System and method for carbon-health collaborative accounting of power module full life cycle

Abstract

This invention relates to a method for collaborative carbon-health accounting throughout the entire lifecycle of power modules, comprising: S1: collecting data from power modules and performing cross-stage data association and alignment based on digital twin identifiers; S2: determining the electricity carbon emission factor; calculating dynamic operating carbon emissions based on the carbon emission factor and real-time power loss; S3: calculating carbon emissions during manufacturing, use, and end-of-life stages, and normalizing them into a product-level full lifecycle carbon footprint value; S4: randomly sampling the remaining lifespan and future green electricity penetration rate using a random simulation method, calculating the probability distribution of net carbon gains from tiered utilization, and outputting retirement decision recommendations; S5: calculating cryptographic hash values, digitally signing key data packets, and generating zero-knowledge proofs. This method achieves seamless data integration across the entire manufacturing-operation-retirement chain, collaborative dynamic accounting of carbon emissions and health status, repeatable calibration of thermal resistance aging parameters, probabilistic quantification of retirement decisions, and reliable anchoring of data sources.

Description

Technical Field

[0001] This invention relates to the fields of power semiconductor devices and green low-carbon technology, and in particular to a power module full life cycle carbon-health collaborative accounting system and method. Background Technology

[0002] Driven by the accelerating global carbon neutrality process and the formation of international carbon barriers in trade, transparent management of carbon emissions throughout the entire lifecycle of power modules has become an essential requirement for the industry. Currently, the industry mainly relies on health management (PHM) systems to assess module reliability and on static life cycle assessment (LCA) methods to calculate carbon footprints. However, existing technologies have the following significant drawbacks: First, carbon management and health management goals are disconnected. Existing PHM systems focus solely on reliability and fail to account for efficiency declines caused by module aging (such as increased thermal resistance) as "hidden carbon emissions," resulting in distorted carbon data in the later stages of operation.

[0003] Secondly, there are fundamental errors in the accuracy of carbon accounting. Existing methods mostly use static thermal models, ignoring the fact that thermal resistance increases with aging, forming a positive feedback chain of "thermal resistance degradation → junction temperature error → loss deviation → cumulative error in carbon accounting," making it difficult to guarantee the accuracy of dynamic carbon accounting. At the same time, thermal resistance aging parameters are often given directly without complete experimental calibration methods and statistical significance analysis, leading to doubts about the repeatability and generalization ability of the parameters.

[0004] Third, data silos and trust issues are prominent. Process energy consumption data from the manufacturing phase is difficult to integrate with dynamic data from the operation and decommissioning phases, creating data silos. Existing blockchain-based evidence storage solutions only hash the data on the chain, proving that the data has not been tampered with, but failing to verify the authenticity of the original data itself, resulting in a "garbage in, garbage out" trust vulnerability. Furthermore, existing digital twins mostly use proprietary identifier systems, lacking interoperability and being out of sync with mainstream international standards (such as IEC 63278).

[0005] Fourth, the decision-making process for decommissioning lacks quantitative basis. The prediction of the remaining lifespan of decommissioned modules is uncertain, and the type of energy used in the secondary utilization scenario directly affects carbon gains. Existing technologies lack probability-based quantitative decision-making tools.

[0006] Fifth, it is out of step with international carbon accounting standards. Existing technology does not fully consider the specific requirements of international standards such as ISO 14067 for electricity models, functional units, recycling and distribution, and data quality, resulting in accounting results that cannot be used for official carbon labeling declarations.

[0007] Therefore, how to achieve carbon-health collaborative management, integrate data across the entire life cycle, improve the accuracy of dynamic accounting, and meet international standards has become a pressing technical challenge in this field. Summary of the Invention

[0008] The present invention aims to provide a carbon-health collaborative accounting system and method for the entire life cycle of power modules to overcome the shortcomings of the prior art. The technical problem to be solved by the present invention is achieved through the following technical solutions.

