simulating the cumulative energy consumption characteristics of products during long-term use.

[0038] If the lifecycle includes a disposal phase, the independent sub-model for the disposal phase is configured as a Transformer structure. Specifically, the disposal and recycling process typically involves multiple recycling techniques (such as physical crushing and chemical degradation), the sorting and processing of different materials, and complex coupling of environmental factors. These factors do not exhibit obvious time-series relationships but rather a global correlation dependency. Therefore, for the disposal phase, the independent sub-model is configured as a Transformer structure. The Transformer utilizes its core self-attention mechanism to process all input influencing factors in parallel, automatically focusing on the key factors with the greatest impact on carbon emissions. This allows it to capture long-distance, non-temporally sequential complex correlation features, achieving accurate prediction of carbon footprint factors for the disposal phase.

[0039] In this embodiment, the targeted phased model structure design ensures that each lifecycle stage can be matched with the algorithm most suitable for its data characteristics, thus significantly outperforming the evaluation effect of using a single general model.

[0040] In some embodiments, S103 specifically includes: S1031. Input the real-time influencing factor data of any target stage in the life cycle into the independent sub-model of the target stage to obtain the basic estimated value of the carbon footprint factor of the target stage. S1032. Based on the target area label of the target stage, retrieve the pre-set regionalized carbon footprint factor basic database and extract the regional benchmark data of the target area label. S1033. Generate regional adjustment coefficients based on regional benchmark data, and perform multiplicative weighting on the basic estimated value of carbon footprint factor based on regional adjustment coefficients. S1034. Determine the deviation between real-time influencing factor data and regional benchmark data, generate regional adjustment terms based on the deviation, and perform additive correction on the weighted carbon footprint factor baseline estimate based on the regional adjustment terms to determine the dynamic carbon footprint factor for the target stage.

[0041] Specifically, in practical applications, relying solely on the direct output of an AI model may lead to biases in the prediction of carbon footprint factors in specific new regions due to limitations in the training data. To improve the model's generalization ability and robustness, this application introduces a secondary correction mechanism based on regional benchmarks and environmental adjustments.

[0042] For any target stage in the product lifecycle, S1031 uses the aforementioned trained independent sub-models to perform preliminary inference on the input real-time influencing factor data, and the model output is a basic estimate of the carbon footprint factor. The basic estimate of the carbon footprint factor reflects the theoretical carbon footprint factor under standard operating conditions or the average level of the training dataset, and does not reflect extreme differences in specific regions or minor perturbations in the real-time environment.

[0043] In step S1032, based on the input target region label, a search is performed in a pre-built regionalized carbon footprint factor database to obtain the regional benchmark data corresponding to the target region label. Regional benchmark data refers to the statistically significant baseline level of a specific region, used to subsequently correct the basic estimates of carbon footprint factors. Examples include the average grid emission factor and average production technology efficiency index for a specific region.

[0044] In S1033, a dimensionless regional adjustment coefficient is generated by comparing regional benchmark data with the model's default benchmark. For example, if the target region has a high level of power grid cleanliness or a production technology level above average, a regional adjustment coefficient less than 1 is set. The generated regional adjustment coefficient is used to perform a multiplicative weighted average on the basic estimate of the carbon footprint factor, that is, multiplying the basic estimate of the carbon footprint factor by the regional adjustment coefficient. This can reflect the systematic deviations caused by differences in regional production technology levels, energy structure, etc., and these deviations are usually proportional.

[0045] In S1034, a regional adjustment term is generated based on the deviation between real-time influencing factor data and regional baseline data. This regional adjustment term is then used to additively correct the multiplicative weighted baseline estimate of the carbon footprint factor, resulting in the final dynamic carbon footprint factor. Additive correction involves adding the regional adjustment term to the multiplicative weighted baseline estimate of the carbon footprint factor. The regional adjustment term represents the additional carbon emissions caused by environmental fluctuations. This additive correction reflects incidental increases caused by real-time environmental fluctuations (such as extreme weather events), which are typically the sum of absolute values.

[0046] In this embodiment, by first multiplying by a regional adjustment coefficient and then adding a regional adjustment term, the macro-level regional statistical regularities and micro-level real-time environmental impacts can be combined to ultimately output a dynamic carbon footprint factor that conforms to regional characteristics and responds to real-time changes, thereby improving the accuracy of the assessment results under complex operating conditions.

[0047] Furthermore, since the external influence mechanisms differ at different life cycle stages, regional adjustment coefficients and regional adjustment items also need to be defined specifically.

[0048] In some embodiments, if the target stage is the raw material stage, the regional adjustment coefficient includes the regional mining technology level coefficient, which is used to quantify the impact of regional mining technology level differences on carbon emissions.

[0049] Specifically, in the process of raw material acquisition, the maturity of mining technology and the sophistication of equipment in different regions directly determine energy utilization efficiency. The difficulty of mining and the level of technological equipment vary greatly across different regions. A regional mining technology level coefficient is used to quantify these differences in mining technology levels.

[0050] Optionally, the system compares the mining technology index of the target area (derived from regional benchmark data, integrating indicators such as equipment electrification rate and automation rate) with the standard technology index preset by the model. If the mining technology index of the target area is higher than the standard technology index, the regional mining technology level coefficient is less than 1, reducing the carbon footprint factor; conversely, if the target area's technology is lagging behind, the coefficient is greater than 1. This correction based on technology level can eliminate assessment bias caused by ignoring differences in regional industrialization levels.

[0051] In some embodiments, if the target stage is the production stage, the regional adjustment coefficient includes the regional scale effect coefficient and / or the regional process efficiency coefficient, and the regional scale effect coefficient is negatively correlated with the output of the current batch of products.

[0052] Specifically, in the manufacturing process, the scale of a factory's output has a significant impact on the energy consumption per unit of product. The regional scale effect coefficient is negatively correlated with the output of the current batch of products, reflecting the energy-saving advantages of mass production. The larger the output of the current batch of products, the lower the allocated fixed energy consumption; the smaller the output of the current batch of products, the higher the allocated fixed energy consumption.

