Factory carbon emission monitoring device and monitoring method thereof
By combining multimodal sensing units and digital twin models with blockchain evidence storage, the accuracy and reliability issues of factory carbon emission monitoring have been solved, enabling dynamic accounting and visualized management of carbon emissions throughout the entire process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-15
- Publication Date
- 2026-04-07
AI Technical Summary
Existing carbon emission monitoring technologies for factories are insufficient for accurate, dynamic, and full-process accounting. They cannot trace the carbon emissions of specific production processes and equipment, cannot identify high-emission links, and the data is susceptible to dilution and mixing. Traditional databases are at risk of being tampered with and cannot be used as reliable evidence.
Multimodal sensing units are used to collect carbon emission data in real time. Data cleaning, spatiotemporal alignment and full-process simulation are performed through edge computing and digital twin models. Combined with blockchain notarization, accurate accounting and reliable notarization of carbon emissions are achieved.
It enables comprehensive and seamless monitoring of factory carbon emissions, improves accounting accuracy and reliability, provides second-level anomaly warnings and carbon emission reduction simulations, and reduces audit and trust costs.
Smart Images
Figure CN121810316A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of factory carbon emission monitoring technology, specifically to a factory carbon emission monitoring device and its monitoring method. Background Technology
[0002] With the ongoing global efforts to address climate change, achieving carbon peaking and carbon neutrality has become a core strategy for my country's economic and social development. As a major source of energy consumption and greenhouse gas emissions, the accurate monitoring, accounting, and management of carbon emissions in the industrial sector are crucial for achieving macro-level emission reduction targets. Accurate and reliable carbon emission data are the foundation for carbon trading, carbon audits, performance evaluation of emission reductions, the development of scientific carbon quotas, and the guidance of low-carbon technology investment. However, current carbon emission monitoring and accounting systems at the factory level still face a series of severe challenges in terms of technology and practical application, making it difficult to meet the growing demands for precision, real-time data, and verifiability.
[0003] Currently, carbon emission accounting methods in factories mainly rely on the emission factor method and the material balance method. The emission factor method calculates emissions by multiplying fuel consumption and activity level data by a fixed empirical emission factor. This method is simple to operate, but its accuracy is limited, failing to reflect the impact of actual operating conditions such as equipment efficiency and changes in operating conditions, leading to significant discrepancies between the calculated results and actual emissions. While the material balance method is theoretically more accurate, in complex industrial production, the diverse transformation pathways of carbon in the process flow, incomplete measurement points, and difficulties in data closure often hinder its practical application, especially in process industries involving complex chemical reactions or large quantities of intermediate products.
[0004] At the monitoring technology level, existing continuous emission monitoring systems are mainly installed at the factory's main exhaust outlet or main chimney. Their monitoring results (such as CO2 concentration and flue gas flow rate) can reflect the instantaneous status of direct emissions at the end. However, this method has obvious defects: (1) It can only give the final total emission amount and cannot trace the emission responsibility to specific production processes, equipment or even products, making it difficult to identify high emission links; (2) It is powerless against fugitive emissions, escape emissions (such as pipeline leaks) and indirect emissions (such as carbon emissions corresponding to purchased electricity and steam); (3) Flue gas concentration is easily affected by dilution, mixing and other factors, and the representativeness of the measurement is questionable. On the other hand, the distributed energy metering and equipment operating parameters (DCS / PLC data) widely existing in factories contain carbon flow information, but these data are usually scattered in independent monitoring systems with different formats, different sampling frequencies, asynchronous time, and lack of unified spatial correlation, making it difficult to directly apply to refined carbon accounting.
[0005] Furthermore, existing carbon emission data is mostly collected, aggregated, and calculated manually after the fact, a process lacking transparency and susceptible to human interference. In scenarios involving significant economic interests, such as carbon trading and carbon tariffs, it is difficult to serve as reliable evidence at the judicial or financial level. Traditional centralized database storage methods also face the risk of data tampering. No solutions have yet been proposed to address these technical issues. Summary of the Invention
[0006] To address the problems in related technologies, this invention proposes a factory carbon emission monitoring device and its monitoring method to overcome the aforementioned technical problems existing in the current related technologies. The purpose of this invention is to achieve accurate, dynamic, and full-process accounting of carbon emissions, break through the limitations of traditional methods, solve the existing problems of industrial data fusion, ensure the reliability and timeliness of the accounting model, build a reliable and auditable data chain, and see the impact prediction on the entire process of carbon emissions in real time. At the same time, it can also achieve second-level anomaly warning and preliminary analysis to ensure the accuracy of the final accounting results.
