Gold recycling carbon footprint credible storage and intelligent traceability system based on blockchain and ai technology

Abstract

This invention discloses a trusted carbon footprint storage and intelligent traceability system for gold recycling based on blockchain and AI technologies, belonging to the intersection of green finance technology and environmental protection technology. The system adopts a four-layer architecture: the data acquisition layer binds batch IDs through GS1 encoding and integrates multi-source data through IoT sensors, a mobile app, and edge preprocessing; the AI ​​processing layer accurately calculates the carbon footprint based on a dynamic LCA model, optimizes process parameters through deep reinforcement learning, and simulates emission reduction potential; the blockchain storage layer uses a consortium blockchain (PBFT consensus) + IPFS storage, deploying smart contracts for storage and verification to automatically generate CDCEER carbon credits; the application layer provides traceability queries, ESG dashboards, regulatory audits, and green finance interfaces. This invention solves the problems of data fragmentation, inefficient calculation, and difficulty in value conversion, achieving trusted management and value conversion of the entire carbon footprint process, helping enterprises reduce costs and increase efficiency, comply with emission reduction regulations, and promote the green upgrading of the recycled precious metals industry.

Description

Technical Field

[0001] This invention relates to the field of green finance technology and environmental protection technology, specifically a trusted evidence storage and intelligent traceability system for carbon footprint of gold recycling based on blockchain and AI technology. Background Technology

[0002] Gold recycling, as a core component of the green circular economy, plays an irreplaceable role in alleviating the pressure of mining and reducing environmental impact. According to the World Gold Council, the carbon emissions from recycled gold production are only 1 / 30th that of primary gold (approximately 20-30 kg CO2e / oz vs. 600-800 kg CO2e / oz). The transparency of the carbon footprint of gold recycling has become a key indicator for international supply chain compliance and green finance allocation: the EU's Carbon Border Adjustment Mechanism (CBAM) explicitly requires imported recycled gold to declare its full life-cycle carbon footprint, otherwise it will face high tariffs; ESG investment institutions incorporate the accuracy of carbon footprint calculations (e.g., batch-level vs. enterprise-level) into the rating system for precious metal companies, directly affecting financing availability. Against this backdrop, it is crucial to build an accurate, reliable, and traceable carbon footprint management system.

[0003] Despite the urgent demand, the industry faces systemic technical obstacles. Firstly, data fragmentation and low credibility: the recycling chain involves multiple entities, including individual recyclers (decentralized, lacking standardized records), logistics providers (fragmented multimodal transport data), and refineries (closed process parameters). Data collection relies on paper documents or isolated Excel spreadsheets, forming "information silos." Under the centralized database storage model, data is easily tampered with (e.g., falsely reporting recycling volume, concealing energy consumption), and lacks cross-entity collaborative verification mechanisms, making it difficult to meet the ISO 14064 audit requirements for "data traceability and non-repudiation." Secondly, the accounting is crude and lacks optimization guidance: existing carbon accounting mostly uses static estimates based on industry averages (such as a one-size-fits-all approach of "150kg CO2e emissions per ton of recycled gold"), ignoring batch differences (such as energy consumption differences of up to 30% in purifying waste of different purities) and process fluctuations (such as the impact of smelting temperature deviations on energy consumption). This results in accounting results deviating from reality by more than 20%, and makes it impossible to identify high-emission links (such as high empty-load rates in transportation and excessive use of refining chemicals), leading to blind emission reduction measures by enterprises. Thirdly, the environmental value is difficult to convert: even if some enterprises achieve emission reduction through technological upgrades, their results lack a "credible quantification-third-party verification-assetization" path, making it impossible to connect with the carbon trading market (such as CCER) or green finance (such as carbon-backed loans). Emission reduction investments are difficult to obtain economic returns, creating a vicious cycle of "insufficient motivation for emission reduction."

[0004] Current industry attempts to develop technological solutions have failed to overcome the aforementioned bottlenecks. Traditional ERP systems can only manage internal work orders and inventory processes, failing to cover multi-source data from decentralized recycling scenarios. Furthermore, data stored on centralized servers is susceptible to tampering. Centralized database traceability solutions (such as SQL-based batch tracking) can record basic circulation information but lack integration with carbon accounting models, thus failing to output carbon footprint data. Single blockchain traceability systems (such as Hyperledger Fabric-based product traceability) focus on data storage but lack AI-driven precise accounting and process optimization modules, and have not established interfaces for carbon asset verification and financial transformation. In summary, existing technologies either focus on specific aspects (such as traceability or storage only) or fail to form a complete closed loop of "credible data collection → precise accounting → intelligent optimization → value transformation," resulting in gold recycling carbon footprint management remaining at a "passive compliance" level for a long time, failing to unleash the economic and environmental value of green transformation.

[0005] To address this, we propose a trusted and intelligent traceability system for the carbon footprint of gold recycling based on blockchain and AI technologies. Summary of the Invention

[0006] To achieve the above objectives, this invention provides the following technical solution: a trusted evidence storage and intelligent traceability system for the carbon footprint of gold recycling based on blockchain and AI technologies, comprising a data acquisition layer, an AI processing layer, a blockchain evidence storage layer, and an application layer that interact sequentially, wherein:

[0007] Data Acquisition Layer: Used to achieve multi-source acquisition of data throughout the entire lifecycle of gold recycling, industrial identifier binding, and edge intelligent preprocessing, specifically including:

[0008] The multi-source heterogeneous data acquisition module, through an IoT sensor group (including an XRF spectrometer, smart meter, and GPS locator), a mobile APP (supporting individual recyclers to enter the weight, purity, time, and geographical location of the recycled materials), and an ERP / GPS interface (connecting to the TMS system of transportation companies and the MES system of refineries), collects data on the composition and origin of waste materials in the recycling process, the load / mileage / fuel type in the transportation process, and the energy consumption (electricity / gas), material consumption (chemical usage), and process parameters (melting temperature, oxygen flow rate, and reaction time) in the refining process.

[0009] The industrial identification binding module uses the GS1-128 coding standard to assign a unique batch ID to each batch of recycled gold. Through the API interface, the batch ID is mapped to GPS trajectory data, ERP work order number, and XRF spectrometer detection data to achieve full-process data connectivity with a single code.

[0010] The edge intelligence preprocessing module is deployed at the data acquisition source (such as IoT gateway, mobile APP local), and uses the isolated forest algorithm or LOF algorithm for anomaly detection. It cleans sensor data in real time with ±3σ as the noise filtering threshold, improving the data accuracy to ≥95%.

[0011] AI processing layer: used to achieve accurate carbon footprint calculation, intelligent optimization of process parameters, and simulation of emission reduction potential, specifically including:

[0012] The dynamic LCA carbon footprint calculation module constructs a batch-level accounting model based on the Life Cycle Assessment (LCA) theory. It integrates the emission factor library of the Ministry of Ecology and Environment's "Guidelines for the Compilation of Provincial Greenhouse Gas Inventories" and the industry benchmark library of the World Gold Council (WGC). Input parameters include real-time energy consumption (unit: kWh), material consumption (unit: kg), transport load (unit: ton), fuel type (diesel / gasoline / biodiesel), and mileage (unit: km). The output is the batch carbon footprint (unit: kg CO2e), with an accounting error rate of ≤5%.

[0013] The deep reinforcement learning optimization engine uses "minimizing energy consumption per unit of gold purification" as the reward function and constructs a policy model containing an Actor-Critic network to dynamically adjust the smelting temperature (range 1100-1350℃), oxygen flow rate (range 150-250m³ / h), and reaction time (range 2-4 hours) in the refining process, thereby reducing energy consumption by 12%-18%.

[0014] The process emission reduction simulation module has a built-in virtual process flow model (covering crushing, smelting, and electrolytic purification sub-modules). It compares the company's actual carbon emissions with WGC benchmark values ​​(such as 600 kg CO2e / ounce carbon emissions from primary gold smelting) and pre-evaluates the carbon emission reduction and return on investment (ROI) of technical improvement plans (such as replacing induction melting furnaces and optimizing chemical ratios).