[0053] Besides economies of scale, significant differences exist in yield rates and waste control levels for specific manufacturing processes across different regions. A regional process efficiency coefficient is used to reflect the average technological maturity of a target region for a specific manufacturing process. Optionally, the regional process efficiency coefficient is configured to be negatively correlated with the regional average yield rate or process maturity. If the average yield rate of the target region is higher than the benchmark, the coefficient is less than 1, reflecting more efficient resource utilization; conversely, it is greater than 1. This correction can make implicit differences in process quality explicit as differences in carbon emission data.

[0054] In some embodiments, if the target stage is the warehousing and transportation stage, the regional adjustment coefficient includes the regional transportation efficiency coefficient, which is determined based on the current batch transportation volume and the average regional transportation scale.

[0055] Specifically, in the logistics and transportation process, vehicle loading rate, actual load rate, and degree of intensification have a significant impact on carbon emissions per unit of cargo. The regional transportation efficiency coefficient can dynamically reflect the impact of logistics intensification on carbon footprint, and is determined based on the ratio of the current batch transportation volume to the regional average transportation scale.

[0056] Optionally, if the current batch is a fully loaded truckload (large transport volume) and exceeds the regional average for carpooling, the coefficient is less than 1, reflecting the low-carbon advantages of intensive transportation; if the current batch is a small-batch, scattered delivery, the coefficient is greater than 1, reflecting an increase in emissions per unit. This correction can more accurately reflect the dynamic impact of logistics organization efficiency and load factor on carbon footprint.

[0057] In some embodiments, if the target stage is the usage stage, the regional adjustment coefficient includes a regional energy structure coefficient and / or an aging factor. The regional energy structure coefficient is used to quantify the impact of regional energy structure differences on carbon emissions, and the aging factor increases logarithmically with the product usage time.

[0058] Specifically, during the usage phase of electronic products or mechanical equipment, as the years of use increase, component aging and mechanical wear lead to a decrease in energy conversion efficiency. If only the energy consumption data at the time of manufacture is used to assess the carbon footprint throughout the entire life cycle, the results will be underestimated. Therefore, when the target phase is the usage phase, the regional adjustment coefficient is specifically configured as an aging factor, which is configured to increase logarithmically with the product's usage time. Product aging is typically rapid in the initial stage, then levels off but continues to increase. This correction can dynamically compensate for efficiency losses in the later stages of the product's life cycle, making the carbon footprint assessment of long-life products more accurate.

[0059] Besides aging factors, energy structures also differ across regions. Indirect carbon emissions from products during their use phase (especially electrically driven equipment) are highly dependent on the energy structure of the local power grid. Regional energy structure coefficients are used to quantify the impact of regional energy structure differences on carbon emissions.

[0060] Optionally, the system calculates the ratio of the clean energy share of the target region's power grid to a benchmark value. If the target region's power grid is cleaner, this coefficient is less than 1; if the target region relies on high-carbon coal power, this coefficient is greater than 1. This correction accurately reveals the decisive role of energy supply-side cleanliness in the carbon footprint of products.

[0061] In some embodiments, if the target stage is the waste stage, the regional adjustment coefficient includes a regional recycling efficiency coefficient, which is used to quantify the impact of efficiency differences in regional recycling processes on carbon emissions.

[0062] Specifically, the level of sophistication of the waste treatment industry chain varies across different regions, as does the conversion rate of recycling processes. The regional recycling efficiency coefficient is used to quantify these regional differences in recycling process efficiency.

[0063] Optionally, the regional recycling efficiency coefficient is determined based on a comparison between the average recycling conversion rate of the target area and a benchmark value. If the recycling conversion rate of the target area is high, the coefficient is less than 1; if inefficient landfilling or simple incineration is prevalent, the coefficient is greater than 1. This correction can reflect the regional imbalance in end-of-life treatment capacity and provide data support for optimizing the flow of product waste.

[0064] In some embodiments, if the target stage is the raw material stage, the regional adjustment item includes a temperature and humidity deviation adjustment item, which is determined based on the difference between real-time temperature and humidity data and regional reference temperature and humidity.

[0065] Specifically, raw materials are sensitive to ambient temperature and humidity during storage or initial processing. The temperature and humidity deviation adjustment item is determined based on the difference between real-time temperature and humidity data and the regional reference temperature and humidity. If the real-time environment is more humid or colder than the reference, resulting in the need for additional drying or heating energy consumption, the temperature and humidity deviation adjustment item will be a positive value.

[0066] In some embodiments, if the target stage is the production stage or the warehousing and transportation stage, the regional adjustment item includes a regional energy adjustment item to correct for carbon emission deviations caused by differences in regional electricity and fuel emission factors.

[0067] Specifically, the regional energy adjustment term is used to correct for the absolute difference between the general energy factors that the base model may use and the actual regional energy factors. For example, the base model may assume the use of standard diesel, while the target region uses a blended fuel containing biodiesel, or the actual emission factors of the regional power grid may deviate from the model's preset values. The regional energy adjustment term is used to directly correct for carbon emission deviations caused by differences in electricity and fuel emission factors.

[0068] In some embodiments, if the target stage is the usage stage, the regional adjustment item includes an environmental energy consumption adjustment item, which is determined based on the deviation between the real-time usage temperature and the regional reference temperature.

[0069] Specifically, the environmental energy consumption adjustment item primarily targets equipment significantly affected by ambient temperature. This item is determined based on the deviation between the real-time operating temperature and the regional reference temperature. For example, when the temperature exceeds the reference value, the cooling system load increases, generating additional environmental energy consumption, which is additively compensated for through the environmental energy consumption adjustment item.

[0070] In some embodiments, if the target phase is the decommissioning phase, the regional adjustment items include carbon emission reduction adjustment items, which are determined based on the regional carbon emission reduction coefficient and the amount of recycled fuel consumed.

