Marine circular economy whole-chain carbon sink accounting method and system based on ai multi-source data fusion

Abstract

This invention relates to the fields of carbon sequestration and artificial intelligence, and discloses a method and system for full-chain carbon sequestration accounting in the marine circular economy based on AI multi-source data fusion. The method includes: constructing a unified data fusion architecture based on the Transformer model, introducing marine feature labels, and dynamically adjusting feature extraction weights; constructing a full-chain coupled model, embedding a global marine microclimate correction factor library, and introducing marine-specific correction parameters such as salinity and ocean currents; establishing a regional coefficient library and supporting dynamic iterative optimization through third-party verification data; adopting a blockchain architecture to support on-chain data sharing and traceability among multiple global nodes; and combining a Kalman filter data assimilation algorithm to inversely optimize model parameters. This invention is applicable to accurate carbon sequestration accounting across all stages in various marine cities and sea areas worldwide, and can seamlessly connect to carbon trading markets in different regions globally, providing standardized, accurate, and reliable carbon sequestration quantification support for the large-scale development of the global marine circular economy.

Description

Technical Field

[0001] This invention relates to the fields of carbon sequestration and artificial intelligence, specifically to a method and system for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion. Background Technology

[0002] The full-chain carbon sequestration accounting for the marine circular economy covers multiple stages, including waste resource utilization and emission reduction, bio-based carbon sequestration, and marine ecological carbon sink enhancement. It involves multi-source heterogeneous data, including real-time IoT monitoring data, satellite remote sensing spatial data, enterprise production ledger data, and marine environmental observation data. Traditional accounting methods suffer from three major pain points:

[0003] Data integration is difficult: There are large differences in the formats of structured and unstructured data, and manual integration is inefficient. In addition, the characteristics of data in different marine environments vary significantly, and there is a lack of a globally adapted integration mechanism, making it impossible to form a full-chain carbon sequestration quantification system.

[0004] Low calculation accuracy: Traditional segmented calculation ignores the coupling relationship of carbon sinks in each link, does not take into account the microclimate differences in different ocean regions (tropical, temperate and frigid zones) around the world, and the calculation error generally exceeds ±10%. It is also difficult to adapt to the circular economy model of different ocean cities and cannot meet the stringent requirements of global carbon trading for data accuracy.

[0005] Poor traceability credibility and insufficient global adaptability: Data recording relies on manual entry, which is easy to tamper with and difficult to trace. At the same time, existing technologies are mostly limited to a single region (such as a specific port or sea area), lacking the ability to adapt to different marine ecosystems and different coastal economic models around the world, and thus unable to support the coordinated development of the global marine circular economy.

[0006] In existing technologies, most related patents focus on carbon sequestration for a single ecosystem or region, failing to cover the entire marine circular economy chain and neglecting the collaborative issues of heterogeneous fusion of multi-source data, global regional adaptation, and reliable traceability. While the urban green space carbon sequestration method disclosed in CN120996341A offers valuable insights into multi-source data fusion, it is not adapted to marine scenarios and global applications. The dynamic calibration mechanism of the land engineering carbon sequestration regulation system in CN120851389A is worth considering, but it lacks specific design considerations for regional differences in global marine environments. Summary of the Invention

[0007] To address the technical challenges of existing technologies, such as difficulties in data fusion, low accounting accuracy, poor traceability reliability, and insufficient global adaptability, this invention proposes a method and system for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion. The technical solution is as follows:

[0008] A method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion includes the following steps:

[0009] Step 1: Collect multi-source data from IoT monitoring, satellite remote sensing, production, and environment, including waste disposal volume, material production, marine plankton biomass, satellite remote sensing of marine carbon storage, energy consumption data, water temperature, salinity, and light intensity, and standardize the multi-source data.

[0010] Step 2: Using the Transformer hybrid model, feature extraction and fusion are performed on the processed data to generate carbon sink feature vectors;

[0011] Step 3: Based on the IPCC guidelines and the experience of the blue carbon pilot project, construct a full-chain accounting model of waste emission reduction, carbon sequestration of materials, and marine carbon sink enhancement. Optimize the accounting coefficients in the full-chain accounting model using an AI model, input the carbon sink feature vector, and output the total carbon sink amount.