[0015] Blockchain evidence storage layer: used to achieve trusted data storage, automatic carbon asset verification, and cross-entity collaborative verification, specifically including:

[0016] The hierarchical storage module adopts an architecture that combines a consortium blockchain (nodes include gold recyclers, refineries, logistics companies, regulatory agencies, and third-party verification agencies) with IPFS off-chain storage. The raw data (sensor logs, APP input records, ERP work orders) is encrypted and stored in IPFS and a content identifier (CID) is generated. The carbon footprint calculation results, traceability information, and smart contract execution logs are uploaded to the blockchain after being processed by SHA-256 hash.

[0017] The consensus and verification module of the consortium blockchain adopts the PBFT practical Byzantine fault-tolerant consensus mechanism (number of nodes ≥ 4f + 1, where f is the number of faulty nodes) and supports cross-chain verification (such as integration with the EU CBAM registration system and the China Carbon Market MRV platform).

[0018] The smart contract group includes:

[0019] Evidence storage contract: Defines the data on-chain format (including batch ID, timestamp, CID, and hash value), and the trigger condition is that the data collection is completed and passes edge preprocessing verification;

[0020] Carbon asset verification contract: Based on the "Methodology for Voluntary Greenhouse Gas Emission Reduction Projects" (CCER-01-001), it automatically calculates the batch emission reduction (actual carbon emissions of the enterprise - industry benchmark value), generates tradable carbon credit certificates (CDCER), and records the certificate number, issuance date, validity period and trading status;

[0021] Application layer: Used to enable value creation for multiple parties, specifically including:

[0022] Consumer traceability interface: Batch ID is associated with QR code / NFC tag, and the displayed content includes carbon footprint value (kgCO2e / gram), emission reduction comparison with traditional primary gold (percentage), process optimization highlights (such as "smelting energy consumption reduced by 15%), and third-party certification mark;

[0023] Enterprise ESG Dashboard: Integrated carbon emission trend analysis (by daily / weekly / monthly dimensions), automatic generation of emission reduction reports (compliant with ISO14064-1:2018 format), and green finance interface (connected to bank credit systems, dynamically adjusting interest rates based on carbon intensity reduction).

[0024] Regulatory audit interface: Provides a penetrating query function, supports retrieving on-chain hash values ​​and IPFS raw data by batch ID, time range, and subject identity, and outputs standardized reports that comply with CBAM Chapter 4 (Carbon Emission Accounting) and the China Carbon Market MRV System (Monitoring Plan, Report, Verification).

[0025] Preferably, in the industrial identification binding module, the GS1-128 code contains segmented information: the first 6 digits are the recycling area code (refer to GB / T 18354-2021), the middle 8 digits are the batch serial number, and the last 4 digits are the check code, which is attached to the gold waste packaging unit by laser etching or RFID tag.

[0026] Preferably, in the edge intelligent preprocessing module, the number of trees in the isolated forest algorithm is 100-200, the subsample size is 256, and the anomaly scoring threshold is set to 0.6 (scores > 0.6 are judged as abnormal data and marked for manual review).

[0027] Preferably, in the dynamic LCA carbon footprint calculation module, the emission factor library integrates detailed parameters: the regional power grid emission factor corresponding to electricity consumption (e.g., 0.704 kg CO2e / kWh for East China Power Grid), the IPCC Tier 2 factor corresponding to natural gas consumption (0.056 kg CO2e / MJ), and the emission factor corresponding to diesel transportation (2.68 kg CO2e / L).

[0028] Preferably, in the deep reinforcement learning optimization engine, the Actor network output layer consists of three consecutive actions (temperature, oxygen flow rate, and reaction time), and the Critic network uses state-action pair value assessment to guide policy updates. The training data comes from a historical dataset of process parameters and energy consumption (sample size ≥ 100,000).

[0029] Preferably, in the process emission reduction simulation module, the virtual process flow model is constructed using Aspen Plus software and includes energy balance equations (such as smelting thermal efficiency = effective heat / total input heat) and material balance equations (such as gold recovery rate = weight after purification / initial weight of waste).

[0030] Preferably, in the blockchain evidence storage layer, the consortium blockchain nodes authenticate their identities through CA certificates, and the node permissions are set in a hierarchical manner: recyclers / logistics companies can only write data for their own process, refineries can read data from the entire chain, regulatory agencies have read-only and audit permissions, and third-party verification agencies can execute data verification instructions.

[0031] Preferably, in the smart contract group, the emission reduction calculation logic of the carbon asset verification contract is as follows: when the company's actual carbon emissions are less than the industry benchmark value, the emission reduction = (benchmark value - actual carbon emissions) × batch gold weight (grams); when the actual carbon emissions are greater than or equal to the benchmark value, an early warning is triggered and process optimization suggestions are pushed to the company's ESG dashboard.

[0032] Preferably, in the green finance interface of the application layer, the interest rate adjustment rules are as follows: when the carbon intensity (carbon emissions per unit of output) decreases by ≥10% month-on-month, the loan interest rate is reduced by 0.5 percentage points; when the decrease is ≥20%, it is reduced by 1.0 percentage point; when the decrease is <5%, the benchmark interest rate is maintained, and an upward adjustment triggers an ESG risk warning. The regulatory audit interface supports multiple standard switching: when the "EU CBAM mode" is selected, a report including carbon emissions in the transportation process (according to the EU ETS emission factor) and indirect emissions in the refining process (purchased electricity) is output; when the "China carbon market mode" is selected, the output conforms to the format of the "Guidelines for Accounting Methods and Reporting of Greenhouse Gas Emissions of Enterprises (Non-ferrous Metal Smelting)". In the consumer traceability interface, the information associated with the QR code is anonymized and displayed using zero-knowledge proof (ZKP) technology, disclosing only the total carbon footprint and the comparison of emission reductions, without revealing the company's trade secrets (such as specific process parameters and supplier information).

[0033] A blockchain and AI-based method for recording and tracing the carbon footprint of gold recycling includes the following steps:

[0034] S1 Data Acquisition and Binding:

[0035] The IoT sensor array collects data on waste composition and origin during the recycling process, load / mileage / fuel type during transportation, and energy consumption and process parameters during refining.

[0036] A unique batch ID is generated using GS1-128 encoding, and a two-way mapping is established between the batch ID and GPS trajectory, ERP work order, and component data.

[0037] Edge computing nodes are deployed, and the isolated forest algorithm is used to filter out abnormal data, achieving a data accuracy rate of ≥95% after cleaning.

[0038] S2 Carbon Footprint Dynamic Accounting:

[0039] Input real-time energy consumption, material consumption, and transportation parameters into the dynamic LCA model, and integrate the emission factor library of the Ministry of Ecology and Environment and the WGC industry benchmark values;

[0040] Carbon footprint is calculated using the formula: Carbon Footprint = Σ(Energy Consumption in Each Process × Corresponding Emission Factor) + Σ(Transport Load × Mileage × Fuel Emission Factor), with an error rate of ≤5%.

[0041] When the actual carbon emissions are less than the baseline value, CDCE certificates are generated according to the formula: Emission reduction = (Baseline value - Actual carbon emissions) × Gold weight;

[0042] S3 process parameter optimization and control:

[0043] Construct an Actor-Critic network policy model with the reward function being the minimization of energy consumption per unit of gold refining.

[0044] The model is trained based on historical process data and outputs continuous action commands for melting temperature, oxygen flow rate, and reaction time in real time.

[0045] After optimizing the parameters, compare the changes in energy consumption. If the actual energy saving is ≥12%, then update the model parameters.

[0046] S4 Blockchain Trusted Evidence Storage:

[0047] The original data is encrypted and stored in IPFS to generate a CID. The hash value, batch ID, timestamp, and CID are packaged and uploaded to the blockchain.

[0048] The PBFT consensus mechanism is used to complete cross-chain verification, supporting data integration with the EU CBAM system and the Chinese carbon market platform.