[0009] According to a first aspect of this application, a method for carbon-health co-accounting throughout the entire lifecycle of a power module is provided, comprising: S1: Data from the manufacturing, operation, and decommissioning phases of the power module are collected through end-side sensors and data interfaces, and cross-phase data association and alignment are performed based on digital twin identifiers; the junction temperature of the power module is estimated in real time using an electrothermal network model that includes online thermal resistance correction, and the online thermal resistance correction is dynamically performed based on the equivalent number of thermal cycles extracted from the case temperature data through a sliding window thermal cycle counting algorithm; S2: Determine the power carbon emission factor based on the energy type supplied by the power module, and differentiate between the market-based method and the location-based method. The calculation method of the carbon emission factor should be consistent within the same accounting period. Calculate the dynamic operating carbon emissions based on the carbon emission factor and real-time power loss. S3: Based on energy consumption and material data during the manufacturing stage, dynamic carbon emissions during operation, and recycling and processing data during the decommissioning stage, calculate the carbon emissions during the manufacturing, use, and end-of-life stages, and normalize them into a product-level full life-cycle carbon footprint value. S4: Based on the digital twin of the decommissioned power module and the expected energy structure of the cascade utilization scenario, random sampling is performed on the remaining lifespan and future green electricity penetration rate through random simulation method to calculate the probability distribution of net carbon gain from cascade utilization, and decommissioning decision suggestions are output according to the preset confidence threshold. S5: Calculate cryptographic hash values ​​for manufacturing parameters, operational data summaries, decommissioning decision data, and full lifecycle carbon footprint values ​​respectively; digitally sign key data packets; store the hash values ​​and signature information on the blockchain; and generate zero-knowledge proofs. The inputs to the zero-knowledge proofs include the public key hash of the digital signature of the data source and proof of signature validity.

[0010] Preferably, the digital twin identifier adopts an internationally unique identifier, including manufacturer code, product type code, batch number and serial number, and is written to the module's non-volatile memory or QR code; the digital twin model encapsulates manufacturing, operation and decommissioning data into sub-models and associates them with unique identifiers.

[0011] Preferably, the sliding window thermal cycle counting algorithm uses a fixed-length window to cache historical shell temperature data and applies simplified rainflow counting to extract closed thermal cycles, achieving a time complexity of O(1) and a memory footprint of less than 50KB; the online thermal resistance correction formula is: R_th(jc)(t) = R_th0 × (1 + α·N_eq^β); Where R_th(jc)(t) is the dynamic junction thermal resistance corrected at time t, R_th0 is the factory thermal resistance baseline value, N_eq is the equivalent thermal cycle number, α and β are thermal resistance aging parameters, and the value of β ranges from 0.85 to 1.15.

[0012] Preferably, the thermal resistance aging parameters α and β are calibrated through accelerated aging experiments. The training group is fitted using the nonlinear least squares method, and the average relative error of the verification group is less than 5%. The parameters are calibrated separately for different packaging processes and can be updated remotely in the cloud.

[0013] Preferably, the electrothermal network model is a Foster-Cauer hybrid model, and the junction temperature is estimated using a sliding mode observer. The observer equation is: x_hat_dot = A·x_hat + B·u + L·sign(y - C·x_hat) Where x_hat is the state vector estimate, A, B, and C are the system matrices, u is the power loss input, y is the measured case temperature, and L is the observer gain matrix determined by pole placement; under standard pulse load test conditions, the junction temperature estimation error margin is less than ±5℃.

[0014] Preferably, the determination method for the electricity carbon emission factor is as follows: when using a market-based approach, the residual electricity portfolio emission factor or a specific emission factor agreed upon in a power purchase agreement is used, in conjunction with an energy attribute certificate or power purchase agreement; when using a location-based approach, the officially published regional average electricity portfolio factor is used; the formula for calculating the dynamic operating carbon emissions is: C_rate(t) = P_loss(t) × EF_electricity(t) Where P_loss(t) is the real-time power loss at time t, EF_electricity(t) is the carbon emission factor of electricity at time t, and the carbon emission during the operation phase is calculated by integrating this rate over time. For renewable energy direct connection scenarios, the green electricity ratio is obtained in real time through the communication interface of the electricity meter or inverter; for ordinary grid access points, the green electricity ratio is allocated on an hourly basis.

[0015] Preferably, the formula for calculating the life-cycle carbon footprint in S3 is: LCACF = (C_manufacturing + C_use + C_EoL) / (P_rated × T_lifetime) Where P_rated is the module's rated power, T_lifetime is the expected service life, and LCACF is measured in gCO2-eq / kW·h, which complies with the functional unit requirements of ISO 14067. The formula for calculating carbon emissions during the manufacturing stage is as follows: C_manufacturing=Σ(E_process,i×EF_electricity)+Σ(M_material,j×EF_material,j)+ C_overhead; Where E_process,i is the energy consumption of the i-th process, M_material,j is the amount of the j-th material, EF_material,j is the corresponding carbon emission factor, and C_overhead is the allocation of public energy consumption. The carbon emissions at the end of the life cycle stage are calculated using the circular footprint formula: C_EoL = E_recycling - R_2 × V_sub × A_sub Where R_2 is the output recycling rate, E_recycling is the energy consumption for recycling, V_sub is the carbon gain from replacing virgin materials, and A_sub is the replacement ratio; Data quality was assessed using the Pedigree Matrix, with a score of less than 3.0. Uncertainty was also quantified using Monte Carlo simulation, with expanded uncertainty less than ±15%.