[0012] Step 4: Record the data from the above process using blockchain technology, including multi-source data, fusion results, and accounting results, and generate a unique traceability identifier;

[0013] Step 5: Use third-party verification data and Kalman filter data assimilation algorithm for calibration, and update the localized calculation coefficient library.

[0014] Furthermore, the sampling frequency in step 1 is ≥ 1 time / minute, and the standardization process uses the Z-Score method combined with regional adaptive thresholds to remove outliers exceeding 3 times the standard deviation. The multi-source data is then unified into a globally universal format through an edge computing gateway.

[0015] Furthermore, in step 2, the Transformer hybrid model is built based on the PyTorch framework, the input layer integrates 15-dimensional structured data and 256-dimensional remote sensing image features, and the encoder is equipped with a 6-layer multi-head attention mechanism.

[0016] Furthermore, step 2 employs Sigmoid gating to dynamically allocate feature weights and utilizes a genetic algorithm to optimize the feature fusion structure.

[0017] Furthermore, the specific formula for the whole-chain accounting model in step 3 is: total carbon amount = waste emission reduction × regional emission reduction coefficient + material production × regional carbon sequestration coefficient + sea area × regional carbon sink coefficient × f(T,H,R) × G(t), where f(T,H,R) is the global marine microclimate correction factor, which includes three-dimensional parameters of temperature, humidity and salinity, and G(t) is the marine biomass growth factor.

[0018] Furthermore, the coefficients in the formula are accounting coefficients based on a regional coefficient database, which covers global marine cities and sea areas and is classified in three dimensions: climate zone, marine type, and economic model. The climate zone dimension includes tropical, subtropical, temperate, and frigid zones; the marine type dimension includes coastal industrial, port trade, ecological protection, and comprehensive types; and the economic model dimension includes waste resource utilization-led, material manufacturing-led, marine aquaculture-led, and integrated recycling-led.

[0019] The regional coefficient library supports dynamic updates via API interface. When a new region is connected, the coefficients can be quickly calibrated using a small amount of localized data.

[0020] Furthermore, step 4 supports global multi-node access, including marine city management agencies, carbon trading platforms, and third-party verification agencies in different countries / regions. Nodes share data and manage permissions through smart contracts. The entire accounting process data is stored on the blockchain to generate a unique traceability identifier. This identifier is compatible with the data interface standards of the global carbon trading market, including the Chinese carbon trading market, the EU ETS, and the California cap-and-trade program, and supports cross-regional carbon sink verification and trading.

[0021] Furthermore, in step 5, the Kalman filter data assimilation algorithm optimizes key parameters in reverse through third-party verification data, including the accounting coefficient, correction factor weight, maximum carboxylation rate, and biomass growth coefficient.

[0022] A marine circular economy full-chain carbon sink accounting system based on AI multi-source data fusion, used to implement any of the above-described methods, including:

[0023] The data acquisition module is used to collect IoT monitoring data, satellite remote sensing data, production data, and environmental data.

[0024] The standardization processing module is used to standardize multi-source data.

[0025] The multi-source data fusion module is used for feature extraction and fusion of data;

[0026] The carbon sequestration module is used to calculate the total carbon sequestration.

[0027] The digital traceability module generates a unique traceability identifier;

[0028] The model calibration module is used to optimize the parameters of the accounting system.

[0029] The system supports global multi-region parameter configuration and data interface adaptation.

[0030] Beneficial effects

[0031] This invention breaks through the limitations of single regions. Through a global regional coefficient database, microclimate correction factors, and an adaptive data fusion mechanism, it can seamlessly adapt to the circular economy scenarios of various marine cities and sea areas worldwide, with an adaptation rate of ≥95%. Through full-chain coupled modeling, global environmental correction, and data assimilation calibration, the calculation error of this method is ≤±5%, improving accuracy by 50% compared to traditional methods, meeting the data accuracy requirements of different global carbon trading markets. By using AI-automated processing to replace manual aggregation, the full-chain calculation time is shortened from 3 days to within 2 hours, improving efficiency by 80%, and data from multiple global regions can be processed in parallel. The data traceability of this invention is reliable and controllable. The blockchain consortium chain ensures that the data in the entire calculation process is tamper-proof, supports global cross-regional verification, improves the efficiency of third-party verification by 60%, and enhances the global credibility of carbon sink trading. This invention provides a standardized carbon sink accounting solution for the global marine circular economy, promotes the coordinated development of the marine circular economy in different countries / regions, and helps achieve global dual-carbon goals. Attached Figure Description