[0071] Specifically, waste recycling is often accompanied by energy recovery or material recycling, which can offset carbon emissions. The carbon reduction adjustment is determined based on the regional carbon reduction coefficient and the actual amount of recycled fuel consumed. Typically, the carbon reduction adjustment is a negative value or used to deduct from total emissions, reflecting the positive environmental benefits of the recycling process.

[0072] In this embodiment, by defining refined correction parameters for different life cycle stages, the general correction logic is translated into a concrete implementation that conforms to physical laws, thereby further improving the accuracy of carbon footprint assessment.

[0073] In some embodiments, the phased AI evaluation model invoked in S102 is trained based on the following steps: S1021. Construct a regionalized carbon footprint factor database that includes multi-level administrative divisions and environmental characteristic labels; S1022. Extract historical data by region and stage from the regionalized carbon footprint factor basic database to construct a training dataset. The samples in the training dataset include historical influencing factor data and corresponding historical carbon footprint factor true values. S1023. Using historical influencing factor data as input and historical carbon footprint factor true values ​​as labels, train each independent sub-model.

[0074] Specifically, S1021 pre-constructs a multi-dimensional regional carbon footprint factor database containing multi-level administrative divisions and environmental feature labels. The multi-level administrative divisions refer to the organization of data according to a hierarchical structure such as "national-province-city-district," ensuring that the data granularity can be scaled down to specific cities or even industrial parks. Environmental feature labels are used to tagged each region, such as "subtropical monsoon climate," "plain terrain," and "coastal / inland," for subsequent model extraction of regional environmental benchmarks. This regional carbon footprint factor database stores historical carbon footprint factors, regional power grid factors, and industry benchmark energy consumption for each region.

[0075] In S1022, historical data by region and stage are extracted from the regionalized carbon footprint factor base database to construct a training dataset for training independent sub-models at different stages. Each training sample includes at least two parts: the input features are historical influencing factor data and the corresponding regional label, and the true value of the label is the measured or calculated true value of the carbon footprint factor.

[0076] In S1023, supervised training of each independent sub-model is performed using training samples. The constructed training samples are input into the corresponding neural network model structure for supervised learning. During training, the model automatically learns and extracts the complex nonlinear mapping relationship between influencing factor data and carbon footprint factors at each stage. Optionally, the error between the predicted value and the true label value is minimized using the backpropagation algorithm. Optionally, regional feature data is used as prior conditions to fine-tune the model parameters, enabling the model to distinguish emission pattern differences between different regions.

[0077] This application embodiment achieves a closed-loop process from data management to model optimization by constructing a dedicated regional carbon footprint factor base database.

[0078] In some embodiments, in order to achieve dynamic carbon footprint factor assessment, the real-time influencing factor data collected in this application covers the following three key dimensions: Environmental data is used to capture non-linear changes in energy consumption caused by environmental fluctuations. Environmental temperature data specifically includes temperature, humidity, and altitude data. Real-time collection of temperature and humidity data is conducted in production workshops, transportation routes, or the operating environment. This data directly affects the energy efficiency of air conditioning, refrigeration, or heating equipment. Altitude affects air density, which in turn affects the combustion efficiency of internal combustion engines in transportation vehicles and the performance of cooling equipment.

[0079] Technology-level data is used to differentiate emissions across different technological approaches and equipment levels. This data can specifically include process type, equipment model, and fuel type. Process type identifies the specific production process; different process types have drastically different energy consumption benchmarks. Equipment model identifies the specific production equipment, reflecting its energy efficiency level. Fuel type identifies the type of energy used; different energy types have different carbon footprint factors.

[0080] Logistics dimension data is used to achieve high-precision logistics carbon footprint accounting, especially for transportation assessment in complex terrain areas. Logistics dimension data can specifically include transportation mode, transportation distance, road slope, etc. Transportation mode can be categorized as road, rail, waterway, or air transport. Transportation distance refers to the actual logistics mileage. Road slope refers to the terrain slope data collected for the transportation route. Compared to calculating only horizontal distance, incorporating slope data can more accurately calculate the additional fuel consumption caused by vehicles overcoming gravity.

[0081] The real-time influencing factors differ across different stages. For example: In the raw materials stage, real-time influencing factors include raw material type, raw material consumption, transportation distance, transportation mode, fuel type, fuel consumption, time, temperature, and humidity. In the production stage, real-time influencing factors include production auxiliary material type, production auxiliary material consumption, electricity consumption of production equipment, production fuel type, fuel consumption, production process type, production output, time, temperature, and humidity. In the warehousing and transportation stage, real-time influencing factors include transportation distance, transportation mode, fuel type, fuel consumption, time, temperature, humidity, and route gradient. In the usage stage, real-time influencing factors include energy type, energy consumption, ingredient type, ingredient consumption, time, temperature, and humidity. In the waste disposal stage, real-time influencing factors include recycling method, recycling process, fuel type, fuel consumption, time, temperature, and humidity.

[0082] This application's embodiments introduce fine-grained, multi-dimensional, real-time influencing factor data, enabling the model to perceive subtle changes in the physical world and output more accurate dynamic carbon footprint factor prediction results.

[0083] The technical solutions provided in this application will be further illustrated below with several specific examples.

[0084] Example 1: Independent Sub-model of the Raw Materials Stage ① Input parameter acquisition: Real-time influencing factor data: Raw material type (T) m ), Acquisition Method (M) o ), transportation distance (D) t ), mode of transport (M) t ), fuel type (Ft ), fuel consumption (C) f ), time (T), temperature (Temp), humidity (H).

[0085] Target area label (R) m ): Associated regional benchmark data B m (R) m ), including: Regional benchmark mining technology level B m k Regional benchmark transport efficiency B m tr Regional reference temperature and humidity B m env .

[0086] ② Feature fusion processing Discrete feature embedding: transforming categorical variables into computable vectors that satisfy: =Embedding(T m， M o M t F t R m ); Standardization: Eliminating the influence of dimensions due to different regions and magnitudes, satisfying: ; in, Use it in region R m mean within and standard deviation Standardize it.