[0016] Preferably, the stochastic simulation method described in S4 embeds the cyclic footprint formula into the simulation process. The input variables include the probability distribution of remaining lifetime, future value of recycled materials, and future green electricity penetration rate, and the output is the probability that the net carbon gain is greater than zero. When the probability exceeds a preset confidence threshold and the remaining lifetime meets the minimum requirement, a tiered utilization suggestion is output. For existing power modules without historical operating data, a prior distribution is constructed based on modules in the same batch that have historical operating data, and updated in real time in conjunction with the currently collected data under a Bayesian framework.

[0017] Preferably, the digital signature in S5 is executed by a hardware security module built into the key metering equipment, supporting the national cryptographic algorithms SM2 / SM3 / SM4 and internationally recognized algorithms; the cryptographic hash value of the key data packet is constructed into a Merkle tree, and the root hash value is uploaded to the blockchain for evidence storage; the zero-knowledge proof enables a third-party auditor to verify the compliance of the carbon footprint statement without obtaining the original process data.

[0018] According to a second aspect of this application, a power module lifecycle carbon-health co-accounting device employing the above-described power module lifecycle carbon-health co-accounting method is provided, characterized in that it comprises: The data acquisition and alignment module is used to collect cross-stage data and perform time-series alignment based on digital twin identifiers; The dynamic junction temperature estimation module estimates the junction temperature based on the electrothermal network model and online thermal resistance correction. The dynamic carbon emission accounting module distinguishes between market-based and location-based carbon emission factors to calculate operational carbon emissions. The full life cycle carbon footprint accounting module calculates carbon emissions during the manufacturing, use, and end-of-life stages and normalizes them into product-level carbon footprint values. The retirement decision-making module uses random simulation to calculate the probability of net carbon gain from tiered utilization and outputs decision recommendations. The trusted evidence storage module digitally signs key data packets, stores them on the blockchain, and generates zero-knowledge proofs.

[0019] The embodiments of the present invention have the following advantages: First, by using internationally standardized digital twin identifiers to align the three-stage data across time periods, carbon data silos are eliminated, and data connectivity across the entire manufacturing-operation-retirement chain is achieved, laying a data foundation for accurate full life-cycle accounting.

[0020] Secondly, the incremental loss caused by module aging is quantified into additional carbon emissions through online thermal resistance correction technology, filling the "hidden carbon" gap in traditional methods; a sliding window thermal cycle counting algorithm with a time complexity of O(1) is adopted, with a memory occupation of less than 50KB, so that the above correction can be executed in real time on resource-constrained automotive-grade microcontrollers, and the calculation accuracy can reach more than 97%.

[0021] Furthermore, the full process of large-sample experimental calibration of thermal resistance aging parameters (with a sample size of no less than 30) is disclosed, covering mainstream packaging processes such as silver sintering, tin-silver-copper solder, and transient liquid phase diffusion soldering. It provides goodness of fit, 95% confidence interval, and Weibull distribution verification to ensure the repeatability and feasibility of the parameters.

[0022] Furthermore, by embedding the cycle footprint formula into Monte Carlo simulation, the net carbon gain of decommissioned and reused units is quantified in the form of a probability distribution. Based on the confidence threshold, quantifiable decommissioning decision-making suggestions are given, effectively avoiding the risks of blind reuse.

[0023] Finally, at the hardware trust root level, an automotive-grade hardware security module conforming to the AEC-Q100 Grade 0 standard is introduced to digitally sign the original data. Combined with zero-knowledge proofs, auditability is achieved while protecting business privacy, fundamentally solving the trust vulnerability in blockchain evidence storage that "only prevents tampering, but not forgery". Attached Figure Description

[0024] Figure 1 This is a flowchart of the steps of a power module full life cycle carbon-health co-accounting method according to the present invention; Figure 2 This is a schematic diagram of the structure of a power module full life cycle carbon-health collaborative accounting system according to the present invention. Detailed Implementation

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] It should be noted that the above detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0027] In this specification, "thermal resistance" specifically refers to the junction-to-case thermal resistance R_th(jc) of the power module, "junction temperature" specifically refers to the junction temperature T_j of the active region of the power device, "equivalent thermal cycle count" refers to the cumulative count value N_eq of the actual variable amplitude thermal cycle converted into an equivalent constant amplitude thermal cycle after processing by the rainflow counting method, and "dynamic operating carbon emissions" refers to the carbon emission accounting results of the operating phase that takes into account the changes in energy structure over time.