[0032] Figure 1 A flowchart of a full-chain carbon sink accounting method for the marine circular economy based on AI multi-source data fusion;

[0033] Figure 2 This is a diagram illustrating the architecture of a full-chain carbon sequestration accounting system for the marine circular economy based on AI-driven multi-source data fusion.

[0034] Figure 3 This is a diagram illustrating the classification architecture of the regional coefficient library in an embodiment of the present invention.

[0035] Figure 4 This is a flowchart illustrating the on-chain interaction process of blockchain technology in an embodiment of the present invention. Detailed Implementation

[0036] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0037] like Figure 1 As shown, the present invention provides a method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion, comprising:

[0038] Multi-source data acquisition: Deploy IoT sensors to monitor waste treatment volume and material output, obtain marine chlorophyll concentration through satellite remote sensing, and collect factory energy consumption data;

[0039] Standardization: Data is standardized using Z-Score to remove outliers;

[0040] Multi-source data fusion: A Transformer model was built based on the PyTorch framework. The input layer contained 15-dimensional structured data and remote sensing image features. The encoder was equipped with a 6-layer multi-head attention mechanism, and the decoder output a carbon sink feature vector. The model was trained using 50,000+ sets of historical accounting data and iterated for 150 rounds to ensure that the explained variance of the fused features was ≥90%.

[0041] Full-chain accounting: Referring to the IPCC 2019 guidelines, the accounting formula is set as follows: Total carbon emissions = Waste emission reduction × Emission reduction coefficient + Material production × Carbon sequestration coefficient + Sea area × Carbon sequestration coefficient. The coefficient library is optimized through AI model (emission reduction coefficient 0.8 tons CO2 / ton of waste, carbon sequestration coefficient 0.2 tons CO2 / ton of material, carbon sequestration coefficient 0.05 tons CO2 / square kilometer).

[0042] Blockchain traceability: Integrating the accounting model with the blockchain module into the digital platform to generate a unique traceability identifier and connect to the carbon trading market data interface;

[0043] Model calibration and feedback optimization: The calculation results of 100 batches were verified by a third-party verification agency. The calculation error was stable within ±4.5%, which met the transaction requirements. The calculation model was calibrated and the regional coefficient library was dynamically updated.

[0044] Based on the above method, the marine circular economy full-chain carbon sink accounting system based on AI multi-source data fusion in this embodiment, such as... Figure 2 As shown:

[0045] 1. Data Acquisition Module

[0046] This module deploys LoRa wireless IoT sensors (supporting globally universal communication protocols) to monitor waste disposal volume (accuracy ±1 ton), bio-based material production (accuracy ±0.1 ton), and marine plankton biomass at a frequency of ≥1 time / minute; it acquires remote sensing data such as global ocean carbon storage and chlorophyll concentration with a resolution of 10m through a Sentinel-2 / 3 satellite combination, providing comprehensive global coverage; it connects to the ERP systems of different companies worldwide to collect production data such as energy consumption (accuracy ±10kWh) and raw material consumption, supporting compatibility with mainstream data formats; and it integrates with a global ocean buoy network to acquire environmental data such as water temperature, salinity, light intensity, and ocean currents, forming an integrated global data acquisition network that adapts to the circular economy data collection needs of different marine cities.

[0047] 2. Standardized processing module

[0048] Data is processed using Z-Score standardization to remove outliers, such as energy consumption data that exceeds 3 times the standard deviation.