[0087] By concatenating features, discrete features, standardized continuous features, and regional benchmark data are integrated to provide comprehensive input features for subsequent models. ,satisfy: ; ③Basic estimation The LSTM structure is chosen to model the dynamic relationships during the raw material acquisition stage. The core formula is as follows: Forget gate: controls the proportion of historical information retained, satisfying: ; in, W is the sigmoid function. xf W hf Let b be the weight matrix. f For the bias term, h t-1 The state that was hidden in the previous moment.

[0088] Input gate: Controls the weights of the current input features on the model state update, satisfying: ; Cell state update: Accumulate carbon emission information throughout the entire process, satisfying: tanh(W) xc X m，t +W hc h t-1 +b c ); in, For element-wise multiplication, the tanh function outputs candidate states; cell state c t, Carbon emission information is dynamically accumulated through the synergy of forget gates and input gates.

[0089] Output gate and hidden state: Extract the carbon emission feature representation at the current time step, satisfying: ; ; Among them, output gate o t Controlling cell state c t The output ratio, the hidden state h t It is a carbon emission characteristic vector that comprehensively reflects the historical cumulative effect and current characteristics.

[0090] Basic carbon footprint factor estimate: The final state of the LSTM is transformed into a basic carbon footprint factor value, satisfying: ; Among them, h t To hide the state, use the weights W of the fully connected layer. m1 Bias b m1 Converted into specific numerical values, this represents the theoretical carbon footprint factor (unit: kgCO2e) without considering regional differences.

[0091] ④Regional Adjustment Regional adjustment coefficient (multiplicative): Specifically, this is the regional mining technology level coefficient, which quantifies the multiplicative impact of differences in regional mining technology levels on carbon emissions, satisfying the following: ; Regional Adjustment Item (Additive): This specifically refers to the temperature and humidity deviation adjustment item, which quantifies the additive impact of regional temperature and humidity deviations on energy consumption. It is determined based on the difference between real-time temperature and humidity data and the regional reference temperature and humidity, and satisfies the following: ; in, This represents the deviation between real-time temperature data and the regional reference temperature; the coefficient 0.02 is a constant, indicating that for every 1°C deviation, an additional 0.02 times the reference value is added, W. m3 The weights are adapted to the temperature-sensitive properties of different raw materials.

[0092] ⑤ Final output The dynamic carbon footprint factor adjusted for raw material stage meets the following requirements: CF 原材料 =CF m,bas ×α m,R +β m,R .

[0093] Example 2: Independent Sub-model of the Production Stage ① Input parameter acquisition: Real-time influencing factor data, production auxiliary material type (T) a ), Consumption of auxiliary materials (C) a ), equipment power consumption (E) e ), fuel type (F p ), fuel consumption (C) p ), process type (P) t ), output (Y), time (T), temperature (Temp), humidity (H).

[0094] Target area label (R) p ): Associated regional benchmark data B p (R) p ), including: regional electricity carbon footprint factor B p el Regional fuel carbon footprint factor B p fuel Regional process efficiency B p pr Regional scale effect coefficient B p sc .

[0095] ② Feature fusion processing By integrating process type, auxiliary material type, and auxiliary material consumption per unit output, the following conditions are met: =ProcessEmbedding(P t T a ) + Norm (C a / Y); Among them, ProcessEmbedding() captures the matching relationship between process and auxiliary materials and converts it into a vector. Norm(Ca / Y) is used to standardize the auxiliary material consumption per unit output to eliminate the influence of batch size.

[0096] Divide the batch's electricity consumption and fuel consumption by the output Y to obtain the energy consumption per unit output. Then, standardize the energy consumption of batches of different sizes to make them comparable, satisfying the following: E en =Norm(E e / Y,C p / Y); Feature splicing is performed, integrating process, energy, environmental, and regional characteristics to meet the following requirements: .

[0097] ③Basic estimation Choosing a CNN-LSTM hybrid architecture to extract process spatial features and production temporal features satisfies: ; Among them, local features are extracted using a 3×3 convolution kernel (Conv2d).

[0098] Capture energy consumption fluctuations in a continuous production process, satisfying: ; Process complexity correction involves increasing model focus on high-complexity processes to satisfy the following: ; Among them, P t Encode the process type. Transform the LSTM final state into a basic carbon footprint factor value, satisfying: ; Among them, the process features extracted by CNN are input into LSTM step by step, h t To be in a hidden state, h T This is the final hidden state at the end of production. Transformed into a basic estimate of the total carbon footprint for the production phase through a fully connected layer; This is the basic estimate of the carbon footprint factor.

[0099] ④Regional Adjustment Regional adjustment coefficient (multiplicative): Specifically, this is the regional scale effect coefficient, which quantifies the impact of regional production scale on unit energy consumption, satisfying the following: ; in, This is the ratio of the current batch output to the regional average production scale.

[0100] Regional adjustment term (additive): Specifically, this is the regional energy adjustment term, which corrects for carbon emission deviations caused by differences in regional energy structures, satisfying the following: ; in, (This refers to the regional benchmark carbon emissions corresponding to electricity consumption for production.) W represents the regional baseline carbon emissions corresponding to fuel consumption. p2 For training the weights of the model, b p2 This is the bias value.

[0101] ⑤ Final Output: Divide the total carbon emissions after multiplicative weighting and additive supplementation by the output Y to obtain the dynamic carbon footprint factor per unit of product at each production stage, satisfying: CF 生产 = .

[0102] Example 3: Independent Sub-model of the Warehousing and Transportation Stage ① Input parameter acquisition: Real-time influencing factor data: Transportation distance (D) t ), mode of transport (M) t ), fuel type (F t ), fuel consumption (C) f Time (T), Temperature (Temp), Humidity (H), and Road Slope Data (S) p ).

[0103] Target area label (R) t ): Associated regional benchmark data B t (R) t This includes regional benchmark transport carbon emissions, regional benchmark warehousing energy consumption, and regional road condition coefficients.