[0007] To achieve the above objectives, the present invention provides the following technical solution: a factory carbon emission monitoring device, comprising: The edge sensing module includes multimodal sensing units installed at key carbon source nodes within the factory. These multimodal sensing units are used to collect real-time process data directly related to carbon emissions. The key carbon source nodes include main energy input ports, key combustion / reaction equipment, main chimneys / exhaust outlets, and large energy-consuming equipment. The data fusion and edge computing module includes an edge computing gateway communicatively connected to the multimodal sensing unit. The edge computing gateway incorporates a data cleaning module, a multi-source spatiotemporal alignment module, and a preliminary emission factor library. The data cleaning module performs outlier removal and smoothing on the received raw process data. The multi-source spatiotemporal alignment module timestamps and maps the spatial locations of sensor data from different physical locations and sampling frequencies, forming a data sequence within a unified spatiotemporal framework. The preliminary emission factor library stores basic emission factors based on equipment model and fuel type. The digital twin and precise accounting module includes a carbon monitoring server, which is communicatively connected to the edge computing gateway. The carbon monitoring server internally constructs a full-process carbon flow digital twin model synchronized with the actual factory's technological flow. This full-process carbon flow digital twin model receives data sequences processed by the data fusion and edge computing module, drives the model's operation, and dynamically simulates and quantifies the flow, transformation, and final emission of carbon elements in various factory processes, with material balance and energy balance as its core, outputting a carbon emission intensity map. The visualization and interaction module includes a management terminal connected to the carbon monitoring server, which is used to display the carbon emission intensity map, real-time emission data of each node, early warning information, and simulation results of carbon emission reduction methods in real time.
[0008] Preferably, the multimodal sensing unit includes: High-precision fuel flow meters and online component analyzers are used to measure fuel consumption and carbon content in real time. Multi-parameter monitoring instruments at the discharge outlets of key equipment are used to monitor flue gas flow rate, CO2, CO, CH4 concentration, temperature, and pressure parameters in real time. Distributed power consumption monitoring terminals are used to collect real-time power consumption data for individual items and equipment. The visual recognition module is used to assist in calculating the consumption of solid fuels through image recognition and volume scanning.
[0009] Preferably, the multi-source spatiotemporal alignment module uses the BeiDou / GPS synchronous clock as a reference to stamp all sensor data with microsecond-level timestamps; establishes a factory coordinate system and binds the device IDs of several multimodal sensing units to their physical coordinates; for data with different sampling frequencies, an interpolation algorithm based on process event triggering is used to align all data streams to a unified virtual time axis, wherein the process events include equipment start / stop signals, batch start signals, and batch end signals.
[0010] Preferably, the end-to-end carbon flow digital twin model includes: The static process topology network is constructed based on the actual process flow diagram of the factory, defining all equipment nodes, pipeline connections and carbon flow paths from raw material input to product / waste output. The dynamic mechanism calculation unit is embedded in each device node of the static process topology network. The dynamic mechanism calculation unit calculates the carbon output, carbon conversion rate and carbon emissions of the node based on the carbon flow and energy flow of the input material and the process parameters of the node, and on the basis of chemical reaction equation, combustion equation and heat balance equation. The real-time data-driven engine is used to inject the real-time measurement values in the data sequence as boundary conditions or correction parameters into the corresponding dynamic mechanism calculation unit, driving the entire model to perform synchronous simulation calculations at the second / minute level.
[0011] Preferably, the dynamic mechanism calculation unit adopts a hybrid calculation mode: For combustion and reaction processes with known stoichiometric ratios, deterministic mechanistic models are used for calculations. For biochemical and complex reaction processes, data-driven models are used for fitting calculations. The carbon monitoring server has a built-in model verification module. The model verification module periodically compares the virtual measurement values calculated by the model with the actual sensor measurement values. If the deviation exceeds the threshold, the model parameter calibration is automatically triggered.
[0012] Preferably, it also includes a blockchain evidence storage unit connected to the carbon monitoring server; the blockchain evidence storage unit is used to generate a data block from the key carbon emission accounting result data package output by the full-process carbon flow digital twin model, the key carbon emission accounting result data package including the carbon emission amount of each process, the total emission amount, the corresponding supporting process data hash value and timestamp, and upload it to the blockchain network for distributed evidence storage, forming an immutable carbon emission audit clue.
[0013] To achieve the above objectives, the present invention also provides the following technical solution: A monitoring method for a factory carbon emission monitoring device includes the following steps: Step S1: Real-time acquisition of full-process data of the factory through multimodal sensing units set at key carbon source nodes in the factory. In step S2, the edge computing gateway receives the process data, performs data cleaning and multi-source spatiotemporal alignment processing in sequence to form a unified standardized data sequence, and uses a preliminary emission factor library to perform a fast and rough emission estimation. Step S3: The carbon monitoring server receives the standardized data sequence and drives the full-process carbon flow digital twin model to perform real-time simulation. Step S4: The full-process carbon flow digital twin model is based on material and energy balance, combined with real-time data, to dynamically calculate the inflow, retention and outflow of carbon elements in each process link, accurately calculate direct and indirect emissions, and generate process-level carbon emission intensity map. Step S5: Visualize the carbon emission intensity map, real-time calculation results, and early warning information on the management terminal; Step S6: Package the key accounting results and evidence chain data, and upload them to the blockchain for storage.
[0014] Preferably, in step S4, the precise calculation specifically includes: For direct emissions, the full-process carbon flow digital twin model takes high-precision fuel measurement and multi-parameter monitoring at the emission outlet as direct inputs, and calibrates combustion efficiency through model inversion and data assimilation to achieve minute-level dynamic calculation of emissions; For indirect emissions, the model maps distributed power consumption data to specific process equipment, and combines real-time grid emission factors and the carbon intensity of in-plant self-generated power to allocate and trace the power consumption to the final product.