[0049] 3. Multi-source data fusion module

[0050] The multi-source data fusion module processes multi-source data including:

[0051] Structured data (waste disposal volume, energy consumption, etc.): Multi-scale one-dimensional convolution is used to extract temporal features, while embedding marine feature labels (latitude, salinity, temperature level) to enhance the distinguishability of data from different regions around the world.

[0052] Remote sensing image data: Spatial distribution characteristics of marine carbon storage are extracted using a CNN network, and key spectral bands in different marine areas are identified by combining a band attention mechanism to adapt to the differences in remote sensing data of tropical, temperate, and frigid marine areas.

[0053] Environmental data: The spatial autocorrelation index (Moran's I) was used to analyze the spatial clustering characteristics of environmental factors, and the global ocean zoning standard was introduced to ensure the consistency of environmental data characteristics in different regions.

[0054] The multi-source data fusion module dynamically assigns weights to each modal feature using the Sigmoid function, with the weight coefficients adaptively adjusted according to the sea area type (e.g., increasing the weight of plankton biomass data in high-salinity sea areas and increasing the weight of production data in industrial marine cities). A 6-layer multi-head attention Transformer model is built based on the PyTorch framework. The input layer integrates 15-dimensional structured data and 256-dimensional remote sensing image features. The model is trained using sample data from different regions around the world to ensure its generalization ability in various marine scenarios, generating a 256-dimensional carbon sink feature vector that explains ≥90% of the variance.

[0055] 4. Carbon Sequestration Calculation Module

[0056] The carbon sink accounting module is based on the IPCC 2019 guidelines and the experience of major global blue carbon pilot projects (Xiangshan Port in China, Great Barrier Reef in Australia, Bergen Port in Norway, etc.), and constructs an accounting formula with a global correction factor: ,in, Total carbon content (tons of CO2); Waste emission reduction (tons); This is the regional emission reduction coefficient (retrieved from the global regional coefficient database, such as 0.8 for coastal China, 0.75 for coastal Northern Europe, and 0.82 for coastal Southeast Asia). Material output (tons); The coefficient represents the regional carbon sequestration coefficient (dynamically updated in the global coefficient database); S represents the sea area (square kilometers). This is the regional sinking coefficient (adapted to different marine ecosystem types). The global marine microclimate correction factor, considering the three-dimensional parameters of temperature (T), humidity (H), and salinity (R), is formulated as follows: ,in , , These are the baseline environmental parameters for the corresponding sea area; It is a marine biomass growth factor, obtained by fitting time-series biomass data from different sea areas around the world, and adapted to the growth characteristics of different marine ecosystems.

[0057] The carbon sequestration accounting module has built a regional coefficient database covering over 200 typical marine cities and sea areas worldwide, such as... Figure 3 As shown, the classification is based on a three-dimensional framework of "climate zone - ocean type - economic model":

[0058] Primary classification (climate zones): Classified into four categories based on global marine climate characteristics to ensure coverage of marine environmental differences across different climate zones:

[0059] Tropical: Latitude 0°-23.5°, typical examples are Singapore and Kuala Lumpur, suitable for high salinity and high temperature marine environments;

[0060] Subtropical: latitude 23.5°-30°, typical examples are Shenzhen and Dubai, suitable for mild, saline and humid environments with seasonal temperature fluctuations;

[0061] Temperate zone: latitude 30°-60°, typical examples are Xiangshan Port in China and Hamburg in Germany, suitable for distinct seasons and medium salinity environment;

[0062] Arctic: latitudes above 60°, typical examples are Bergen Port in Norway and Anchorage in the United States, suitable for low temperature, high salinity, long day / polar night environments.

[0063] Secondary classification (marine type): Based on the functional positioning of coastal cities, they are divided into 4 categories to match different basic scenarios of marine circular economy:

[0064] Coastal industrial type: The core feature is the concentrated emission of industrial waste, and the coefficient is optimized with a focus on emission reduction coefficients;

[0065] Port trade type: The core characteristics are large demand for logistics packaging and frequent turnover. The coefficient matching focuses on balancing carbon sequestration and emission reduction coefficients.