[0104] ② Feature fusion processing Transportation characteristics processing: Standardize transportation distance, total energy consumption, and energy consumption per unit distance (efficiency indicators) to meet the following requirements: =Embedding(M t F t ) + Norm (D t C f C f / D t ).

[0105] Warehouse environment characteristic processing: Quantify the impact of storage time and environmental conditions, satisfying: E stor =Norm(T, Temp, H).

[0106] Regional benchmark feature processing: Incorporating regional transportation and warehousing benchmark data to meet the following requirements: E reg =Norm(B t (R) t ).

[0107] Features are fused to satisfy: X t =Concat(E trans E stor E reg ).

[0108] ③Basic estimation Choose a bidirectional LSTM structure: A forward LSTM is used to model the forward transportation process from origin to destination, satisfying the following: ; Backward LSTM models the "destination-destination" return process, satisfying: ; Integrating forward and reverse transport characteristics to satisfy: ; By concatenating the hidden states of the forward and backward LSTMs, the overall characteristics of the round-trip transportation can be fully reflected.

[0109] ④Regional Adjustment Regional adjustment coefficient (multiplicative): Specifically, this is the regional transport efficiency coefficient, reflecting the multiplicative impact of regional road conditions and transport network on energy consumption, satisfying the following: ; in, Road condition coefficients in regional baseline data Regional Adjustment Item (Additive): This specifically refers to the terrain slope adjustment item, which compensates for the additional fuel consumption deviation caused by overcoming terrain slope, satisfying the following: in, Standardized road slope data.

[0110] ⑤ Final output The dynamic carbon footprint factor during the warehousing and transportation phase satisfies: .

[0111] Example 4: Independent Sub-models in the Usage Phase ① Input parameter acquisition: Real-time influencing factor data: Energy type (E t Energy consumption (C) e ), ingredient type (I t ), ingredient consumption (C) i Time (T), Temperature (Temp), Humidity (H) Target area label (R) u ): Associated regional benchmark data B u (R) u ), including: regional energy carbon footprint factor B u en User habit coefficient B u habAmbient reference temperature B u env .

[0112] ② Feature fusion processing The energy and ingredient characteristics are integrated to meet the following requirements: E en-in =Embedding(E t I t ) + Norm (C e C i C e / T); The features are concatenated and fused to satisfy: X u =Concat(E en-in Norm(T, Temp, H), B u (R) u )).

[0113] ③Basic estimation The core formula for choosing the GRU structure is as follows: Update gate: Controls the percentage of historical states retained, satisfying: ; in, Here, Wz and Uz are the sigmoid functions, and Wz and Uz are the weight matrices.

[0114] Reset gate: Controls the influence of historical states on the current candidate state, satisfying: ; Hidden state update, via updating gate z t Balanced historical state h t-1 With the current candidate state This forms the energy consumption characteristics at the current moment: ; Cumulative state calculation: s t =s t-1 +h t .

[0115] Baseline carbon footprint estimates: =W u1 sT+b u1 ; Where sT represents the cumulative feature over the entire usage cycle, through a fully connected layer (W u1 b u1 This is converted into theoretical carbon emission values.

[0116] ④Regional Adjustment Regional adjustment coefficient (multiplicative): includes regional energy structure coefficient and aging factor.

[0117] The regional energy structure coefficient quantifies the multiplicative impact of differences in regional energy structure / type on carbon emissions, satisfying the following: ; in, For the region Medium energy type Carbon footprint factor; through (Weight) and (Bias) Convert it into an adjustment factor.

[0118] The product aging factor increases logarithmically with product usage time, satisfying the following: t =1+0.05×log(T+1); Where T is the usage duration, and log(T+1) ensures that the aging effect increases over time but at a slower rate.

[0119] Regional Adjustment Item (Additive): This specifically refers to the environmental energy consumption adjustment item, which corrects for the additional absolute carbon emissions caused by real-time temperature deviations, satisfying the following: in, This represents the deviation between the actual operating temperature and the regional reference temperature. To quantify the additional energy consumption caused by environmental deviations, W u3 Model training weights, b u3 Bias value.

[0120] By correcting the base value using the regional energy structure coefficient and environmental energy consumption adjustment term, and then multiplying it by the aging factor, the carbon footprint factor output for the entire lifecycle of the usage phase is obtained, satisfying the following: CF 使用 =(CF u，bas × + )× t .

[0121] Example 5: Independent Sub-model of the Abandonment Phase ① Input parameter acquisition: Real-time influencing factor data: recovery method (R m ), recycling process (R) p ), fuel type (F r ), fuel consumption (C) r ), time (T), temperature (Temp), humidity.

[0122] Region Label (R) v ): Associated regional benchmark data B v (R) v ), including: regional benchmark recovery efficiency B v eff Regional carbon emission reduction coefficient B v red Processing benchmark B v dis .

[0123] ② Feature fusion processing Integration of Recycling and Processing Characteristics: Integrating recycling methods, processes, fuel types, and energy consumption Norm (C) r Standardize the total energy consumption for recycling to eliminate differences between recycling batches of different scales and meet the following requirements: E rec =Embedding(R m R p F r ) + Norm (C r ); Environmental characteristic integration: Standardize recycling processing time, ambient temperature, and humidity to make the environmental impact of recycling in different areas comparable and meet the following requirements: E env =Norm(T, Temp, H); Regional baseline feature fusion: Standardize regional baseline recovery efficiency, treatment baseline, and carbon emission reduction coefficient, and unify them with other feature dimensions to meet the following requirements: E reg =Norm(B v (R) v ); Perform feature concatenation to satisfy: X v =Concat(E rec E env E reg ).

[0124] ③Basic estimation We choose the Transformer architecture and employ a multi-head attention mechanism. The core formula is as follows: ; ; in, The weight matrix is ​​used, with each head capturing a type of process correlation; the outputs of each head are concatenated using Concat, and then multiplied by... (Weights) are used to obtain associated features.