[0049] The smart contract automatically triggers the processes of evidence storage, verification, and carbon asset generation, and the execution log is uploaded to the blockchain using SHA-256 hashing.

[0050] S5 Multi-Dimensional Source Tracing Query:

[0051] Batch IDs are read via QR code / NFC to retrieve consortium blockchain evidence data and original IPFS files;

[0052] Display the desensitized carbon footprint values, emission reduction comparisons, process optimization information, and third-party certification labels;

[0053] Regulatory agencies, through authorized authentication, retrieve the entire chain of hash values ​​and raw data to generate standardized audit reports.

[0054] Compared with existing technologies, this invention provides a trusted evidence storage and intelligent traceability system for the carbon footprint of gold recycling based on blockchain and AI technologies, which has the following beneficial effects:

[0055] 1. This gold recycling carbon footprint trusted storage and intelligent traceability system, based on blockchain and AI technologies, completely solves the core pain points of "low data credibility and difficulty in auditing" in gold recycling carbon footprint management through the combination of "blockchain storage + IPFS layered storage + smart contract verification", and builds an industry-level trust foundation. Regarding data immutability, the original data is encrypted and stored in IPFS (generating a unique CID). Carbon footprint hashes and traceability information are processed using SHA-256 and then uploaded to the blockchain (PBFT consensus in the consortium blockchain). Any modification will break hash consistency, meeting the audit requirements of ISO 14064 for "data traceability and non-repudiation". In terms of cross-entity collaborative verification, the consortium blockchain nodes cover recyclers, refineries, logistics providers, and regulatory agencies (such as the Shanghai Environmental Science Research Institute). Through CA certificate hierarchical permissions (such as recyclers only writing their own data, and regulators having the ability to perform penetrating queries), information silos are eliminated. In terms of international standard compatibility, the regulatory audit interface supports switching between multiple standards such as the EU CBAM and the Chinese carbon market MRV (such as the CBAM model outputting the EU ETS emission factor accounting value for the transportation link). The application time is shortened from the traditional 2 weeks to 2 days, and the audit pass rate is 100%. Compared to traditional centralized databases (which are easily tampered with) and single blockchain traceability systems (which only store evidence but do not verify it), this invention is the first to achieve a closed loop of "trustworthy data collection → automatic verification → compliant output", becoming a "passport" for enterprises to cope with green trade barriers (such as EU carbon tariffs).

[0056] 2. This gold recycling carbon footprint trusted evidence storage and intelligent traceability system, based on blockchain and AI technologies, transforms environmental protection investments into quantifiable economic benefits through an economic closed loop of "AI-precise accounting + process optimization + carbon assetization," solving the problem of "insufficient motivation for emission reduction." Regarding significant reductions in operating costs, the dynamic LCA model (error rate ≤5%) replaces industry average estimations, avoiding redundant costs caused by "over-calculation"; the deep reinforcement learning optimization engine uses "unit gold purification energy consumption" as the reward function to dynamically adjust parameters such as smelting temperature and oxygen flow (e.g., optimized from 1300℃ to 1250℃ in the example), achieving a 12-18% reduction in energy consumption (annualized electricity savings exceeding one million yuan); the process emission reduction simulation module (Aspen Plus constructs a virtual process flow) pre-evaluates the ROI of technical improvement solutions (e.g., the ROI of replacing the induction melting furnace reaches 12.8%), avoiding blind investment. Regarding the realization of carbon assets and their financial value, carbon asset certification contracts automatically generate CDCE certificates based on the CCER methodology (e.g., a batch emission reduction of 175.6 kg CO2e in the example), supporting carbon trading (unit price 50 yuan / kg, quarterly profit exceeding one million yuan); the green finance interface links the reduction in carbon intensity to loan interest rates (e.g., a 15.7% reduction triggers a 0.5% interest rate reduction), resulting in annualized interest savings of over 250,000 yuan. Compared to the dilemma of traditional solutions where "emission reduction only increases costs," this invention enables enterprises to achieve a positive cycle of "emission reduction → cost reduction → revenue increase," with an annualized comprehensive return increase of over 30%.

[0057] 3. This gold recycling carbon footprint trusted storage and intelligent traceability system, based on blockchain and AI technologies, injects new momentum into the recycled precious metals industry through "standardized tools + multi-party collaborative interfaces," promoting the value reconstruction of the entire industry chain. Regarding inclusive empowerment of SMEs, the system supports flexible "cloud-edge-device" deployment (such as small recyclers sharing edge gateways and adopting SaaS-based AI services), simplifies data entry items at the data collection layer (only weight, purity, and GPS location are required), and keeps the carbon accounting error rate below 6%. Small and micro enterprises with an annual processing capacity of 500 tons can profit over one million yuan through carbon trading, breaking down the barrier that "only large enterprises can manage carbon effectively." In terms of supply chain collaborative transformation, industrial identification binding (GS1-128 coding "one code to the end") connects the entire process of recycling, logistics, and refining. Consumers scan codes for traceability (zero-knowledge proof desensitization display of carbon footprint and emission reduction comparison) to increase the premium of recycled gold jewelry (e.g., the "79% emission reduction compared to virgin gold" label drives a 20% increase in sales), forcing upstream recyclers to standardize data entry. Regulatory audit interfaces support cross-chain verification (e.g., connection with the Singapore carbon trading platform), helping Chinese companies participate in the international carbon market. In terms of ecology, the system promotes the replacement of virgin gold with recycled gold (carbon emissions are only 1 / 30 of virgin gold). According to the example calculation, processing 5,000 tons of waste gold annually can reduce the land occupation (approximately 100 hectares) and water consumption (approximately 5 million tons) corresponding to mining. At the same time, the ESG dashboard automatically generates ISO 14064 reports, helping companies attract green investment through ESG ratings (e.g., MSCI rating improvement), forming a positive feedback loop of "corporate emission reduction - capital favor - industrial upgrading". Detailed Implementation

[0058] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example

[0060] An Example of a Trusted Evidence-Building and Intelligent Traceability System for Gold Recycling Carbon Footprint Based on Blockchain and AI Technologies

[0061] A trusted and intelligent traceability system for the carbon footprint of gold recycling, based on blockchain and AI technologies, comprises a data acquisition layer, an AI processing layer, a blockchain storage layer, and an application layer that interact sequentially.

[0062] Data Acquisition Layer: Used to achieve multi-source acquisition of data throughout the entire lifecycle of gold recycling, industrial identifier binding, and edge intelligent preprocessing, specifically including:

[0063] The multi-source heterogeneous data acquisition module, through an IoT sensor group (including an XRF spectrometer, smart meter, and GPS locator), a mobile APP (supporting individual recyclers to enter the weight, purity, time, and geographical location of the recycled materials), and an ERP / GPS interface (connecting to the TMS system of transportation companies and the MES system of refineries), collects data on the composition and origin of waste materials in the recycling process, the load / mileage / fuel type in the transportation process, and the energy consumption (electricity / gas), material consumption (chemical usage), and process parameters (melting temperature, oxygen flow rate, and reaction time) in the refining process.

[0064] The industrial identification binding module uses the GS1-128 coding standard to assign a unique batch ID to each batch of recycled gold. Through the API interface, the batch ID is mapped to GPS trajectory data, ERP work order number, and XRF spectrometer detection data to achieve full-process data connectivity with a single code.

[0065] The edge intelligence preprocessing module is deployed at the data acquisition source (such as IoT gateway, mobile APP local), and uses the isolated forest algorithm or LOF algorithm for anomaly detection. It cleans sensor data in real time with ±3σ as the noise filtering threshold, improving the data accuracy to ≥95%.

[0066] AI processing layer: used to achieve accurate carbon footprint calculation, intelligent optimization of process parameters, and simulation of emission reduction potential, specifically including:

[0067] The dynamic LCA carbon footprint calculation module constructs a batch-level accounting model based on the Life Cycle Assessment (LCA) theory. It integrates the emission factor library of the Ministry of Ecology and Environment's "Guidelines for the Compilation of Provincial Greenhouse Gas Inventories" and the industry benchmark library of the World Gold Council (WGC). Input parameters include real-time energy consumption (unit: kWh), material consumption (unit: kg), transport load (unit: ton), fuel type (diesel / gasoline / biodiesel), and mileage (unit: km). The output is the batch carbon footprint (unit: kg CO2e), with an accounting error rate of ≤5%.