[0028] like Figure 1 As shown, the power module full life cycle carbon-health co-accounting method of this application mainly includes the following steps: S1: Full lifecycle multimodal data acquisition and dynamic junction temperature estimation. Collect process energy consumption and quality inspection data during the power module manufacturing stage, electrical parameters and multi-physics field state data during the operation stage, and recycling and processing data during the decommissioning stage. Cross-stage data association and alignment are performed based on digital twin identifiers conforming to the IEC 63278 series standards. During the operation stage, an improved electrothermal network model with online thermal resistance correction is used for dynamic junction temperature estimation. The thermal resistance correction is based on the equivalent thermal cycle number N_eq extracted from the shell temperature fluctuation through the improved sliding window rainflow counting method (ISWR) and is performed online in real time. The ISWR algorithm has a time complexity of O(1) per sampling point and a memory usage of <50KB, which is suitable for real-time operation of automotive-grade microcontrollers.

[0029] Specifically, during the manufacturing phase, a unique digital twin identifier is generated for each power module. The identifier format is compatible with the ISO / IEC 15459 international unique identifier standard and includes the manufacturer code, year, product type code, production batch number, and serial number, for example, the format "CMP-2026-SIC-001-0001". The identifier encoding is solidified in two ways: by writing it into the power module's non-volatile memory (EEPROM) and by printing it as a Data Matrix QR code. The non-volatile memory supports offline reading, ensuring that the identifier remains valid throughout the entire lifecycle of the module.

[0030] The digital twin model is organized according to a sub-model structure, encapsulating manufacturing data, operational data, and decommissioning data into independent sub-models, each linked by a unique identifier. The edge controller and cloud platform exchange data using the OPC UA protocol, ensuring real-time data streams are synchronized with the digital twin model. When the power module enters the decommissioning stage, the identification code is retained in the recycling tag, supporting digital twin reconstruction in tiered utilization scenarios and achieving full lifecycle traceability.

[0031] For example, data collection during the manufacturing phase can be divided into two schemes based on the production mode: For vertically integrated manufacturing (IDM) models, measured energy consumption and material consumption data for each process are directly collected from the Manufacturing Execution System (MES), Enterprise Resource Planning (ERP), and Programmable Logic Controllers (PLCs). Specific data collected includes: unit energy consumption (kWh / wafer) for wafer manufacturing processes (including epitaxial growth, photolithography, ion implantation, etc.); energy consumption and protective atmosphere consumption for packaging processes (including sintering / reflow soldering, molding, electroplating, etc.); energy consumption for testing processes (including static parameter testing, dynamic parameter testing, and aging testing); usage of key materials such as silicon wafers, ceramic substrates, bonding wires, and molding compounds; and the allocated value of common energy consumption such as factory lighting and air conditioning based on product output. Key metering equipment has a built-in hardware security module that digitally signs each uploaded data packet, and the signing public key is registered on the blockchain.

[0032] For fabless models, the Scope 3 carbon footprint statement provided by the wafer supplier that complies with the Semiconductor Climate Coalition (SCC) standard is preferred; if the supplier's data is missing, the industry average data for the corresponding wafer type in the Ecoinvent 3.9 database is used, and the corresponding data quality level (DQR) is clearly marked in the accounting report.

[0033] Data acquisition during operation includes: continuous acquisition of electrical parameters such as case temperature T_c(t), bus voltage, and phase current of the power module by end-side sensors, as well as the proportion of power supply energy types of the converter. For direct connection scenarios of renewable energy sources such as photovoltaic and wind power, the green electricity ratio is obtained in real time through the inverter communication interface; for ordinary grid connection points, the green electricity ratio is allocated at the hourly level according to the power purchase agreement or the day-ahead green electricity forecast curve published by the grid.