[0125] Encoder output: The multi-head attention features are processed through two fully connected layers (W1 and W2 are weights, b1 and b2 are biases) and a ReLU activation function to enhance key association features, satisfying the following: ; The basic estimated value satisfies: CF v，bas =W v1 H+b v1 .

[0126] ④Regional Adjustment Regional adjustment factor (multiplicative): Specifically, this is the regional recycling efficiency factor, which quantifies the impact of efficiency differences in regional recycling processes on carbon emissions, satisfying the following: ; in, For the region Mid-cycle recycling process The efficiency coefficient.

[0127] Regional adjustment term (additive): This is specifically a carbon emission reduction adjustment term, which serves as a negative bias to supplement emission reduction gains, satisfying the following: in, This represents the regional emission reduction coefficient.

[0128] ⑤ Final output First, the baseline value is corrected using the recycling efficiency coefficient, then the regional carbon reduction adjustment term is subtracted to obtain the carbon footprint factor output for the entire waste disposal process, satisfying the following: CF 废弃 =CF v，bas × .

[0129] Example 5: Carbon Footprint Assessment For the carbon footprint assessment request of the target product, the life cycle stages of the target product are divided, and the activity data, real-time influencing factor data and target area labels corresponding to each stage are obtained; the real-time influencing factor data and target area labels are fed into the pre-trained independent sub-models of each stage to obtain the dynamic carbon footprint factor of each stage; based on the dynamic carbon footprint factor and activity data of each stage, the total carbon footprint of the target product is determined.

[0130] Among them, the total carbon footprint factor of the target product within the accounting region can be determined based on the dynamic carbon footprint factor at each stage, satisfying the following: CF 总 =CF 原材料 +CF 生产 +CF 运输 +CF 使用 +CF 废弃; Among them, CF 总 To determine the carbon footprint factor of the target product within the accounting region, relevant information, including regional and influencing factor information, is reviewed and stored in the regionalized carbon footprint factor database. The carbon footprint factor should be dynamically updated. In the event of significant changes, such as changes in production processes, the model should be reassessed to generate new carbon footprint factors, which will then be reviewed and updated.

[0131] The basic database fields for regionalized carbon footprint factors should include: Unique identifier for factors: such as region-carbon factor-year-serial number; Factor Name: Carbon Footprint Factor of Product xx in xxx Region; Factor categories: such as cradle to gate, cradle to grave; Applicable area: xxx province xx; Applicable fields: such as server manufacturing; Factor value: The core value of the carbon footprint factor (in "carbon dioxide equivalent (CO2e)"). Unit of measurement: The unit corresponding to the factor value, such as kgCO2e / kWh, kgCO2e / t (tons); Accounting boundaries: Time dimension, the time range corresponding to the factor; Statistical range, describing the statistical range of the factor; Data source organization: Data publishing unit; Release / Update Date: [Date] Factor description (notes): Special applicable conditions or precautions for supplementary factors; Review status: Has the factor been approved? Dynamic iteration: Set review points to avoid using outdated data; The review process includes, but is not limited to, comparisons with existing carbon footprint factors and threshold determination.

[0132] The product carbon footprint assessment device provided in this application is described below. The product carbon footprint assessment device described below can be referred to in correspondence with the product carbon footprint assessment method described above.

[0133] Figure 2 This is a schematic diagram of the structure of the product carbon footprint assessment device provided in the embodiments of this application, as shown below. Figure 2 As shown, the device includes: The first acquisition module 201 is used to respond to the carbon footprint assessment request for the target product, divide the life cycle stages of the target product, and acquire the activity data, real-time influencing factor data and target area labels corresponding to each stage. Module 202 is invoked to invoke a pre-built phased AI assessment model. The phased AI assessment model is configured to include independent sub-models for different life cycle stages. Different independent sub-models establish a mapping relationship between influencing factor data and carbon footprint factors by learning historical data by region and stage. The second acquisition module 203 is used to obtain the dynamic carbon footprint factor for each stage based on real-time influencing factor data, target area labels, and independent sub-models for different life cycle stages. Module 204 is used to determine the total carbon footprint of the target product based on the dynamic carbon footprint factor and activity data at each stage.

[0134] In some embodiments, the life cycle stages include one or more of the following: raw material stage, production stage, warehousing and transportation stage, use stage, and disposal stage; The model structure of the independent sub-models for different lifecycle stages satisfies one or more of the following: If the lifecycle stage includes the raw material stage, the independent sub-model of the raw material stage is configured as a Long Short-Term Memory (LSTM) network structure. If the lifecycle stage includes a production stage, the independent sub-model of the production stage is configured as a combination of a convolutional neural network and a long short-term memory network in a CNN-LSTM structure. If the lifecycle stage includes the warehousing and transportation stage, the independent sub-model of the warehousing and transportation stage is configured as a bidirectional long short-term memory network Bi-LSTM structure. If the lifecycle phase includes a usage phase, the independent sub-model of the usage phase is configured as a gated recurrent unit (GRU) structure. If the lifecycle phase includes a discard phase, the independent sub-model of the discard phase is configured as a Transformer structure.

[0135] In some embodiments, obtaining the dynamic carbon footprint factor at each stage includes: Input real-time influencing factor data of any target stage in the life cycle into an independent sub-model of the target stage to obtain the basic estimated value of the carbon footprint factor of the target stage. Based on the target area label of the target stage, a pre-set regionalized carbon footprint factor database is retrieved, and regional benchmark data of the target area label is extracted. Regional adjustment coefficients are generated based on regional baseline data, and the basic estimated values ​​of carbon footprint factors are multiplicatively weighted based on the regional adjustment coefficients. Determine the deviation between real-time influencing factor data and regional benchmark data, generate regional adjustment terms based on the deviation, and perform additive correction on the weighted carbon footprint factor baseline estimate based on the regional adjustment terms to determine the dynamic carbon footprint factor for the target stage.