[0068] The deep reinforcement learning optimization engine uses "minimizing energy consumption per unit of gold purification" as the reward function and constructs a policy model containing an Actor-Critic network to dynamically adjust the smelting temperature (range 1100-1350℃), oxygen flow rate (range 150-250m³ / h), and reaction time (range 2-4 hours) in the refining process, thereby reducing energy consumption by 12%-18%.

[0069] The process emission reduction simulation module has a built-in virtual process flow model (covering crushing, smelting, and electrolytic purification sub-modules). It compares the company's actual carbon emissions with WGC benchmark values ​​(such as 600 kg CO2e / ounce carbon emissions from primary gold smelting) and pre-evaluates the carbon emission reduction and return on investment (ROI) of technical improvement plans (such as replacing induction melting furnaces and optimizing chemical ratios).

[0070] Blockchain evidence storage layer: used to achieve trusted data storage, automatic carbon asset verification, and cross-entity collaborative verification, specifically including:

[0071] The hierarchical storage module adopts an architecture that combines a consortium blockchain (nodes include gold recyclers, refineries, logistics companies, regulatory agencies, and third-party verification agencies) with IPFS off-chain storage. The raw data (sensor logs, APP input records, ERP work orders) is encrypted and stored in IPFS and a content identifier (CID) is generated. The carbon footprint calculation results, traceability information, and smart contract execution logs are uploaded to the blockchain after being processed by SHA-256 hash.

[0072] The consensus and verification module of the consortium blockchain adopts the PBFT practical Byzantine fault-tolerant consensus mechanism (number of nodes ≥ 4f + 1, where f is the number of faulty nodes) and supports cross-chain verification (such as integration with the EU CBAM registration system and the China Carbon Market MRV platform).

[0073] The smart contract group includes:

[0074] Evidence storage contract: Defines the data on-chain format (including batch ID, timestamp, CID, and hash value), and the trigger condition is that the data collection is completed and passes edge preprocessing verification;

[0075] Carbon asset verification contract: Based on the "Methodology for Voluntary Greenhouse Gas Emission Reduction Projects" (CCER-01-001), it automatically calculates the batch emission reduction (actual carbon emissions of the enterprise - industry benchmark value), generates tradable carbon credit certificates (CDCER), and records the certificate number, issuance date, validity period and trading status;

[0076] Application layer: Used to enable value creation for multiple parties, specifically including:

[0077] Consumer traceability interface: Batch ID is associated with QR code / NFC tag, and the displayed content includes carbon footprint value (kgCO2e / gram), emission reduction comparison with traditional primary gold (percentage), process optimization highlights (such as "smelting energy consumption reduced by 15%), and third-party certification mark;

[0078] Enterprise ESG Dashboard: Integrated carbon emission trend analysis (by daily / weekly / monthly dimensions), automatic generation of emission reduction reports (compliant with ISO14064-1:2018 format), and green finance interface (connected to bank credit systems, dynamically adjusting interest rates based on carbon intensity reduction).

[0079] Regulatory audit interface: Provides a penetrating query function, supports retrieving on-chain hash values ​​and IPFS raw data by batch ID, time range, and subject identity, and outputs standardized reports that comply with CBAM Chapter 4 (Carbon Emission Accounting) and the China Carbon Market MRV System (Monitoring Plan, Report, Verification).

[0080] Specifically, in the industrial identification binding module, the GS1-128 code contains segmented information: the first 6 digits are the recycling area code (refer to GB / T 18354-2021), the middle 8 digits are the batch serial number, and the last 4 digits are the check code, which are attached to the gold scrap packaging unit by laser etching or RFID tag.

[0081] Specifically, in the edge intelligence preprocessing module, the number of trees in the isolated forest algorithm is 100-200, the subsample size is 256, and the anomaly scoring threshold is set to 0.6 (scores > 0.6 are judged as abnormal data and marked for manual review).

[0082] Specifically, in the dynamic LCA carbon footprint calculation module, the emission factor library integrates detailed parameters: electricity consumption corresponds to the regional power grid emission factor (such as 0.704 kg CO2e / kWh for East China Power Grid), natural gas consumption corresponds to the IPCC Tier 2 factor (0.056 kg CO2e / MJ), and diesel transportation corresponds to 2.68 kg CO2e / L.

[0083] Specifically, in the deep reinforcement learning optimization engine, the Actor network output layer consists of three consecutive actions (temperature, oxygen flow rate, and reaction time), and the Critic network uses state-action pair value assessment to guide policy updates. The training data comes from a historical dataset of process parameters and energy consumption (sample size ≥ 100,000).

[0084] Specifically, in the process emission reduction simulation module, the virtual process flow model is constructed using Aspen Plus software, which includes energy balance equations (such as smelting thermal efficiency = effective heat / total input heat) and material balance equations (such as gold recovery rate = weight after purification / initial weight of waste).

[0085] Specifically, in the blockchain evidence storage layer, consortium blockchain nodes authenticate their identities through CA certificates, and node permissions are set in a hierarchical manner: recyclers / logistics companies can only write data for their own process, refineries can read data from the entire chain, regulatory agencies have read-only and audit permissions, and third-party verification agencies can execute data verification instructions.

[0086] Specifically, in the smart contract group, the calculation logic for the emission reduction of the carbon asset verification contract is as follows: when the company's actual carbon emissions are less than the industry benchmark, the emission reduction = (benchmark value - actual carbon emissions) × batch gold weight (grams); when the actual carbon emissions are greater than or equal to the benchmark, an early warning is triggered and process optimization suggestions are pushed to the company's ESG dashboard.

[0087] Specifically, in the green finance interface at the application layer, the interest rate adjustment rules are as follows: when carbon intensity (carbon emissions per unit of output) decreases by ≥10% month-on-month, the loan interest rate is reduced by 0.5 percentage points; when the decrease is ≥20%, it is reduced by 1.0 percentage point; when the decrease is <5%, the benchmark interest rate is maintained, and an increase triggers an ESG risk warning. The regulatory audit interface supports multiple standard switching: when the "EU CBAM mode" is selected, the output includes reports on carbon emissions in the transportation process (according to the EU ETS emission factor) and indirect emissions in the refining process (purchased electricity); when the "China carbon market mode" is selected, the output conforms to the format of the "Guidelines for Accounting Methods and Reporting of Greenhouse Gas Emissions of Enterprises (Non-ferrous Metal Smelting)". In the consumer traceability interface, the information associated with the QR code is anonymized and displayed using zero-knowledge proof (ZKP) technology, disclosing only the total carbon footprint and the comparison of emission reductions, without revealing the company's trade secrets (such as specific process parameters and supplier information).

[0088] A blockchain and AI-based method for recording and tracing the carbon footprint of gold recycling includes the following steps:

[0089] S1 Data Acquisition and Binding:

[0090] The IoT sensor array collects data on waste composition and origin during the recycling process, load / mileage / fuel type during transportation, and energy consumption and process parameters during refining.

[0091] A unique batch ID is generated using GS1-128 encoding, and a two-way mapping is established between the batch ID and GPS trajectory, ERP work order, and component data.

[0092] Edge computing nodes are deployed, and the isolated forest algorithm is used to filter out abnormal data, achieving a data accuracy rate of ≥95% after cleaning.

[0093] S2 Carbon Footprint Dynamic Accounting:

[0094] Input real-time energy consumption, material consumption, and transportation parameters into the dynamic LCA model, and integrate the emission factor library of the Ministry of Ecology and Environment and the WGC industry benchmark values;

[0095] Carbon footprint is calculated using the formula: Carbon Footprint = Σ(Energy Consumption in Each Process × Corresponding Emission Factor) + Σ(Transport Load × Mileage × Fuel Emission Factor), with an error rate of ≤5%.