[0034] Dynamic junction temperature estimation is achieved based on an improved electrothermal network model, which introduces an online thermal resistance correction element into the traditional Foster thermal network. Specifically, real-time junction temperature estimation is performed using a sliding mode observer (SMO). The state-space equation of the sliding mode observer is as follows: x_hat_dot = A·x_hat + B·u + L·sign(y - C·x_hat) Where x_hat is the estimated value of the temperature vector of the thermal network nodes, A, B, and C are the system matrix, input matrix, and output matrix of the thermal network, respectively, u is the power loss, y is the shell temperature measurement value, L is the observer gain matrix determined by pole placement, and sign(·) is the sign function. The observer gain L is designed using pole placement to ensure that the dynamic convergence time of the state estimation error is less than 2 seconds. A sliding mode observer is used instead of an extended Kalman filter (EKF) mainly to avoid the divergence risk of the EKF when the system observability is insufficient, thus enhancing the robustness of the estimation. Under standard pulsed load conditions, the junction temperature estimation error margin is less than ±5℃.

[0035] The sliding window thermal cycle counting algorithm (ISWR) is as follows: the core of online thermal resistance correction is to obtain the equivalent number of thermal cycles N_eq in real time. This application uses an improved sliding window thermal cycle counting algorithm (ISWR) to extract N_eq from the shell temperature time series T_c(t). Specifically, the algorithm maintains a sliding window of fixed length L (preferably L is 1000 data points). Within the window, a simplified variant of rainflow counting conforming to the ASTM E1049-85 standard is applied, extracting only closed loops and discarding residual terms to reduce the computational load. For each new sampling point, it only needs to be compared with the extreme points within the window, with a time complexity of O(1) and a space complexity of O(L).

[0036] The measured performance on an ARM Cortex-M4 processor (100MHz) is as follows: CPU utilization is less than 8%, RAM usage is less than 50KB, and single thermal cycle extraction latency is less than 5ms, which meets the real-time processing requirements of a 10kHz case temperature sampling rate. Compared with the standard rainflow counting method, the ISWR algorithm has an equivalent thermal cycle extraction error of less than 3%, meeting engineering accuracy requirements.

[0037] The online correction of thermal resistance specifically involves: after obtaining the equivalent number of thermal cycles N_eq, performing online correction of the dynamic junction thermal resistance according to the power-law degradation model. R_th(jc)(t) = R_th0 × (1 + α·N_eq^β) Where R_th0 is the factory-specified nominal thermal resistance baseline value, α and β are thermal resistance aging parameters, α>0, and β ranges from 0.85 to 1.15 to ensure that the thermal resistance increases monotonically with the number of cycles and the rate of increase slows down. N_eq is the cumulative equivalent number of thermal cycles up to time t. The physical meaning of the thermal resistance aging parameters α and β originates from the failure mechanism of the packaging interface layer: for silver sintered interface layers, thermal cycling causes microcrack propagation in the sintered silver layer; for tin-silver-copper solder interface layers, thermal cycling causes solder creep-fatigue damage and thickening of the intermetallic compound layer; for transient liquid phase diffusion soldering (TLP) interface layers, thermal cycling causes diffusion growth of the high-melting-point intermetallic compound layer. Due to the different failure mechanisms and material properties of different packaging processes, the α and β parameters must be calibrated separately for specific packaging processes.

[0038] The specific calibration procedure for thermal resistance aging parameters is as follows: Thermal resistance aging parameters α and β are determined through the following calibration procedure: The first step is to randomly select no fewer than 30 samples from the same batch of power modules and randomly divide them into a training group (about two-thirds) and a validation group (about one-third). The second step is to perform accelerated aging tests on the training group samples. The test conditions are: temperature cycling range of -40℃ to 150℃ (temperature difference of 190℃), heating rate of 15℃ / min, high and low temperature residence time of 15 minutes each, and cycle period of 60 minutes. The test is paused every 200 cycles, and the static thermal resistance is measured according to the JESD51-14 standard at a reference temperature of 25℃. The third step involves using the (N, R_th) data pairs of each cycle point of each sample as training data. The Levenberg-Marquardt nonlinear least squares algorithm is used to fit the parameters of the power law model R_th(N) = R_th0 × (1 + α·N^β). The fitted values ​​of α and β, the goodness of fit R², and the 95% confidence interval are output. The training group sample size n≥30, the validation group determination coefficient R²>0.9, and the residuals conform to the Weibull distribution through the KS test (p>0.05). The fourth step is to verify the model's prediction accuracy on the validation group samples. The average relative error between the predicted and measured thermal resistance values ​​of the validation group samples should be less than 5%, the maximum relative error should be less than 10%, and the root mean square error (RMSE) should be less than 5%. The fifth step is to store the calibrated α and β parameters and their corresponding packaging process labels into the cloud platform parameter library. During the operation of the power module, the cloud platform will periodically refit the parameters based on the continuously transmitted operating data and issue updates through the OTA mechanism to achieve continuous adaptive optimization of the parameters.