[0136] In some embodiments, the regional adjustment factor satisfies one or more of the following: If the target stage is the raw material stage, the regional adjustment coefficient includes the regional mining technology level coefficient, which is used to quantify the impact of differences in regional mining technology levels on carbon emissions. If the target stage is the production stage, the regional adjustment coefficient includes the regional scale effect coefficient and / or the regional process efficiency coefficient. The regional scale effect coefficient is negatively correlated with the output of the current batch of products. If the target stage is the warehousing and transportation stage, the regional adjustment coefficient includes the regional transportation efficiency coefficient, which is determined based on the current batch transportation volume and the average regional transportation scale. If the target stage is the usage stage, the regional adjustment coefficient includes the regional energy structure coefficient and / or aging factor. The regional energy structure coefficient is used to quantify the impact of regional energy structure differences on carbon emissions, and the aging factor increases logarithmically with the product usage time. If the target stage is the waste stage, the regional adjustment coefficient includes the regional recycling efficiency coefficient, which is used to quantify the impact of regional recycling process efficiency differences on carbon emissions.

[0137] In some embodiments, the region adjustment item satisfies one or more of the following: If the target stage is the raw materials stage, the regional adjustment items include temperature and humidity deviation adjustment items, which are determined based on the difference between real-time temperature and humidity data and regional benchmark temperature and humidity. If the target phase is the production phase or the warehousing and transportation phase, the regional adjustment items include regional energy adjustment items, which are used to correct carbon emission deviations caused by differences in regional electricity and fuel emission factors; If the target phase is the usage phase, the regional adjustment items include environmental energy consumption adjustment items, which are determined based on the deviation between the real-time usage temperature and the regional reference temperature; If the target phase is the decommissioning phase, the regional adjustment items include carbon emission reduction adjustment items, which are determined based on the regional carbon emission reduction coefficient and the amount of recycled fuel consumed.

[0138] In some embodiments, the phased AI evaluation model is trained based on the following steps: Construct a regional carbon footprint factor database that includes multi-level administrative divisions and environmental characteristic labels; A training dataset is constructed by extracting historical data from regional and phased data from a regionalized carbon footprint factor database. The samples in the training dataset include historical influencing factor data and corresponding historical carbon footprint factor true values. Using historical influencing factor data as input and historical carbon footprint factor true values ​​as labels, each independent sub-model is trained.

[0139] In some embodiments, the lifecycle stages of the target product are divided, including: The life cycle stages of the target product are determined based on the Life Cycle Assessment (LCA) method.

[0140] In some embodiments, real-time influencing factor data includes one or more of the following: Environmental dimension data; Technical dimension data; Logistics dimension data.

[0141] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application, such as... Figure 3 As shown, the electronic device may include: a processor 301, a communications interface 302, a memory 303, and a communication bus 304, wherein the processor 301, the communications interface 302, and the memory 303 communicate with each other via the communication bus 304. The processor 301 can call logical instructions in the memory 303 to execute a product carbon footprint assessment method, including: In response to a request for a carbon footprint assessment of a target product, the product's lifecycle stages are divided, and activity data, real-time influencing factor data, and target area labels are obtained for each stage. The pre-built phased AI assessment model is invoked; the phased AI assessment model is configured to include independent sub-models for different life cycle stages, and the different independent sub-models establish the mapping relationship between influencing factor data and carbon footprint factors by learning historical data by region and stage. Based on real-time influencing factor data, target area labels, and independent sub-models for different life cycle stages, dynamic carbon footprint factors for each stage are obtained. Based on the dynamic carbon footprint factor and activity data at each stage, the total carbon footprint of the target product is determined.

[0142] Furthermore, the logical instructions in the aforementioned memory 303 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0143] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a processor-readable storage medium. When the computer program is executed by the processor, the computer is able to perform the product carbon footprint assessment methods provided by the above methods, including: In response to a request for a carbon footprint assessment of a target product, the product's lifecycle stages are divided, and activity data, real-time influencing factor data, and target area labels are obtained for each stage. The pre-built phased AI assessment model is invoked; the phased AI assessment model is configured to include independent sub-models for different life cycle stages, and the different independent sub-models establish the mapping relationship between influencing factor data and carbon footprint factors by learning historical data by region and stage. Based on real-time influencing factor data, target area labels, and independent sub-models for different life cycle stages, dynamic carbon footprint factors for each stage are obtained. Based on the dynamic carbon footprint factor and activity data at each stage, the total carbon footprint of the target product is determined.

[0144] In another aspect, this application also provides a processor-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the product carbon footprint assessment methods provided by the methods described above, including: In response to a request for a carbon footprint assessment of a target product, the product's lifecycle stages are divided, and activity data, real-time influencing factor data, and target area labels are obtained for each stage. The pre-built phased AI assessment model is invoked; the phased AI assessment model is configured to include independent sub-models for different life cycle stages, and the different independent sub-models establish the mapping relationship between influencing factor data and carbon footprint factors by learning historical data by region and stage. Based on real-time influencing factor data, target area labels, and independent sub-models for different life cycle stages, dynamic carbon footprint factors for each stage are obtained. Based on the dynamic carbon footprint factor and activity data at each stage, the total carbon footprint of the target product is determined.

[0145] Processor-readable storage media can be any available medium or data storage device that the processor can access, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MOs), etc.), optical storage (such as CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (such as ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs)).

[0146] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method of product carbon footprint assessment, characterized by, include: In response to a carbon footprint assessment request for a target product, the life cycle stages of the target product are divided, and activity data, real-time influencing factor data, and target area labels corresponding to each stage are obtained. The pre-built phased AI assessment model is invoked; the phased AI assessment model is configured to include independent sub-models for different life cycle stages, and the different independent sub-models establish a mapping relationship between influencing factor data and carbon footprint factors by learning historical data by region and stage. Based on the real-time influencing factor data, the target area label, and the independent sub-models for different life cycle stages, the dynamic carbon footprint factor for each stage is obtained. Based on the dynamic carbon footprint factors at each stage and the activity data, the total carbon footprint of the target product is determined.