[0015] Preferably, in step S6, the key accounting results and evidence chain data include: accounting time interval, total carbon emissions of the factory, carbon emissions of each process, timestamp sequence hash of the real-time data used, model version number and calibration parameter snapshot, and digital signature of the accounting results.
[0016] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention is a factory carbon emission monitoring device and its monitoring method. By setting up a multimodal sensing unit, it directly and in real time measures core process data such as fuel carbon content and key emission parameters. Based on the full-process carbon flow digital twin model, it performs accurate dynamic simulation of material and energy balance, thereby improving the accuracy of the calculation. By simulating the flow and transformation of carbon elements in various processes of the factory through the digital twin model, it can generate a process-level carbon emission intensity map, clearly showing the carbon footprint of each piece of equipment and each production line, enabling managers to accurately locate high emission and efficiency bottlenecks. It not only accurately calculates direct emissions such as fuel combustion, but also scientifically and reasonably allocates the corresponding indirect emissions such as purchased electricity and steam to specific products and processes through distributed power consumption monitoring and model mapping, thereby achieving full-caliber and comprehensive monitoring of factory carbon emissions. (2) This invention is a factory carbon emission monitoring device and its monitoring method. By setting up a multi-source spatiotemporal alignment module, based on the Beidou / GPS high-precision clock and the factory coordinate system, and using an intelligent alignment algorithm based on process event triggering, the sensor data of different frequencies and locations are unified into a consistent spatiotemporal framework, providing high-quality and highly consistent input for subsequent model calculations, and ensuring the authenticity and reliability of digital twin simulation; by setting up a model verification module, the model simulation results can be automatically compared with the actual measured values on a regular basis. Once the deviation exceeds the threshold, an automatic warning or parameter calibration process can be triggered. (3) This invention is a factory carbon emission monitoring device and its monitoring method. By setting up a blockchain storage unit, key accounting results, supporting data hashes, timestamps and other information are packaged and uploaded to the chain to form an immutable data fingerprint. This provides each carbon emission report with legally credible electronic evidence, effectively preventing data fraud. This allows the monitoring data to be directly used in scenarios with extremely high requirements for data authenticity, such as carbon trading, green finance, and carbon tariff accounting, which greatly reduces audit costs and trust costs. By setting up visualization and interactive modules, not only is the data displayed, but carbon emission reduction simulation functions are also provided. Managers can adjust process parameters, change raw materials or energy efficiency equipment in the full-process carbon flow digital twin model, and see the impact prediction on the full-process carbon emissions in real time, thereby improving the factory's low-carbon management capabilities. Attached Figure Description
[0017] Figure 1 This is a schematic diagram of the overall framework structure of the present invention. Detailed Implementation
[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
[0019] Example Please see Figure 1 This invention proposes a technical solution for a factory carbon emission monitoring device and its monitoring method: A factory carbon emission monitoring device includes: The edge sensing module includes multimodal sensing units installed at key carbon source nodes within the plant. These units collect real-time process data directly related to carbon emissions. Key carbon source nodes include main energy input points, key combustion / reaction equipment, main chimneys / exhaust outlets, and large energy-consuming equipment. Specifically, the main energy input points serve as the absolute benchmark for plant-wide carbon balance. Monitoring key combustion / reaction equipment directly yields crucial process parameters such as conversion efficiency and byproducts. The main chimney / exhaust outlet serves as an important verification point, not the sole data source. Individual metering of large energy-consuming equipment is fundamental to accurate allocation of indirect emissions. The data fusion and edge computing module includes an edge computing gateway communicatively connected to the multimodal sensing unit. The edge computing gateway incorporates a data cleaning module, a multi-source spatiotemporal alignment module, and a preliminary emission factor library. The data cleaning module performs outlier removal and smoothing on the received raw process data. The multi-source spatiotemporal alignment module timestamps and maps the spatial locations of sensor data from different physical locations and sampling frequencies, forming a data sequence within a unified spatiotemporal framework. The preliminary emission factor library stores basic emission factors based on equipment model and fuel type. Specifically, the data cleaning module uses a composite algorithm based on statistics (such as the 3σ principle), rules (such as range exceeding limits and sudden changes in rate of change), and models (prediction bias) to automatically identify and process abnormal data caused by sensor failures, signal interference, and transmission packet loss, ensuring the stability of downstream calculations. The multi-source spatiotemporal alignment module uses nanosecond-level timing signals from BeiDou / GPS as its source, assigning a unified absolute timestamp to all data and establishing a digital twin coordinate system for the factory. This lays the foundation for accurately locating the data source in the model. For data with different sampling frequencies (e.g., flow meters once per second, component analyzers once every 10 seconds), simple linear interpolation can produce significant errors when operating conditions change abruptly. This invention employs an interpolation algorithm based on process event triggering. When a signal such as "boiler ignition" is received, the system knows that the fuel composition may undergo a step change, ensuring the physical continuity and accuracy of the data sequence. The preliminary emission factor library, using aligned data and basic emission factors at the edge, can quickly calculate the approximate emissions of each node. When the rapid estimate of a node spikes instantaneously, an alarm can be immediately issued to the site. By performing rough calculations upfront, the server focuses on high-fidelity simulation. The digital twin and precise accounting module includes a carbon monitoring server, which is communicatively connected to the edge computing gateway. The carbon monitoring server internally constructs a full-process carbon flow digital twin model synchronized with the actual factory's technological flow. This full-process carbon flow digital twin model receives data sequences processed by the data fusion and edge computing module, drives the model's operation, and dynamically simulates and quantifies the flow, transformation, and final emission of carbon elements in various factory processes, with material balance and energy balance as its core, outputting a carbon emission intensity map. The visualization and interaction module includes a management terminal connected to the carbon monitoring server, which is used to display the carbon emission intensity map, real-time emission data of each node, early warning information, and simulation results of carbon emission reduction methods in real time. Specifically, the carbon emission intensity map presented by the management terminal is a heat map or Sankey diagram, which intuitively shows where carbon flows in, where it concentrates, and where it is emitted. Managers can view real-time carbon intensity, historical trends, and comparisons with industry benchmarks.