[0039] Specifically, the parameter calibration examples for the three mainstream packaging processes are as follows: the fitting result for the silver sintered interface layer is α = 2.1 × 10⁻⁶. -4 β=0.92, goodness of fit R²=0.94, 95% confidence intervals are α∈[1.9×10 -4 , 2.3×10 -4 ] and β∈[0.89, 0.95], the average prediction error of the validation group was 3.8%; the fitting result of the tin-silver-copper (SAC305) solder interface layer was α=5.8×10 -4 The fitting result for the interface layer of transient liquid phase diffusion welding (TLP, Cu-Sn system) was α = 3.5 × 10⁻⁶, with β = 0.88, goodness of fit R² = 0.91, and average prediction error of 4.2% for the validation group. -4 The results showed that β=0.95, goodness of fit R²=0.93, and the average prediction error of the validation group was 3.9%. These parameters are specific to the process formulation and thermal cycling conditions. If the material formulation is changed or the thermal cycling test conditions are altered, they must be recalibrated according to the same procedure.

[0040] S2 operates dynamically in carbon emission accounting, strictly distinguishing the calculation methods for electricity carbon emission factors: when using a market-based approach, the residual electricity portfolio emission factor or a specific emission factor agreed upon in the power purchase agreement is used; when using a location-based approach, the regional average electricity portfolio factor (AEF) is used. The calculation methods for the carbon emission factors remain consistent within the same accounting period.

[0041] By combining the module's real-time power loss P_loss(t) and ΔP_loss(t), the dynamic operating carbon emissions C_mix(t) considering energy type are calculated. The power supply ratio is obtained using one of the following methods depending on the actual scenario: Direct connection scenario: For dedicated converters that are directly connected to photovoltaic and wind power, the green electricity ratio can be obtained in real time through the communication interface of the electricity meter or inverter. Grid scenario: For ordinary grid access points where the energy source cannot be distinguished in real time, the time-average method is adopted, and the proportion of green electricity is allocated at the hourly level according to the power purchase agreement or the day-ahead green electricity forecast curve published by the grid.

[0042] In some embodiments, this step combines the real-time electrical and thermal parameters of the power module with the power supply energy structure to calculate the dynamic carbon emissions during operation. The formula for calculating dynamic operating carbon emissions is: C_rate(t) = P_loss(t) × EF_electricity(t) Wherein, P_loss(t) is the real-time power loss of the power module at time t, calculated from the current, voltage, and junction temperature parameters; EF_electricity(t) is the electricity carbon emission factor at time t; the carbon emissions during the operation phase are calculated by integrating this rate over time. Two methods are strictly distinguished in determining the electricity carbon emission factor: when using the market-based approach, the residual power portfolio emission factor or a specific emission factor agreed upon in the power purchase agreement is used, which must be used in conjunction with Energy Attribute Certificates (EACs) or Power Purchase Agreements (PPAs); when using the location-based approach, the officially published regional average power portfolio factor (AEF) is used. The calculation method for the carbon emission factor remains consistent within the same accounting period to ensure methodological consistency in the accounting results.

[0043] Furthermore, in step S2, the additional loss power caused by health degradation can optionally be quantified. Specifically, a semantic network is constructed including manufacturing carbon nodes, operating carbon nodes, decommissioned carbon nodes, failure mode nodes, and energy type nodes. A lightweight graph neural network (GNN) is used to embed the quantified relationships between nodes, inferring the current failure probability and the resulting additional loss power ΔP_loss(t). Summing P_loss(t) and ΔP_loss(t) and substituting it into the above formula can further improve the accuracy of carbon emission quantification. This step is an optional optimization scheme and can be omitted in resource-constrained scenarios.

[0044] S3 product-level full lifecycle carbon footprint accounting In some embodiments, this step calculates the product-level full lifecycle carbon footprint declaration value (LCACF) based on the multi-stage data collected and calculated in S1 and S2.

[0045] The LCACF calculation formula is as follows: LCACF = (C_manufacturing + C_use + C_EoL) / (P_rated × T_lifetime) Where P_rated is the rated power of the power module (unit: kW), T_lifetime is the expected service life (unit: hours), and LCACF is in kg CO2-eq / kW, which complies with the functional unit requirements of ISO 14067.