[0096] When the actual carbon emissions are less than the baseline value, CDCE certificates are generated according to the formula: Emission reduction = (Baseline value - Actual carbon emissions) × Gold weight;

[0097] S3 process parameter optimization and control:

[0098] Construct an Actor-Critic network policy model with the reward function being the minimization of energy consumption per unit of gold refining.

[0099] The model is trained based on historical process data and outputs continuous action commands for melting temperature, oxygen flow rate, and reaction time in real time.

[0100] After optimizing the parameters, compare the changes in energy consumption. If the actual energy saving is ≥12%, then update the model parameters.

[0101] S4 Blockchain Trusted Evidence Storage:

[0102] The original data is encrypted and stored in IPFS to generate a CID. The hash value, batch ID, timestamp, and CID are packaged and uploaded to the blockchain.

[0103] The PBFT consensus mechanism is used to complete cross-chain verification, supporting data integration with the EU CBAM system and the Chinese carbon market platform.

[0104] The smart contract automatically triggers the processes of evidence storage, verification, and carbon asset generation, and the execution log is uploaded to the blockchain using SHA-256 hashing.

[0105] S5 Multi-Dimensional Source Tracing Query:

[0106] Batch IDs are read via QR code / NFC to retrieve consortium blockchain evidence data and original IPFS files;

[0107] Display the desensitized carbon footprint values, emission reduction comparisons, process optimization information, and third-party certification labels;

[0108] Regulatory agencies, through authorized authentication, retrieve the entire chain of hash values ​​and raw data to generate standardized audit reports.

[0109] Through the above technical solution, this invention, by combining "blockchain notarization + IPFS hierarchical storage + smart contract verification," completely solves the core pain points of "low data credibility and difficulty in auditing" in gold recycling carbon footprint management, thus constructing an industry-level trust foundation. Regarding data immutability, the original data is encrypted and stored in IPFS (generating a unique CID). Carbon footprint hashes and traceability information are processed using SHA-256 and then uploaded to the blockchain (PBFT consensus in the consortium blockchain). Any modification will destroy hash consistency, meeting the ISO 14064 audit requirements for "data traceability and non-repudiation." Regarding cross-entity collaborative verification, the consortium blockchain nodes cover recyclers, refineries, logistics providers, and regulatory agencies (such as the Shanghai Environmental Science Research Institute). Through CA certificate hierarchical permissions (e.g., recyclers only write their own data, while regulators can perform penetrating queries), information silos are eliminated. Regarding international standard compatibility, the regulatory audit interface supports switching between multiple standards such as the EU CBAM and the Chinese carbon market MRV (e.g., the CBAM model outputs the EU ETS emission factor calculation value for the transportation link), shortening the application time from the traditional two weeks to two days, with a 100% audit pass rate. Compared to traditional centralized databases (easily tampered with) and single blockchain traceability systems (only storing evidence without verification), this invention achieves a closed loop of "trustworthy data collection → automatic verification → compliant output" for the first time, becoming a "passport" for enterprises to cope with green trade barriers (such as EU carbon tariffs). Through an economic closed loop of "AI-precise accounting + process optimization + carbon assetization," environmental protection investment is transformed into quantifiable economic benefits, solving the problem of "insufficient motivation for emission reduction." Regarding significant reductions in operating costs, the dynamic LCA model (error rate ≤5%) replaces industry average estimation, avoiding redundant costs caused by "over-calculation"; the deep reinforcement learning optimization engine uses "energy consumption per unit of gold purification" as the reward function, dynamically adjusting parameters such as smelting temperature and oxygen flow (e.g., optimizing from 1300℃ to 1250℃ in the example), achieving a 12-18% reduction in energy consumption (annualized electricity savings exceeding one million yuan); the process emission reduction simulation module (Aspen Plus constructs a virtual process flow) pre-evaluates the ROI of technical improvement solutions (e.g., the ROI of replacing the induction melting furnace reaches 12.8%), avoiding blind investment. Regarding the realization of carbon assets and their financial value, carbon asset certification contracts automatically generate CDCE certificates based on the CCER methodology (e.g., a batch emission reduction of 175.6 kg CO2e in the example), supporting carbon trading (unit price 50 yuan / kg, quarterly profit exceeding one million yuan); the green finance interface links the reduction in carbon intensity to loan interest rates (e.g., a 15.7% reduction triggers a 0.5% interest rate reduction), saving over 250,000 yuan in interest expenses annually. Compared to the dilemma of traditional solutions where "emission reduction only increases costs," this invention enables enterprises to achieve a positive cycle of "emission reduction → cost reduction → revenue increase," with an annualized comprehensive return increase of over 30%. Through "standardized tools + multi-party collaborative interfaces," it injects new momentum into the recycled precious metals industry through digitalization, greening, and financialization, promoting the value reconstruction of the entire industry chain.In terms of empowering SMEs, the system supports flexible deployment across the cloud, edge, and terminal (e.g., small recyclers sharing edge gateways and adopting SaaS-based AI services). The data collection layer simplifies data entry (only weight, purity, and GPS location are required), and the carbon accounting error rate remains below 6%. Small and micro enterprises with an annual processing capacity of 500 tons can profit over one million yuan through carbon trading, breaking down the barrier that "only large enterprises can manage carbon effectively." Regarding supply chain collaborative transformation, industrial label binding (GS1-128 coding for end-to-end consistency) connects the entire process of recycling, logistics, and refining. Consumers scan codes for traceability (zero-knowledge proof-based desensitized display of carbon footprint and emission reduction comparisons) increases the premium of recycled gold jewelry (e.g., a "79% emission reduction compared to virgin gold" label drives a 20% increase in sales), forcing upstream recyclers to standardize data entry. The regulatory audit interface supports cross-chain verification (e.g., integration with the Singapore carbon trading platform), helping Chinese enterprises participate in the international carbon market. In terms of ecology, the system promotes the replacement of primary gold with recycled gold (carbon emissions are only 1 / 30 of primary gold). According to the implementation example, processing 5,000 tons of waste gold annually can reduce the land occupation (approximately 100 hectares) and water consumption (approximately 5 million tons) corresponding to mining. At the same time, the ESG dashboard automatically generates ISO 14064 reports, helping companies attract green investment through ESG ratings (such as MSCI rating upgrades), forming a positive ecological feedback loop of "corporate emission reduction - capital favor - industrial upgrading".

[0110] Implementation details of each layer

[0111] 1. Data Acquisition Layer: "One Code to the End" Data Integration

[0112] Step 1: Industrial Identification Binding

[0113] Encoding rules: Adopt the GS1-128 standard, the format is (01)69412345678903(10)20240520001, where:

[0114] (01) is the application identifier (AI), followed by a 13-digit manufacturer identification code (e.g., “6941234567890”, which the company applies for from GS1China).

[0115] (10) is the batch number AI, followed by an 8-digit serial number (e.g., “20240520” is the date, and “001” is the batch number for that day).

[0116] The checksum is generated using the GS1 Check Digit Calculator (in this example, the checksum is 3).

[0117] Identification: The code is laser-etched onto the gold scrap packaging iron box (size 50cm×50cm×30cm), and an RFID tag (NXP UCODE 8, storage capacity 512bit) is generated and attached to the box, supporting long-distance reading.

[0118] Step 2: Multi-source data acquisition and edge preprocessing

[0119] In the recycling process: Individual recyclers use a commercial APP to scan the RFID tag on the metal box and enter the recycling weight (e.g., "10.5kg"), purity (Au content of 99.2% as detected by XRF spectrometer), and origin (GPS location "Zhangjiang Road, Pudong New Area, Shanghai"). When the data passes through the edge gateway, the isolated forest algorithm (150 trees, 256 subsamples, and anomaly scoring threshold of 0.6) filters out abnormal values ​​(e.g., a mistakenly recorded "100kg" is marked as abnormal and requires manual verification).