[0066] Ecological protection type: The core characteristics are marine ecological sensitivity and carbon sink dependence on natural carbon sinking, with the coefficient adaptation focusing on optimizing the carbon sinking coefficient;

[0067] Comprehensive type: The core feature is the coexistence of multiple cyclic modes, and the coefficient adaptation focuses on achieving dynamic balance of all coefficients.

[0068] Three-level classification (economic model): Divided into 4 categories according to the core driving direction of the marine circular economy, with clear priorities for coefficient optimization:

[0069] Waste resource utilization is the dominant factor: the priority of coefficient matching is emission reduction coefficient > carbon sequestration coefficient > carbon sink coefficient;

[0070] Materials manufacturing is the primary factor: the priority for coefficient matching is carbon sequestration coefficient > emission reduction coefficient > carbon sequestration coefficient;

[0071] Marine aquaculture is the dominant factor: the priority of coefficient matching is: carbon sequestration coefficient > emission reduction coefficient > carbon fixation coefficient;

[0072] Comprehensive cycle-driven: coefficient adaptation priority is dynamic balance optimization of all coefficients.

[0073] The regional coefficient library supports dynamic updates via API. When a new region is added, the coefficients can be quickly calibrated using a small amount of localized data without needing to reconstruct the model.

[0074] 5. Digital traceability module

[0075] Adopting the Hyperledger Fabric consortium blockchain architecture, it supports access from multiple nodes globally (marine city management agencies, carbon trading platforms, and third-party verification agencies in different countries / regions). Nodes share data and manage permissions through smart contracts. It stores all data from the entire accounting process (raw data, fusion results, accounting results, regional coefficients, and correction factors) on the blockchain, generating a unique traceability identifier. This identifier is compatible with the data interface standards of major global carbon trading markets (China's carbon trading market, the EU ETS, California's cap-and-trade program, etc.), supporting cross-regional carbon sink verification and trading.

[0076] On-chain interaction process as follows Figure 4 As shown, the timeline runs from left to right, and the core steps and timing are as follows:

[0077] 4-1 Enterprise side (T0 time): Upload the entire accounting process data (raw data / fusion results / total accounting amount) to the consortium blockchain;

[0078] 4-2 Blockchain Node (Time T1, T1-T0=5s): Generate a unique traceability identifier (e.g., "TX-20250101-XS001") and verify it through block consensus;

[0079] 4-3 Third-party verification agency (time T2, T2-T1=2h): retrieve on-chain data for verification, upload verification report and put it on the chain;

[0080] 4-4 Carbon Trading Platform (T3 time, T3-T2=1h): Verify traceability identifiers and verification reports, and connect to the Zhejiang Province carbon trading interface;

[0081] 4-5 Feedback loop (time T4, T4-T3=12h): The transaction result is sent back to the enterprise, and the data is synchronized to the model calibration module.

[0082] 6. Model Calibration Module

[0083] The model calibration module employs a Kalman filter data assimilation algorithm, integrating third-party verification data from various marine cities worldwide to inversely optimize key model parameters (regional coefficients, correction factor weights, and biomass growth coefficients), as detailed below:

[0084] We collected accounting results from different regions around the world and third-party measured data to calculate the discrepancies. ;

[0085] Based on the deviation values, state equations and observation equations for multiple regions around the world are constructed, and model parameters are optimized to ensure that parameter adjustments are in line with the ecological rationality of different regions.

[0086] The global regional coefficient database is updated quarterly, and a full model iteration optimization is performed annually to ensure that the calculation error of the model under different global environments remains stable within ±5% in the long term.

[0087] It provides a localized calibration interface, allowing newly connected marine cities to quickly adapt to regional characteristics by inputting 50+ sets of localized data, with a calibration cycle of ≤7 days.

[0088] Example 1: Application of the China Xiangshan Port Marine Circular Economy Demonstration Base (Temperate Coastal Industrial Type)

[0089] Data Acquisition and Standardization: Twenty LoRa IoT sensors were deployed, covering the waste treatment station, bio-based material production line, and marine aquaculture area, collecting data at a frequency of once per minute; data from a 500-square-kilometer sea area was acquired via Sentinel-2 satellite; data on annual energy consumption of 1.2 million kWh and raw material consumption of 200,000 tons were collected by integrating with the factory's ERP system; and water temperature (annual average 18-22℃) and salinity (28-32‰) were obtained using three marine buoys. Z-Score standardization was employed, and three abnormal data points from typhoon days were removed, automatically matching the regional coefficient for "temperate-coastal industrial-waste resource-dominated" areas.