2. The product carbon footprint assessment method according to claim 1, characterized in that, The life cycle stages include one or more of the following: raw material stage, production stage, warehousing and transportation stage, use stage, and disposal stage; The model structure of the independent sub-models for different lifecycle stages satisfies one or more of the following: If the life cycle stage includes the raw material stage, the independent sub-model of the raw material stage is configured as a Long Short-Term Memory (LSTM) network structure. If the lifecycle stage includes the production stage, the independent sub-model of the production stage is configured as a CNN-LSTM structure combining a convolutional neural network and a long short-term memory network. If the lifecycle stage includes the warehousing and transportation stage, the independent sub-model of the warehousing and transportation stage is configured as a bidirectional long short-term memory network (Bi-LSTM) structure. If the lifecycle phase includes the usage phase, the independent sub-model of the usage phase is configured as a gated recurrent unit (GRU) structure. If the lifecycle phase includes the obsolescence phase, the independent sub-model of the obsolescence phase is configured as a Transformer structure.

3. The product carbon footprint assessment method according to claim 1, characterized in that, The process of obtaining the dynamic carbon footprint factor at each stage includes: Input the real-time influencing factor data of any target stage in the life cycle into the independent sub-model of the target stage to obtain the basic estimated value of the carbon footprint factor of the target stage; Based on the target area label of the target stage, a pre-set regionalized carbon footprint factor database is retrieved, and regional benchmark data of the target area label is extracted. A regional adjustment coefficient is generated based on the regional benchmark data, and the basic estimated value of the carbon footprint factor is multiplicatively weighted based on the regional adjustment coefficient. The deviation between the real-time influencing factor data and the regional benchmark data is determined. A regional adjustment term is generated based on the deviation value. An additive correction is performed on the weighted carbon footprint factor baseline estimate based on the regional adjustment term to determine the dynamic carbon footprint factor for the target stage.

4. The product carbon footprint assessment method according to claim 3, characterized in that, The regional adjustment coefficient satisfies one or more of the following: If the target stage is the raw material stage, the regional adjustment coefficient includes the regional mining technology level coefficient, which is used to quantify the impact of regional mining technology level differences on carbon emissions. If the target stage is the production stage, the regional adjustment coefficient includes the regional scale effect coefficient and / or the regional process efficiency coefficient, and the regional scale effect coefficient is negatively correlated with the output of the current batch of products. If the target stage is the warehousing and transportation stage, the regional adjustment coefficient includes the regional transportation efficiency coefficient, which is determined based on the current batch transportation volume and the average regional transportation scale. If the target stage is the usage stage, the regional adjustment coefficient includes a regional energy structure coefficient and / or an aging factor. The regional energy structure coefficient is used to quantify the impact of regional energy structure differences on carbon emissions, and the aging factor increases logarithmically with the product usage time. If the target stage is the abandonment stage, the regional adjustment coefficient includes the regional recycling efficiency coefficient, which is used to quantify the impact of regional recycling process efficiency differences on carbon emissions.

5. The product carbon footprint assessment method according to claim 3, characterized in that, The regional adjustment item meets one or more of the following criteria: If the target stage is the raw material stage, the regional adjustment item includes a temperature and humidity deviation adjustment item, which is determined based on the difference between real-time temperature and humidity data and regional benchmark temperature and humidity. If the target stage is the production stage or the warehousing and transportation stage, the regional adjustment item includes a regional energy adjustment item, which is used to correct carbon emission deviations caused by differences in regional electricity and fuel emission factors. If the target stage is the usage stage, the regional adjustment item includes an environmental energy consumption adjustment item, which is determined based on the deviation between the real-time usage temperature and the regional reference temperature; If the target phase is the abandonment phase, the regional adjustment items include carbon emission reduction adjustment items, which are determined based on the regional carbon emission reduction coefficient and the amount of recycled fuel consumed.

6. The product carbon footprint assessment method according to claim 1, characterized in that, The phased AI evaluation model is trained based on the following steps: Construct a regional carbon footprint factor database that includes multi-level administrative divisions and environmental characteristic labels; A training dataset is constructed by extracting historical data by region and stage from the regionalized carbon footprint factor basic database. The samples in the training dataset include historical influencing factor data and corresponding historical carbon footprint factor true values. Using the historical influencing factor data as input and the historical carbon footprint factor true value as label, each independent sub-model is trained.

7. The product carbon footprint assessment method according to claim 1, characterized in that, The process of dividing the lifecycle stages of the target product includes: The target product's life cycle stages are determined based on the Life Cycle Assessment (LCA) method.

8. The product carbon footprint assessment method according to claim 1, characterized in that, The real-time influencing factor data includes one or more of the following: Environmental dimension data; Technical dimension data; Logistics dimension data.

9. A product carbon footprint assessment device, characterized in that, include: The first acquisition module is used to respond to a carbon footprint assessment request for a target product, divide the life cycle stages of the target product, and acquire activity data, real-time influencing factor data, and target area labels corresponding to each stage. The calling module is used to call a pre-built phased AI evaluation model; the phased AI evaluation model is configured to include independent sub-models for different life cycle stages, and the different independent sub-models establish a mapping relationship between influencing factor data and carbon footprint factors by learning historical data of different regions and stages; The second acquisition module is used to obtain the dynamic carbon footprint factor for each stage based on the real-time influencing factor data, the target area label, and the independent sub-models for different life cycle stages. The determination module is used to determine the total carbon footprint of the target product based on the dynamic carbon footprint factor of each stage and the activity data.

10. An electronic device comprising a processor and a memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the product carbon footprint assessment method as described in any one of claims 1 to 8.

11. A processor-readable storage medium, characterized in that, The processor-readable storage medium stores a computer program for causing the processor to execute the product carbon footprint assessment method as described in any one of claims 1 to 8.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the product carbon footprint assessment method as described in any one of claims 1 to 8.

Images