[0120] During transportation: The GPS locator of the logistics vehicle uploads its latitude and longitude (e.g., "31.2304°N, 121.4737°E") every 5 minutes, and the load sensor transmits the weight in real time (e.g., "8.2 tons, after deducting the empty vehicle's own weight of 1.3 tons, the actual load is 6.9 tons). The fuel type (diesel, grade 0#) is selected and entered through the driver's APP.

[0121] Refining process: The PLC controller of the smelting furnace collects temperature (e.g., "1280℃"), oxygen flow rate (e.g., "210m³ / h"), and reaction time (cumulative "2.5 hours") every 10 minutes; the smart meter records the power consumption (e.g., "1850kWh / ton") and the gas meter records the natural gas consumption (e.g., "120m³ / ton").

[0122] Step 3: Data Linking and Uploading

[0123] The edge gateway uploads the pre-processed data (including batch ID, timestamp, and sensor values) to the cloud Kafka message queue via the MQTT protocol, waiting for the AI ​​processing layer to consume it.

[0124] 2. AI Processing Layer: Precise Calculation and Intelligent Optimization

[0125] Step 1: Dynamic LCA Carbon Footprint Calculation

[0126] Model input: Batch data extracted from a Kafka queue (e.g., batch ID "20240520001", weight 10.5kg):

[0127] Recycling process: XRF testing showed Au 99.2% and a color loss of 0.8% (corresponding to 0.1 kg CO2e / kg carbon emissions, according to WGC standards).

[0128] Transportation: Load capacity 6.9 tons, distance 50km (Shanghai Zhangjiang → Jiangsu Kunshan Refinery), diesel fuel consumption 0.3L / km (actual measurement), emission factor 2.68kg CO2e / L (IPCC Tier 2), carbon emissions = 6.9 tons × 50km × 0.3L / km × 2.68kg / L ÷ 1000 = 27.7kg CO2e;

[0129] Refining stage: Electricity consumption 1850 kWh / ton (10.5 kg corresponds to 19.425 kWh), East China Power Grid emission factor 0.704 kg CO2e / kWh (data from the Ministry of Ecology and Environment in 2023), carbon emissions = 19.425 × 0.704 = 13.7 kg CO2e; Natural gas 120 m³ / ton (10.5 kg corresponds to 1.26 m³), ​​IPCC factor 0.056 kg CO2e / MJ (calorific value of natural gas 36 MJ / m³), carbon emissions = 1.26 × 36 × 0.056 = 2.5 kg CO2e.

[0130] Model output: Total carbon footprint = 0.1 × 10.5 (recycling) + 27.7 (transportation) + 13.7 + 2.5 (refining) = 46.4 kg CO2e, with an error rate of ≤5% (compared with the measured value of 48.2 kg by a third-party verification agency, the error is 3.7%).

[0131] Step 2: Deep Reinforcement Learning (DRL) Process Optimization

[0132] Model Training: The Actor-Critic network was trained using the company's historical data from January to June 2023 (100,000 samples, including correlation data on temperature, oxygen flow rate, reaction time, and energy consumption).

[0133] Actor Network (Policy Network): Input state (current energy consumption, gold recovery rate), output action (temperature adjustment ΔT, oxygen flow rate adjustment ΔQ, reaction time adjustment Δt);

[0134] Critic Network (Value Network): Evaluates the value of state-action pairs to guide the Actor Network in updating its strategy.

[0135] Real-time optimization: For batch "20240520001", the original process (1300℃, 220m³ / h, 3h) had an energy consumption of 19.425kWh. The DRL model output optimized parameters (1250℃, 200m³ / h, 2.5h), predicting that the energy consumption would be reduced to 16.2kWh (a reduction of 16.6%). After actual execution, the measured energy consumption was 16.5kWh, verifying the effectiveness of the optimization.

[0136] Step 3: Process Emission Reduction Simulation

[0137] A virtual process flow model was built using Aspen Plus to simulate a technical upgrade scheme of "replacing the induction melting furnace (thermal efficiency from 65% to 80%)". After inputting the parameters of the new equipment, the model output that the energy consumption of the refining process is reduced to 14.8 kWh / ton, and the annual emission reduction is (19.425-14.8)×5000 tons=23,100 kWh, corresponding to a carbon emission reduction of 23,100×0.704=16,300 kg CO2e. The return on investment (ROI) is (carbon trading revenue + energy savings) / equipment cost=(16,300×50 yuan / kg + 23,100×0.8 yuan / kWh) / 2 million yuan=12.8% (feasible).

[0138] 3. Blockchain Evidence Storage Layer: Trusted Evidence Storage and Carbon Asset Verification

[0139] Step 1: Consortium Blockchain Deployment and Node Management

[0140] Node composition: 4 consensus nodes (enterprises, refineries, logistics providers, regulatory agencies (Shanghai Environmental Science Research Institute)) + 2 observation nodes (third-party verification agencies, carbon exchanges), using PBFT consensus (number of nodes 4 = 3f + 1, f = 1, tolerating 1 malicious node).

[0141] Identity authentication: Each node registers using a CA certificate (issued by CFCA), with tiered access control.

[0142] Enterprises / logistics providers: can only write data for their own process (e.g., enterprises write refined data, logistics providers write transportation data).

[0143] Refinery: Reads full-chain data (for quality traceability);

[0144] Regulatory authorities: Read-only + audit permissions (can access all on-chain data);

[0145] Verification authorities: execute data verification instructions (such as comparing the original IPFS data with the on-chain hash).

[0146] Step 2: Data upload to the blockchain and smart contract execution

[0147] Off-chain storage: Raw data (such as XRF detection reports, electricity meter reading screenshots) is encrypted (AES-256) and stored in IPFS to generate a CID (such as "QmXoypizjW3WknFiJnKLwHCnL72vedxjQkDDP1mXWo6uco").

[0148] On-chain evidence storage: The evidence storage contract (written in Solidity) is automatically triggered to package the batch ID, timestamp (e.g., 1716182400, corresponding to 2024-05-20 08:00:00), CID, carbon footprint hash (SHA-256(46.4kg CO2e)=a1b2c3...) onto the chain, and the transaction hash is "0x5f3d...".

[0149] Carbon Asset Verification: The verification contract is based on the CCER-01-001 methodology. The emission reduction is calculated as follows: emission reduction = industry benchmark value (WGC standard 150kg CO2e / ton) - actual carbon emission. The total carbon emission of a batch of 10.5kg is 46.4kg, which is 4.42kg / kg = 4420kg / ton. The carbon emission of primary gold is about 600kg / ounce = 21148kg / ton. Therefore, the emission reduction = 21148 - 4420 = 16728kg CO2e / ton. The emission reduction of this batch = 16728 × 0.0105 tons = 175.6kg CO2e. A CDCE certificate (number CDCE-2024-0520-001, valid for 3 years) is generated, and the application layer green finance interface is triggered.

[0150] 4. Application Layer: Enabling Value from Multiple Parties

[0151] Step 1: Consumer Traceability

[0152] When recycled gold jewelry leaves the factory, a QR code (associated with batch ID "20240520001") is printed on the product tag. After the consumer scans the code, the product is desensitized using zero-knowledge proof (ZKP) technology.

[0153] Carbon footprint: 4.42 kg CO2e / 100g (i.e., 44.2 kg CO2e / kg, total carbon emissions of this batch of 10.5 kg is 46.4 kg).

[0154] Emissions reduction comparison: "79% reduction compared to primary gold (21,148 kg / ton primary gold vs. 4,420 kg / ton this product)";

[0155] Process highlights: "Smelting energy consumption reduced by 16.6%, using an induction melting furnace (thermal efficiency 80%)";

[0156] Certification mark: "Certified by Shanghai Academy of Environmental Sciences, CDER certificate number CDER-2024-0520-001".

[0157] Step 2: Enterprise ESG Dashboard

[0158] Kanban integrates Tableau visualization tools to display:

[0159] Carbon emission trend: In the past 6 months, carbon intensity has decreased from 5.1 kg CO2e / 10,000 yuan of output value to 4.3 kg (a decrease of 15.7%).