[0020] Furthermore, the multimodal sensing unit includes: High-precision fuel flow meters and online component analyzers are used to measure fuel consumption and carbon content in real time. Multi-parameter monitoring instruments at the discharge outlets of key equipment are used to monitor flue gas flow rate, CO2, CO, CH4 concentration, temperature, and pressure parameters in real time. Distributed power consumption monitoring terminals are used to collect real-time power consumption data for individual items and equipment. The visual recognition module is used to assist in calculating the consumption of solid fuels through image recognition and volume scanning.
[0021] In this embodiment, the high-precision fuel flow meter and online component analyzer ensure the accuracy of the total amount. Online component analysis (such as near-infrared spectroscopy and laser gas analysis) can analyze the calorific value, carbon-hydrogen ratio, moisture content, etc. of the fuel in real time and dynamically determine its carbon content factor, completely eliminating the dependence on fixed empirical values. The multi-parameter monitor at the exhaust outlet of key equipment not only monitors CO2, but also simultaneously monitors CO (incomplete combustion index), CH4 (escape emission), etc., providing multi-dimensional data for assessing combustion / reaction efficiency and detecting operational anomalies; Distributed power consumption monitoring terminals acquire real-time power consumption data of workshops, production lines, and even individual equipment through smart meters or power management system (EMS) interfaces, providing a data foundation for drawing power carbon flow maps; The visual recognition module targets solid materials such as coal piles and biomass fuels that are difficult to measure continuously. It uses high-definition cameras and 3D scanning, combined with image recognition algorithms, to estimate changes in material volume periodically or continuously. Combined with density models, it helps to correct and verify solid fuel consumption, thus making up for the shortcomings of traditional measurement methods.
[0022] Furthermore, the multi-source spatiotemporal alignment module uses the BeiDou / GPS synchronous clock as a reference to add microsecond-level timestamps to all sensor data; it establishes a factory coordinate system and binds the device IDs of several multimodal sensing units to their physical coordinates; for data with different sampling frequencies, it uses an interpolation algorithm based on process event triggering to align all data streams to a unified virtual time axis, where process events include equipment start / stop signals, batch start signals, and batch end signals.
[0023] Furthermore, the full-process carbon flow digital twin model includes: The static process topology network is constructed based on the actual process flow diagram of the factory, defining all equipment nodes, pipeline connections and carbon flow paths from raw material input to product / waste output. The dynamic mechanism calculation unit is embedded in each device node of the static process topology network. The dynamic mechanism calculation unit calculates the carbon output, carbon conversion rate and carbon emissions of the node based on the carbon flow and energy flow of the input material and the process parameters of the node, and on the basis of chemical reaction equation, combustion equation and heat balance equation. The real-time data-driven engine is used to inject the real-time measurement values in the data sequence as boundary conditions or correction parameters into the corresponding dynamic mechanism calculation unit, driving the entire model to perform synchronous simulation calculations at the second / minute level.
[0024] In this embodiment, the static process topology network is based on factory knowledge such as P&ID diagrams and material balance sheets. It constructs a directed graph with nodes (equipment) and edges (pipelines / materials), defining all possible carbon flow paths in the virtual factory. For boilers and kilns, the dynamic mechanism calculation unit incorporates combustion reaction equations and thermal efficiency models. Inputting fuel carbon flow and air volume, it calculates the theoretical flue gas volume and CO2 generation, and combines this with measured exhaust data (O2 concentration, temperature) to infer the actual combustion efficiency, thereby calibrating emissions. For reactors and synthesis towers, the dynamic mechanism calculation unit incorporates a chemometrics model. Inputting raw material carbon flow, it calculates the theoretical flue gas volume and CO2 generation based on process parameters such as conversion rate and selectivity. Parameters (obtainable in real-time from DCS) are used to calculate the carbon distribution in products, by-products, and exhaust gases. Data-driven models (such as LSTM neural networks and gradient boosting trees) are employed to predict the carbon conversion and emissions of the unit using historical data accumulated over a long period (feed composition, temperature, pH, output gas, etc.), improving the model's adaptability and practicality to complex industrial scenarios. The real-time data-driven engine dynamically injects the processed real-time data sequence as boundary conditions (such as inlet flow rate), setpoints (such as reaction temperature), or correction parameters (such as efficiency coefficient) into each computing unit, driving the entire virtual factory to run at a speed (seconds / minutes) almost synchronous with the real factory, achieving real-time simulation.