[0046] The formula for calculating carbon emissions (C_manufacturing) during the manufacturing stage is as follows: C_manufacturing = Σ(E_process,i × EF_electricity) + Σ(M_material,j× EF_material,j) + C_overhead Where E_process,i is the energy consumption of the i-th process, EF_electricity is the regional electricity carbon emission factor of the manufacturing location, M_material,j is the amount of the j-th key material used, EF_material,j is the carbon emission factor of the corresponding material, and C_overhead is the carbon emission amount of the plant's public energy consumption allocated to output.

[0047] Specifically, taking the IDM model calculation of a 20kW rated power SiC power module (silver sintered packaging) as an example: the energy consumption of the wafer process is 4 kWh / module; the energy consumption of the packaging process (silver sintering, reflow soldering, molding, and electroplating combined) is about 3 kWh / module; the energy consumption of the testing process is about 5.6 kWh / module; the carbon contained in the protective gas (nitrogen) is about 0.425 kgCO2e / module; the carbon contained in the key materials (ceramic substrate, bonding wire, molding compound, and shell terminals) is about 0.249 kgCO2e / module; the energy consumption of the utility is about 0.25 kgCO2e / module; taking the regional average electricity carbon emission factor of 0.5 kgCO2e / kWh, the total carbon emission of the manufacturing stage, C_manufacturing, is about 7.22 kgCO2e / module.

[0048] Carbon emissions during the usage phase are the integral of dynamic operating carbon emissions over the entire life cycle: C_use = C_mix(t)dt Taking the above example, with an average power loss of 100W and a total energy consumption of 2000 kWh, and a location-based method factor AEF of 0.45 kgCO2 / kWh, then C_use = 900 kgCO2e / module.

[0049] The end-of-life carbon emissions C_EoL are calculated using the circular footprint formula (CFF): C_EoL = E_recycling - R_2 × V_sub × A_sub Where R_2 is the output recycling rate (typically 0.5 to 0.7 in open-circuit recycling scenarios), E_recycling is the energy consumption for recycling (kgCO2e), V_sub is the carbon gain from replacing virgin materials (kgCO2e / kg), and A_sub is the replacement ratio (typically 0.8 to 1.0). Taking the above example, the total mass of the module is 120g, and the copper substrate and shell are recyclable (60% of the mass). Taking R_2=0.6, E_recycling=0.5 kgCO2e, V_sub=2 kgCO2e / kg, and A_sub=0.9, then C_EoL is approximately 0.422 kgCO2e / module.

[0050] It also includes an example of LCACF calculation and uncertainty quantification: Substitute the values ​​of each stage above into the LCACF formula and calculate based on the expected lifespan T_lifetime = 20,000 hours: LCACF = (7.22 + 900 + 0.422) / (20 × 20000) × 1000 = 2.27 g CO2-eq / kW·h;; Furthermore, the three-stage data are scored using the Pedigree Matrix across five dimensions: reliability, completeness, temporal relevance, geographical relevance, and technical relevance. A weighted composite data quality score (DQR) is then calculated, requiring a DQR below 3.0. Simultaneously, the 95% confidence interval for the LCACF is calculated using Monte Carlo simulation (at least 10,000 samples), requiring the expanded uncertainty (k=2) to be below ±15%. Taking the example above, the 95% confidence interval for the LCACF is [2.10, 2.44], with an expanded uncertainty of approximately ±7.3%, meeting the requirements.

[0051] The S4 decommissioned module's tiered utilization low-carbon decision-making is based on the digital twin of the decommissioned module and the expected energy type distribution of the target tiered utilization scenario. The Monte Carlo simulation method is used to randomly sample uncertain parameters such as remaining lifespan and future green electricity penetration rate, and the probability distribution of net carbon gain C_net is calculated based on the CFF formula. Based on the preset confidence threshold, the decision-making suggestions of tiered utilization or direct recycling are output.

[0052] In some embodiments, step S4 provides probability-based decision support for the low-carbon cascade utilization of decommissioned power modules. Specifically, a digital twin of the decommissioned power module is established, its full lifecycle operation record is read, and the probability distribution of its current health status and remaining lifespan (RUL) is estimated. Based on the target cascade utilization scenario, the distribution of energy structure parameters for the future scenario is set, including the distribution of the annual growth rate of future green electricity penetration. The Cycle Footprint Formula (CFF) is embedded into the Monte Carlo simulation process, using uncertain parameters such as remaining lifespan, future green electricity penetration rate, and future recycled material value as input variables. At least 10,000 random samples are performed to calculate the probability distribution of net carbon gain (C_net), and the probability P (C_net > 0) that the net carbon gain is greater than zero is calculated.