[0090] Multi-source data fusion and accounting: The model automatically calls the feature extraction weights of the temperate region to generate a 256-dimensional carbon sink feature vector; it calls the regional coefficients (emission reduction coefficient 0.8, carbon sequestration coefficient 0.2, carbon sink enhancement coefficient 0.05) and embeds the local microclimate correction factor (f(T,H,R)=0.98) and biomass growth factor (G(t)=1.06); the monthly waste treatment volume is 5000 tons, the bio-based material production is 1000 tons, and the covered sea area is 100 square kilometers. The total carbon sink is calculated as 5000×0.8+1000×0.2+100×0.05×0.98×1.06=4205.19 tons of CO2.

[0091] Traceability and Calibration: Data is uploaded to the consortium blockchain, generating a traceability identifier TX-20250101-XS001. The third-party verification result is 4180 tons of CO2 with an error of 0.59%. The data is synchronized to the global regional coefficient database to optimize the coefficient accuracy of temperate coastal industrial areas.

[0092] Example 2: Marine Circular Economy Application in Marina Bay, Singapore (Tropical Port Trade Type)

[0093] Data Acquisition and Standardization: Access IoT monitoring data from the Port of Singapore (waste disposal volume, container cold chain energy consumption), Sentinel-3 satellite marine data, and local marine buoy environmental data (water temperature 27-30℃, salinity 33-35‰); the model automatically identifies the "tropical-port trade-industrialized-integrated circular economy" type and calls the corresponding regional coefficients (emission reduction coefficient 0.82, carbon sequestration coefficient 0.18, carbon sequestration coefficient 0.04);

[0094] Multi-source data fusion and accounting: The feature weights are adjusted through an adaptive data fusion mechanism to generate a carbon sink feature vector; a tropical microclimate correction factor (f(T,H,R)=1.05) is embedded. The monthly waste treatment volume is 3,000 tons, the output of bio-based packaging materials is 800 tons, and the sea area covered is 50 square kilometers. The total carbon sink is calculated as 3,000×0.82+800×0.18+50×0.04×1.05=2642.1 tons of CO2.

[0095] Global trading integration: The traceability label is compatible with the EU ETS interface. After the calculation results are verified by a third party, they can be directly used for cross-regional carbon sink trading. The verification error is 4.2%, which meets the trading requirements.

[0096] Example 3: Marine Circular Economy Application in Bergen Port, Norway (Cold-Region Ecological Protection Type)

[0097] Data collection and standardization: Access to data on marine aquaculture waste treatment in Bergen Port, production data of bio-based fishing gear materials, Sentinel-2 satellite data of the Arctic Ocean, and local buoy data (water temperature 4-12℃, salinity 34-36‰); the model automatically matches the regional coefficients for "cold zone - ecological protection - marine aquaculture-dominated" (emission reduction coefficient 0.75, carbon sequestration coefficient 0.22, carbon sink enhancement coefficient 0.06);

[0098] Multi-source data fusion and accounting: embedding a cold-climate correction factor (f(T,H,R)=0.92), with aquaculture waste treatment volume of 2000 tons, bio-based fishing gear production of 500 tons, and a sea area covered of 150 square kilometers in the current month, the total carbon emissions are calculated as 2000×0.75+500×0.22+150×0.06×0.92=1627.08 tons of CO2;

[0099] Localized calibration: By inputting 60 sets of localized data from the Port of Bergen, the model quickly calibrated the regional coefficients, and the calculation error was reduced to 3.8% after calibration, meeting the requirements of Norwegian local carbon trading.