[0025] Furthermore, the dynamic mechanism calculation unit adopts a hybrid calculation mode: For combustion and reaction processes with known stoichiometric ratios, deterministic mechanistic models are used for calculations. For biochemical and complex reaction processes, data-driven models are used for fitting calculations. The carbon monitoring server has a built-in model verification module. The model verification module periodically compares the virtual measurement values calculated by the model with the actual sensor measurement values. If the deviation exceeds the threshold, the model parameter calibration is automatically triggered.
[0026] In this embodiment, the model verification module periodically compares the virtual measurement values output by the model (such as the simulated CO2 concentration at a certain outlet) with the actual sensor readings. The persistent deviation indicates that the model may be mismatched (such as decreased catalyst activity or equipment scaling). Based on this, the system can automatically trigger the calibration process or issue a diagnostic report to the maintenance personnel, prompting them to pay attention to specific equipment.
[0027] Furthermore, it also includes a blockchain evidence storage unit connected to the carbon monitoring server; the blockchain evidence storage unit is used to generate data blocks from the key carbon emission accounting result data package output by the full-process carbon flow digital twin model, the key carbon emission accounting result data package including the carbon emission amount of each process, the total emission amount, the corresponding supporting process data hash value and timestamp, and upload it to the blockchain network for distributed evidence storage, forming an immutable carbon emission audit clue.
[0028] In this embodiment, the blockchain-based evidence storage unit solves the ultimate problem of ensuring the credibility, auditability, and tradability of carbon emission data. The process is as follows: Evidence chain encapsulation: The final emissions after calibration by the digital twin model within the accounting period, the allocation of each process, and the hash fingerprint of the original data used, model version, calibration parameter snapshot, etc., are packaged together into a structured data package; On-chain evidence storage: Submit the data packet to the blockchain network (which can be a consortium blockchain). The blockchain's consensus mechanism ensures that once the data is on the chain, it cannot be unilaterally tampered with, and the timestamp proves when the data was generated. Generate trusted credentials: After being uploaded to the blockchain, a unique transaction hash (TxID) is generated. This TxID, together with the data packet itself, constitutes a digital carbon asset report with legal and technical credibility. Financial institutions, verification agencies, and government regulatory departments can use this TxID to verify the authenticity and completeness of the report on the blockchain, which greatly simplifies the verification process and reduces trust costs.
[0029] A monitoring method for a factory carbon emission monitoring device includes the following steps: Step S1: Real-time acquisition of full-process data of the factory through multimodal sensing units set at key carbon source nodes in the factory. In step S2, the edge computing gateway receives the process data, performs data cleaning and multi-source spatiotemporal alignment processing in sequence to form a unified standardized data sequence, and uses a preliminary emission factor library to perform a fast and rough emission estimation. Step S3: The carbon monitoring server receives the standardized data sequence and drives the full-process carbon flow digital twin model to perform real-time simulation. Step S4: The full-process carbon flow digital twin model is based on material and energy balance, combined with real-time data, to dynamically calculate the inflow, retention and outflow of carbon elements in each process link, accurately calculate direct and indirect emissions, and generate process-level carbon emission intensity map. Step S5: Visualize the carbon emission intensity map, real-time calculation results, and early warning information on the management terminal; Step S6: Package the key accounting results and evidence chain data, and upload them to the blockchain for storage.
[0030] This embodiment also includes carbon emission tracing and optimization steps: When total carbon emissions or emissions from a particular process fluctuate abnormally, operators can use the management terminal to replay the dynamic simulation of carbon flow during the abnormal time period in the digital twin model. The model highlights the starting point and propagation path of abnormal carbon flow mutations, helping to locate the root cause equipment or process parameters that lead to increased emissions; The model is used to simulate and adjust suspected root cause parameters to predict their impact on final emissions, providing quantitative support for on-site operational adjustments or process optimization.
[0031] Furthermore, in step S4, the precise accounting specifically includes: For direct emissions, the full-process carbon flow digital twin model takes high-precision fuel measurement and multi-parameter monitoring at the emission outlet as direct inputs, and calibrates combustion efficiency through model inversion and data assimilation to achieve minute-level dynamic calculation of emissions; For indirect emissions, the model maps distributed power consumption data to specific process equipment, and combines real-time grid emission factors and the carbon intensity of in-plant self-generated power to allocate and trace the power consumption to the final product.