[0160] Emissions reduction report: Automatically generates ISO 14064-1 format report, including boundary declaration, data sources, and uncertainty analysis;

[0161] Green finance: Due to a 15.7% decrease in carbon intensity, the interest rate reduction rule was triggered (a 0.5% reduction is applied for a decrease of ≥10%). The annual interest rate on bank loans was reduced from 5.0% to 4.5%, resulting in annual interest savings of (loan amount of 50 million) × 0.5% = 250,000 yuan.

[0162] Step 3: Regulatory Audit

[0163] The regulatory agency logs into the audit interface, selects "EU CBAM mode," and enters the batch ID "20240520001." The system outputs:

[0164] Carbon emissions during transportation: 27.7 kg CO2e (calculated based on the EU ETS diesel factor of 2.68 kg / L);

[0165] Indirect emissions from the refining process: 13.7 kg CO2e (for purchased electricity, calculated as 6.8 kg based on the EU grid factor of 0.35 kg CO2e / kWh, taking the higher value).

[0166] Total carbon emissions: 27.7 + 6.8 = 34.5 kg CO2e (meets CBAM application requirements).

[0167] Multi-scenario application case expansion

[0168] Case 1: Small-scale recycler integration (annual processing capacity of 500 tons)

[0169] Simplified deployment: Shared edge gateway (provided by the regional recycling alliance), simplified data entry items in the mobile APP (only weight, condition, and GPS location are retained), AI processing layer adopts cloud SaaS service (no local GPU required), and blockchain nodes only serve as observation nodes (data is written by alliance nodes).

[0170] Results: The carbon accounting error rate is still controlled within 6% (slightly higher than 5% for medium-sized enterprises due to the small amount of data), with an annual emission reduction of 1200 tons × 16728 kg / ton = 20,100 tons of CO2e, and a profit of 1.005 million yuan through carbon trading.

[0171] Case 2: Cross-border Recycling (China → Southeast Asia)

[0172] Challenges: Fragmented cross-border logistics data (Vietnamese logistics providers lack GPS interfaces), and differences in emission factors (Southeast Asian power grid factors are higher than those in China).

[0173] Solution:

[0174] Logistics data: Vietnamese logistics companies manually enter load / mileage into the commercial APP, and the system automatically matches the local emission factors (such as Vietnamese diesel factor 2.75kg CO2e / L).

[0175] Cross-chain verification: Through the Polkadot cross-chain protocol, data from Chinese consortium blockchains is synchronized to the Singapore carbon trading platform blockchain, supporting verification by overseas buyers.

[0176] Key Issues and Solutions

[0177] Data interruption issue: Unstable network for individual recyclers causes data upload delays → Data is cached locally on the edge gateway (up to 7 days), and automatically re-uploaded after the network is restored.

[0178] Model generalization problem: Differences in processes among different enterprises cause fluctuations in DRL optimization results → The AI ​​processing layer has a built-in transfer learning module that fine-tunes the model based on a small amount of data (500 samples) from new enterprises, achieving optimization results within 3 days.

[0179] Privacy concerns: Enterprises worry about the leakage of process parameters → Blockchain uses zero-knowledge proofs (ZKP), the application layer only displays the de-identified carbon footprint, and core parameters are only visible to regulatory agencies.

[0180] Summary of Implementation Results

[0181] Through the above deployment, the company achieved the following within 6 months:

[0182] Carbon footprint accuracy: Batch-level accounting error rate ≤5%, far exceeding the industry average (20%+);

[0183] Reduced energy consumption: Energy consumption in the refining process decreased by 16.6%, resulting in annualized electricity savings of 1.85 million yuan;

[0184] Carbon asset returns: Quarterly emission reduction of 4,380 tons × 16,728 kg / ton = 73.22 million kg CO2e, generating 73.22 million CDERs, with a trading profit of 3.66 million yuan;

[0185] Compliance efficiency: The EU CBAM filing time has been shortened from 2 weeks to 2 days, and the audit pass rate is 100%.

[0186] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A trusted and intelligent traceability system for carbon footprint verification in gold recycling based on blockchain and AI technologies, characterized by: It includes a data acquisition layer, an AI processing layer, a blockchain evidence storage layer, and an application layer that interact sequentially, among which: Data Acquisition Layer: Used to achieve multi-source acquisition of data throughout the entire lifecycle of gold recycling, industrial identifier binding, and edge intelligent preprocessing, specifically including: The multi-source heterogeneous data acquisition module, through an IoT sensor group (including an XRF spectrometer, smart meter, and GPS locator), a mobile APP (supporting individual recyclers to enter the weight, purity, time, and geographical location of the recycled materials), and an ERP / GPS interface (connecting to the TMS system of transportation companies and the MES system of refineries), collects data on the composition and origin of waste materials in the recycling process, the load / mileage / fuel type in the transportation process, and the energy consumption (electricity / gas), material consumption (chemical usage), and process parameters (melting temperature, oxygen flow rate, and reaction time) in the refining process. The industrial identification binding module uses the GS1-128 coding standard to assign a unique batch ID to each batch of recycled gold. Through the API interface, the batch ID is mapped bidirectionally with GPS trajectory data, ERP work order number, and XRF spectrometer detection component data to achieve end-to-end data connectivity with a single code. The edge intelligence preprocessing module is deployed at the data acquisition source (such as IoT gateway, mobile APP local), and uses the isolated forest algorithm or LOF algorithm for anomaly detection. It cleans sensor data in real time with ±3σ as the noise filtering threshold, improving the data accuracy to ≥95%. AI processing layer: used to achieve accurate carbon footprint calculation, intelligent optimization of process parameters, and simulation of emission reduction potential, specifically including: The dynamic LCA carbon footprint calculation module constructs a batch-level accounting model based on the Life Cycle Assessment (LCA) theory. It integrates the emission factor library of the Ministry of Ecology and Environment's "Guidelines for the Compilation of Provincial Greenhouse Gas Inventories" and the industry benchmark library of the World Gold Council (WGC). Input parameters include real-time energy consumption (unit: kWh), material consumption (unit: kg), transport load (unit: ton), fuel type (diesel / gasoline / biodiesel), and mileage (unit: km). The output is the batch carbon footprint (unit: kg CO2e), with an accounting error rate of ≤5%. The deep reinforcement learning optimization engine uses "minimizing energy consumption per unit of gold purification" as the reward function and constructs a policy model containing an Actor-Critic network to dynamically adjust the smelting temperature (range 1100-1350℃), oxygen flow rate (range 150-250m³ / h), and reaction time (range 2-4 hours) in the refining process, achieving a 12%-18% reduction in energy consumption. The process emission reduction simulation module has a built-in virtual process flow model (covering crushing, smelting, and electrolytic purification sub-modules). It compares the company's actual carbon emissions with WGC benchmark values ​​(such as 600 kg CO2e / ounce carbon emissions from primary gold smelting) and pre-evaluates the carbon emission reduction and return on investment (ROI) of technical improvement plans (such as replacing induction melting furnaces and optimizing chemical ratios). Blockchain evidence storage layer: used to achieve trusted data storage, automatic carbon asset verification, and cross-entity collaborative verification, specifically including: The hierarchical storage module adopts an architecture that combines a consortium blockchain (nodes include gold recyclers, refineries, logistics companies, regulatory agencies, and third-party verification agencies) with IPFS off-chain storage. The raw data (sensor logs, APP input records, ERP work orders) is encrypted and stored in IPFS and a content identifier (CID) is generated. The carbon footprint calculation results, traceability information, and smart contract execution logs are uploaded to the blockchain after being processed by SHA-256 hash. The consensus and verification module of the consortium blockchain adopts the PBFT practical Byzantine fault-tolerant consensus mechanism (number of nodes ≥ 4f + 1, where f is the number of faulty nodes) and supports cross-chain verification (such as integration with the EU CBAM registration system and the China Carbon Market MRV platform). The smart contract group includes: Evidence storage contract: Defines the data on-chain format (including batch ID, timestamp, CID, and hash value), and the trigger condition is that the data collection is completed and passes edge preprocessing verification; Carbon asset verification contract: Based on the "Methodology for Voluntary Greenhouse Gas Emission Reduction Projects" (CCER-01-001), it automatically calculates the batch emission reduction (actual carbon emissions of the enterprise - industry benchmark value), generates tradable carbon credit certificates (CDCER), and records the certificate number, issuance date, validity period and trading status; Application layer: Used to enable value creation for multiple parties, specifically including: Consumer traceability interface: Batch ID is associated with QR code / NFC tag, and the displayed content includes carbon footprint value (kgCO2e / gram), emission reduction comparison with traditional primary gold (percentage), process optimization highlights (such as "15% reduction in smelting energy consumption"), and third-party certification mark; Enterprise ESG Dashboard: Integrated carbon emission trend analysis (by daily / weekly / monthly dimensions), automatic generation of emission reduction reports (compliant with ISO14064-1:2018 format), and green finance interface (connected to bank credit systems, dynamically adjusting interest rates based on carbon intensity reduction). Regulatory audit interface: Provides a penetrating query function, supports retrieving on-chain hash values ​​and IPFS raw data by batch ID, time range, and subject identity, and outputs standardized reports that comply with CBAM Chapter 4 (Carbon Emission Accounting) and the China Carbon Market MRV System (Monitoring Plan, Report, Verification).