[0100] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention 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 or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An AI-based multi-source data fusion-based marine circular economy full-chain carbon sink accounting method, characterized in that, Includes the following steps: Step 1: Collect multi-source data from IoT monitoring, satellite remote sensing, production, and environment, including waste disposal volume, material production, marine plankton biomass, satellite remote sensing of marine carbon storage, energy consumption data, water temperature, salinity, and light intensity, and standardize the multi-source data. Step 2: Use the Transformer hybrid model to extract and fuse features from the processed data to generate a carbon sink feature vector; Step 3: Based on the IPCC guidelines and the experience of the blue carbon pilot project, construct a full-chain accounting model of waste emission reduction, carbon sequestration of materials, and marine carbon sink enhancement. Optimize the accounting coefficients in the full-chain accounting model using an AI model, input the carbon sink feature vector, and output the total carbon sink amount. Step 4: Record the data from the above process using blockchain technology, including multi-source data, fusion results, and accounting results, and generate a unique traceability identifier; Step 5: Use third-party verification data and Kalman filter data assimilation algorithm for calibration, and update the localized calculation coefficient library.

2. The method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion as described in claim 1, characterized in that: The sampling frequency in step 1 is ≥ 1 time / minute. The standardization process uses the Z-Score method combined with regional adaptive thresholds to remove outliers exceeding 3 times the standard deviation. The multi-source data is then unified into a globally universal format through an edge computing gateway.

3. The method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion as described in claim 1, characterized in that: In step 2, the Transformer hybrid model is built based on the PyTorch framework. The input layer integrates 15-dimensional structured data and 256-dimensional remote sensing image features, and the encoder is equipped with a 6-layer multi-head attention mechanism.

4. The method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion as described in claim 1, characterized in that: Step 2 uses Sigmoid gating to dynamically allocate feature weights and utilizes a genetic algorithm to optimize the feature fusion structure.

5. The method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion as described in claim 1, characterized in that: The specific formula for the full-chain accounting model in step 3 is: Total carbon amount = Waste emission reduction × Regional emission reduction coefficient + Material production × Regional carbon sequestration coefficient + Sea area × Regional carbon sink coefficient × f(T,H,R) × G(t), where f(T,H,R) is the global marine microclimate correction factor, which includes three-dimensional parameters of temperature, humidity and salinity, and G(t) is the marine biomass growth factor.

6. The method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion as described in claim 5, characterized in that: The coefficients in the formula are calculated based on a regional coefficient database, which covers global marine cities and sea areas and is classified in three dimensions: climate zone, marine type, and economic model. The climate zone dimension includes tropical, subtropical, temperate, and frigid zones; the marine type dimension includes coastal industrial, port trade, ecological protection, and comprehensive; and the economic model dimension includes waste resource utilization, material manufacturing, marine aquaculture, and integrated recycling. The regional coefficient library supports dynamic updates via API interface. When a new region is connected, the coefficients can be quickly calibrated using a small amount of localized data.

7. The method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion as described in claim 1, characterized in that: Step 4 supports global multi-node access, including marine city management agencies, carbon trading platforms, and third-party verification agencies in different countries / regions. Nodes share data and manage permissions through smart contracts. The entire accounting process data is stored on the blockchain to generate a unique traceability identifier. This identifier is compatible with the data interface standards of the global carbon trading market, including the Chinese carbon trading market, the EU ETS, and the California cap-and-trade program, and supports cross-regional carbon sink verification and trading.

8. The method for carbon sequestration across the entire marine circular economy chain based on AI multi-source data fusion as described in claim 1, characterized in that: In step 5, the Kalman filter data assimilation algorithm optimizes key parameters in reverse by using third-party verification data, including the calculation coefficient, correction factor weight, maximum carboxylation rate, and biomass growth coefficient.

9. A marine circular economy full-chain carbon sink accounting system based on AI multi-source data fusion, characterized in that, For implementing the method as described in any one of claims 1 to 8, comprising: The data acquisition module is used to collect IoT monitoring data, satellite remote sensing data, production data, and environmental data. The standardization processing module is used to standardize multi-source data. The multi-source data fusion module is used for feature extraction and fusion of data; The carbon sequestration module is used to calculate the total carbon sequestration. The digital traceability module generates a unique traceability identifier; The model calibration module is used to optimize the parameters of the accounting system. The system supports global multi-region parameter configuration and data interface adaptation.

Images