[0032] In this embodiment, direct emissions accounting is performed: the model uses fuel measurements and key emission point monitoring as strong constraints, and through data assimilation techniques (such as Kalman filtering and ensemble Kalman filtering), it continuously integrates measured data with model predictions, dynamically calibrating key state variables (such as furnace temperature field and reaction process) and parameters (such as heat transfer coefficient and reaction rate) in the model. This ensures that the accounting results not only reflect physicochemical laws but also closely match real-time measurements, achieving a much higher accuracy than simple summation.
[0033] Indirect emission tracing: The model links electricity consumption data to equipment. For example, knowing the electricity consumption of an air compressor during a specific period, combined with the real-time marginal emission factor of the power grid during that period (rather than the annual average factor), allows for accurate calculation of its corresponding carbon emissions. If the factory has its own power plant, the carbon intensity of that power plant (simultaneously calculated by this system) is used as the internal "electricity price," enabling precise allocation of carbon costs within the factory. Ultimately, the carbon footprint of one kilowatt-hour of electricity can be traced back to the ton of product it ultimately contributes.
[0034] Working principle of the invention: Step 1: High-fidelity sensing and collection of comprehensive carbon data It lies in the comprehensive and all-encompassing perception of key nodes in the carbon flow of the factory.
[0035] Multi-dimensional sensing deployment: Based on the analysis of carbon flow paths in the factory, multi-modal sensing units are strategically deployed at key nodes such as main energy input points (e.g., coal / gas pipelines), core conversion equipment (e.g., boilers, reactors), final emission points (e.g., main chimneys), and main energy-consuming equipment (e.g., air compressors, large motors) to form an Internet of Things sensing network.
[0036] Real-time data acquisition: Carbon input at the source: Through high-precision fuel flow meters and online component analyzers, the physical consumption and chemical composition (carbon content, calorific value) of fuel are obtained in real time and continuously, establishing an absolutely accurate input benchmark for the carbon quality balance of the entire plant.
[0037] Process carbon conversion: At the exhaust outlets of key equipment, multi-parameter monitors are used to capture core indicators reflecting conversion efficiency—flue gas / tail gas flow rate, CO2, CO, CH4 concentration, temperature, and pressure. These data directly reflect the completeness of combustion, the progress of the chemical reaction, and escape emissions.
[0038] Indirect carbon sources: Through distributed power consumption monitoring terminals, the electrical energy consumption that powers various process equipment is accurately measured, laying the data foundation for indirect carbon emission accounting.
[0039] Auxiliary and verification: For solid fuel stockpiles, regular volume scanning is performed using a visual recognition module to help correct consumption data and cross-verify it with flow meter data.
[0040] Step 2: Intelligent Fusion and Edge Preprocessing of Multi-Source Heterogeneous Data The collected raw data is characterized by diverse sources, varying formats, different frequencies, and the presence of noise; directly using it for calculations would lead to unreliable results. The system performs crucial data alignment and processing within the edge computing gateway.
[0041] Data cleaning: A composite algorithm (3σ statistics, rule judgment, model prediction) is used to automatically identify and remove abnormal data points such as sensor jumps, signal loss, and instantaneous interference, and to smooth reasonable fluctuations to generate a high-quality data stream.
[0042] Spatiotemporal alignment (core technology): Time unification: All data is based on a high-precision BeiDou / GPS clock and is stamped with a unified microsecond-level timestamp, eliminating timing errors caused by clock asynchrony between systems.
[0043] Spatial Association: Establish a digital coordinate system for the factory, binding the physical location of each sensor to its device ID, so that it has precise coordinates in virtual space.
[0044] Logical Synchronization: For data streams with different sampling frequencies, an "intelligent interpolation algorithm based on process event triggering" is employed. This algorithm does not perform simple linear interpolation, but instead uses key process events such as equipment start-up and shutdown, batch switching, etc., as "anchor points" on the timeline. For example, when the "boiler ignition" event occurs, the algorithm knows that fuel composition and flow rate may undergo a step change, thus employing different data splicing or preservation strategies before and after the event point to ensure that the generated data sequence is physically and logically coherent and accurate. Ultimately, all data is aligned to a unified virtual timeline that strictly corresponds to the physical process.
[0045] Rapid early warning: Utilizing aligned data and a locally stored preliminary emission factor library, the edge gateway can quickly estimate the approximate carbon emission intensity of each node. If the estimated value spikes abnormally within a short period, the system can immediately issue a primary alarm to on-site operators, achieving a second-level response to abnormal emissions.
[0046] Step 3: Dynamic simulation and accurate calculation of carbon flow throughout the entire process based on digital twins The standardized data sequences, after being fused and processed, are uploaded to the carbon monitoring server, driving a high-fidelity, real-time simulation of the entire carbon flow digital twin model. This is the core of achieving accurate accounting.
[0047] Model initialization and construction: Based on the plant's P&ID drawings and process knowledge, a static process topology network is pre-built, defining all equipment, pipelines, and carbon flow paths.
[0048] Hybrid model-driven computation: Mechanism-driven modeling: For processes such as combustion and reactions with known stoichiometry, deterministic mechanistic models based on first principles of physicochemical principles are embedded in the corresponding equipment nodes (such as boilers). This model, when inputted with real-time fuel data, air volume, etc., can calculate the theoretical flue gas composition and carbon emissions.