2. The trusted evidence storage and intelligent traceability system for gold recycling carbon footprint based on blockchain and AI technology as described in claim 1, characterized in that: In the industrial identification binding module, the GS1-128 code contains segmented information: the first 6 digits are the recycling area code (refer to GB / T 18354-2021), the middle 8 digits are the batch serial number, and the last 4 digits are the check code, which is attached to the gold waste packaging unit by laser etching or RFID tag.

3. The trusted evidence storage and intelligent traceability system for gold recycling carbon footprint based on blockchain and AI technology as described in claim 1, characterized in that: In the edge intelligent preprocessing module, the number of trees in the isolated forest algorithm is 100-200, the subsample size is 256, and the anomaly scoring threshold is set to 0.6 (scores > 0.6 are judged as abnormal data and marked for manual review).

4. The trusted evidence storage and intelligent traceability system for gold recycling carbon footprint based on blockchain and AI technology as described in claim 1, characterized in that: The dynamic LCA carbon footprint calculation module integrates detailed parameters in the emission factor library: regional power grid emission factor corresponding to electricity consumption (e.g., 0.704 kg CO2e / kWh for East China Power Grid), IPCC Tier 2 factor corresponding to natural gas consumption (0.056 kg CO2e / MJ), and 2.68 kg CO2e / L corresponding to diesel transportation.

5. The trusted evidence storage and intelligent traceability system for gold recycling carbon footprint based on blockchain and AI technology as described in claim 1, characterized in that: In the deep reinforcement learning optimization engine, the Actor network output layer consists of three consecutive actions (temperature, oxygen flow rate, and reaction time), and the Critic network uses state-action pair value assessment to guide policy updates. The training data comes from a historical dataset of process parameters and energy consumption (sample size ≥ 100,000).

6. The trusted evidence storage and intelligent traceability system for gold recycling carbon footprint based on blockchain and AI technology as described in claim 1, characterized in that: In the process emission reduction simulation module, the virtual process flow model is constructed using Aspen Plus software and includes energy balance equations (such as smelting thermal efficiency = effective heat / total input heat) and material balance equations (such as gold recovery rate = weight after purification / initial weight of waste).

7. The trusted evidence storage and intelligent traceability system for gold recycling carbon footprint based on blockchain and AI technology as described in claim 1, characterized in that: In the blockchain evidence storage layer, consortium blockchain nodes authenticate their identities through CA certificates, and node permissions are set in a hierarchical manner: recyclers / logistics companies can only write data for their own process, refineries can read data from the entire chain, regulatory agencies have read-only and audit permissions, and third-party verification agencies can execute data verification instructions.

8. The trusted evidence storage and intelligent traceability system for gold recycling carbon footprint based on blockchain and AI technology as described in claim 1, characterized in that: In the smart contract group, the emission reduction calculation logic of the carbon asset verification contract is as follows: when the company's actual carbon emissions are less than the industry benchmark value, the emission reduction = (benchmark value - actual carbon emissions) × batch gold weight (grams); when the actual carbon emissions are greater than or equal to the benchmark value, an early warning is triggered and process optimization suggestions are pushed to the company's ESG dashboard.

9. The trusted evidence storage and intelligent traceability system for gold recycling carbon footprint based on blockchain and AI technology as described in claim 1, characterized in that: In the green finance interface of the application layer, the interest rate adjustment rules are as follows: when the carbon intensity (carbon emissions per unit of output) decreases by ≥10% month-on-month, the loan interest rate is reduced by 0.5 percentage points; when the decrease is ≥20%, it is reduced by 1.0 percentage point; when the decrease is <5%, the benchmark interest rate is maintained, and an increase triggers an ESG risk warning. The regulatory audit interface supports multiple standard switching: when the "EU CBAM mode" is selected, the output includes a report on carbon emissions in the transportation process (according to the EU ETS emission factor) and indirect emissions in the refining process (purchased electricity); when the "China carbon market mode" is selected, the output conforms to the format of the "Guidelines for Accounting Methods and Reporting of Greenhouse Gas Emissions of Enterprises (Non-ferrous Metal Smelting)". In the consumer traceability interface, the information associated with the QR code is anonymized and displayed using zero-knowledge proof (ZKP) technology, disclosing only the total carbon footprint and the comparison of emission reductions, without revealing the company's trade secrets (such as specific process parameters and supplier information).

10. A method for recording and tracing the carbon footprint of gold recycling based on blockchain and AI, characterized by: Includes the following steps: S1 Data Acquisition and Binding: The IoT sensor array collects data on waste composition and origin during the recycling process, load / mileage / fuel type during transportation, and energy consumption and process parameters during refining. A unique batch ID is generated using GS1-128 encoding, and a two-way mapping is established between the batch ID and GPS trajectory, ERP work order, and component data. Edge computing nodes are deployed, and the isolated forest algorithm is used to filter out abnormal data, achieving a data accuracy rate of ≥95% after cleaning. S2 Carbon Footprint Dynamic Accounting: Input real-time energy consumption, material consumption, and transportation parameters into the dynamic LCA model, and integrate the emission factor library of the Ministry of Ecology and Environment and the WGC industry benchmark values; Carbon footprint is calculated using the formula: Carbon Footprint = Σ(Energy Consumption in Each Process × Corresponding Emission Factor) + Σ(Transport Load × Mileage × Fuel Emission Factor), with an error rate of ≤5%. When the actual carbon emissions are less than the baseline value, CDCE certificates are generated according to the formula: Emission reduction = (Baseline value - Actual carbon emissions) × Gold weight; S3 process parameter optimization and control: Construct an Actor-Critic network policy model with the reward function being the minimization of energy consumption per unit of gold refining. The model is trained based on historical process data and outputs continuous action commands for melting temperature, oxygen flow rate, and reaction time in real time. After optimizing the parameters, compare the changes in energy consumption. If the actual energy saving is ≥12%, then update the model parameters. S4 Blockchain Trusted Evidence Storage: The original data is encrypted and stored in IPFS to generate a CID. The hash value, batch ID, timestamp, and CID are packaged and uploaded to the blockchain. The PBFT consensus mechanism is used to complete cross-chain verification, supporting data integration with the EU CBAM system and the Chinese carbon market platform. The smart contract automatically triggers the processes of evidence storage, verification, and carbon asset generation, and the execution log is uploaded to the blockchain using SHA-256 hashing. S5 Multi-Dimensional Source Tracing Query: Batch IDs are read via QR code / NFC to retrieve consortium blockchain evidence data and original IPFS files; Display the desensitized carbon footprint values, emission reduction comparisons, process optimization information, and third-party certification labels; Regulatory agencies, through authorized authentication, retrieve the entire chain of hash values ​​and raw data to generate standardized audit reports.