[0049] Data model supplement: For processes with unclear mechanisms, such as fermentation and complex synthesis, a data-driven model trained using historical data (such as an LSTM neural network) is embedded. This model predicts the carbon conversion and emissions of the unit based on real-time input process parameters (temperature, pressure, feed composition).
[0050] This hybrid approach of combining mechanism and data ensures that the model possesses both interpretability and high accuracy for clear processes, while also being able to adapt to the simulation requirements of complex processes.
[0051] Real-time data injection and dynamic calibration: The real-time data-driven engine continuously injects processed sensor data, as boundary conditions, setpoints, or correction parameters, into the corresponding nodes of the digital twin model. More importantly, the system employs data assimilation techniques (such as Kalman filtering). This technique compares and fuses the actual measured values (such as CO2 concentration) at key exhaust outlets with the predicted values at that location in real time, dynamically recalibrating key state variables (such as furnace temperature) and parameters (such as combustion efficiency and reaction rate) within the model.
[0052] Accurate accounting and traceability output: Direct emissions: Using the dynamically calibrated model described above, the actual carbon emissions of each combustion / reaction device are output in minutes, and the total direct emissions of the entire plant are obtained by summing them up.
[0053] Indirect emissions: The model accurately links electricity consumption data to specific equipment and multiplies it by the real-time marginal emission factor obtained from the power sector (or uses the real-time carbon intensity of the self-owned power plant) to calculate the carbon emissions generated by each piece of equipment and each production line due to electricity consumption, and scientifically allocates them to the final product to achieve accurate source tracing of indirect emissions.
[0054] Ultimately, the system generates process-level carbon emission intensity maps, clearly showing where carbon is concentrated and where it is lost in the form of heat maps or Sankey diagrams.
[0055] Step 4: Visual interaction, source tracing diagnosis and decision support The accounting results are presented to managers through visualization and interactive modules.
[0056] Panoramic monitoring: The management terminal displays the total carbon emissions, intensity, trend, and gap with the target for the entire plant and each department in real time.
[0057] Anomaly Root Cause Analysis: When the system issues an alert or the administrator discovers an abnormal emission in a certain process, the carbon flow animation for that period can be replayed in the digital twin model. The model will highlight the starting device and propagation path of the dramatic carbon flow change, like a "carbon flow CT scan," to quickly locate the root cause of the problem (such as a leaking valve, a malfunctioning sensor, or a decrease in the efficiency of a reaction).
[0058] Simulation optimization: Managers can perform "hypothesis analysis" in the twin model, such as simulating the replacement of high-efficiency motors, optimizing reaction temperatures, and adjusting production plans. The system will immediately predict the impact of these measures on the carbon emissions, energy consumption, and costs of the entire process, providing quantitative decision-making basis for emission reduction investment and process optimization.
[0059] Step 5: Trustworthy Evidence Preservation and Value Loop To ensure the credibility of the accounting results and meet the needs of high-credibility scenarios such as auditing and transactions, the system enables a blockchain-based evidence storage unit.
[0060] Evidence chain encapsulation: Package the final accounting results (total amount, number of processes) for a specific accounting period (such as hourly or daily), the hash value of the original data on which it is based, the model version, calibration parameters, etc., into a complete data package.
[0061] On-chain persistence: The data packet is submitted to the blockchain network. Through a distributed consensus mechanism, the data is recorded in a chronologically linked and tamper-proof block, and the timestamp provides legal proof of existence.
[0062] Generate Trustworthy Assets: Upon being uploaded to the blockchain, a unique transaction hash (TxID) is generated. This TxID, together with the data packet, constitutes a counterfeit-proof, traceable digital carbon asset report. The authenticity and completeness of the report can be independently verified through this TxID, enhancing the credibility and circulation value of the data.
[0063] 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 factory carbon emission monitoring device, characterized in that, include: The edge sensing module includes multimodal sensing units installed at key carbon source nodes within the factory. These multimodal sensing units are used to collect real-time process data directly related to carbon emissions. The key carbon source nodes include main energy input ports, key combustion / reaction equipment, main chimneys / exhaust outlets, and large energy-consuming equipment. The data fusion and edge computing module includes an edge computing gateway communicatively connected to the multimodal sensing unit. The edge computing gateway incorporates a data cleaning module, a multi-source spatiotemporal alignment module, and a preliminary emission factor library. The data cleaning module performs outlier removal and smoothing on the received raw process data. The multi-source spatiotemporal alignment module timestamps and maps the spatial locations of sensor data from different physical locations and sampling frequencies, forming a data sequence within a unified spatiotemporal framework. The preliminary emission factor library stores basic emission factors based on equipment model and fuel type. The digital twin and precise accounting module includes a carbon monitoring server, which is communicatively connected to the edge computing gateway. The carbon monitoring server internally constructs a full-process carbon flow digital twin model synchronized with the actual factory's technological flow. This full-process carbon flow digital twin model receives data sequences processed by the data fusion and edge computing module, drives the model's operation, and dynamically simulates and quantifies the flow, transformation, and final emission of carbon elements in various factory processes, with material balance and energy balance as its core, outputting a carbon emission intensity map. The visualization and interaction module includes a management terminal connected to the carbon monitoring server, which is used to display the carbon emission intensity map, real-time emission data of each node, early warning information, and simulation results of carbon emission reduction methods in real time.
2. The factory carbon emission monitoring device according to claim 1, characterized in that, The multimodal sensing unit includes: High-precision fuel flow meters and online component analyzers are used to measure fuel consumption and carbon content in real time. Multi-parameter monitoring instruments at the discharge outlets of key equipment are used to monitor flue gas flow rate, CO2, CO, CH4 concentration, temperature, and pressure parameters in real time. Distributed power consumption monitoring terminals are used to collect real-time power consumption data for individual items and equipment. The visual recognition module is used to assist in calculating the consumption of solid fuels through image recognition and volume scanning.
3. The factory carbon emission monitoring device according to claim 1, characterized in that, The multi-source spatiotemporal alignment module uses the BeiDou / GPS synchronous clock as a reference to stamp all sensor data with microsecond-level timestamps. Establish a factory coordinate system and bind the device IDs of several multimodal sensing units to their physical coordinates; For data with different sampling frequencies, an interpolation algorithm based on process event triggering is used to align all data streams to a unified virtual time axis. The process events include equipment start / stop signals, batch start signals, and batch end signals.
4. The factory carbon emission monitoring device according to claim 1, characterized in that, The full-process carbon flow digital twin model includes: The static process topology network is constructed based on the actual process flow diagram of the factory, defining all equipment nodes, pipeline connections and carbon flow paths from raw material input to product / waste output. The dynamic mechanism calculation unit is embedded in each device node of the static process topology network. The dynamic mechanism calculation unit calculates the carbon output, carbon conversion rate and carbon emissions of the node based on the carbon flow and energy flow of the input material and the process parameters of the node, and on the basis of chemical reaction equation, combustion equation and heat balance equation. The real-time data-driven engine is used to inject the real-time measurement values in the data sequence as boundary conditions or correction parameters into the corresponding dynamic mechanism calculation unit, driving the entire model to perform synchronous simulation calculations at the second / minute level.
5. A factory carbon emission monitoring device according to claim 4, characterized in that, The dynamic mechanism calculation unit adopts a hybrid calculation mode: For combustion and reaction processes with known stoichiometric ratios, deterministic mechanistic models are used for calculations. For biochemical and complex reaction processes, data-driven models are used for fitting calculations. The carbon monitoring server has a built-in model verification module. The model verification module periodically compares the virtual measurement values calculated by the model with the actual sensor measurement values. If the deviation exceeds the threshold, the model parameter calibration is automatically triggered.
6. A factory carbon emission monitoring device according to claim 1, characterized in that, It also includes a blockchain evidence storage unit connected to the carbon monitoring server; the blockchain evidence storage unit is used to generate data blocks from the key carbon emission accounting result data package output by the full-process carbon flow digital twin model, the key carbon emission accounting result data package including the carbon emission amount of each process, the total emission amount, the corresponding supporting process data hash value and timestamp, and upload it to the blockchain network for distributed evidence storage, forming an immutable carbon emission audit clue.
7. A monitoring method based on the factory carbon emission monitoring device according to any one of claims 1 to 6, characterized in that, Includes the following steps: Step S1: Real-time acquisition of full-process data of the factory through multimodal sensing units set at key carbon source nodes in the factory. In step S2, the edge computing gateway receives the process data, performs data cleaning and multi-source spatiotemporal alignment processing in sequence to form a unified standardized data sequence, and uses a preliminary emission factor library to perform a fast and rough emission estimation. Step S3: The carbon monitoring server receives the standardized data sequence and drives the full-process carbon flow digital twin model to perform real-time simulation. Step S4: The full-process carbon flow digital twin model is based on material and energy balance, combined with real-time data, to dynamically calculate the inflow, retention and outflow of carbon elements in each process link, accurately calculate direct and indirect emissions, and generate process-level carbon emission intensity map. Step S5: Visualize the carbon emission intensity map, real-time calculation results, and early warning information on the management terminal; Step S6: Package the key accounting results and evidence chain data, and upload them to the blockchain for storage.
8. The monitoring method of a factory carbon emission monitoring device according to claim 7, characterized in that, In step S4, the precise calculation specifically includes: For direct emissions, the full-process carbon flow digital twin model takes high-precision fuel measurement and multi-parameter monitoring at the emission outlet as direct inputs, and calibrates combustion efficiency through model inversion and data assimilation to achieve minute-level dynamic calculation of emissions; For indirect emissions, the model maps distributed power consumption data to specific process equipment, and combines real-time grid emission factors and the carbon intensity of in-plant self-generated power to allocate and trace the power consumption to the final product.
9. The monitoring method of a factory carbon emission monitoring device according to claim 7, characterized in that, In step S6, the key accounting results and evidence chain data include: accounting time interval, total carbon emissions of the factory, carbon emissions of each process, timestamp sequence hash of the real-time data used, model version number and calibration parameter snapshot, and digital signature of the accounting results.