[0053] When P(C_net > 0) exceeds the preset confidence threshold (preferably, the threshold is set to 80%) and the remaining lifespan meets the minimum service life requirement for the tiered utilization scenario, the system outputs "Recommend tiered utilization"; otherwise, it outputs "Recommend direct recycling".

[0054] For existing power modules without historical operating data, a similar module migration-based approach is adopted: a prior distribution of remaining lifetime is constructed based on modules in the same batch that already have historical operating data, and the posterior distribution is updated in real time under a Bayesian framework as detection data accumulates, thereby outputting decision suggestions and expanding the system's applicability to existing modules.

[0055] Taking a SiC power module that has been decommissioned after 8 years of operation as an example, a digital twin is constructed, and the cascaded utilization scenario is set as "home energy storage" (low-power cycle). An uncertainty distribution is established: The remaining lifespan is RUL N(4.5, 0.68) years; the annual growth rate of green electricity penetration over the next 5 years is g N(6%, 2%). The CFF formula is embedded into a Monte Carlo simulation process, and 10,000 simulations are performed on the net carbon gain C{net}. The simulation results show that P(C{net} > 0) = 97%, which is greater than the preset threshold of 80%, and the system outputs a "recommend tiered utilization" decision.

[0056] For existing modules without historical data, a prior distribution is constructed based on modules in the same batch that already have historical operational data, and the decision is output after Bayesian updating.

[0057] S5 is a trusted carbon storage system based on blockchain and zero-knowledge proofs, which stores the cryptographic hash values ​​of key manufacturing parameters, operational data summaries, decommissioning decision data, and final carbon footprint labels on the blockchain.

[0058] In some embodiments, step S5 combines hardware trust root anchoring with cryptographic techniques to achieve end-to-end trust from the data generation source to blockchain notarization, including: Regarding the hardware root of trust, key metering devices (electricity meters, temperature sensors) incorporate automotive-grade hardware security modules (HSMs) compliant with the AEC-Q100 Grade 0 standard. These modules operate within a temperature range of -40℃ to 150℃ and support Chinese national cryptographic algorithms SM2 / SM3 / SM4 as well as internationally recognized algorithms. Each uploaded data packet is digitally signed, and the signing public key is registered on the blockchain. Optionally, it is compatible with domestically produced automotive-grade security chips such as the National Technology Z32HUB and Unisoc THD89, ensuring a secure and controllable supply chain.

[0059] In terms of blockchain evidence storage, the cryptographic hash values ​​of key manufacturing parameters, operational data summaries, decommissioning decision results, and final carbon footprint labels are constructed into a Merkle tree. The root hash value is periodically uploaded to the blockchain platform for evidence storage, and it is compatible with domestic consortium blockchains such as ChainMaker.

[0060] Regarding zero-knowledge proofs, compliance proofs for the carbon footprint calculation process are generated using zero-knowledge proofs (preferably, the zk-SNARK scheme). The input to the zero-knowledge proof includes not only a data digest but also the hash of the digital signature public key of the data source and the signature validity verification result. This allows a third-party auditing institution to verify the following two points simultaneously without obtaining the original process data: first, the original data was generated by a trusted hardware device (verified via digital signature); second, the calculation process of the entire lifecycle carbon footprint conforms to the established accounting rules (verified via zero-knowledge proof). This mechanism fundamentally solves the trust vulnerability of traditional blockchain evidence storage, which "only prevents tampering, but not forgery."

[0061] like Figure 2 As shown, the power module full lifecycle carbon-health collaborative accounting system in one embodiment of this application adopts a four-layer architecture of "cloud-edge-device" collaboration, specifically including the following layers: The first layer is the end-side data acquisition layer, deployed inside the manufacturing equipment, power module body, and converter. It includes high-precision temperature sensors, current / voltage sensors, energy type identification units, and automotive-grade hardware safety modules that comply with AEC-Q100 Grade 0 standards. It is responsible for the acquisition of raw data and hardware-level digital signatures.

[0062] The second layer is the edge intelligence layer, deployed on the local edge controller. It includes a data preprocessing unit, a lightweight junction temperature estimation and online thermal resistance correction unit based on the ISWR algorithm, and an optional lightweight graph neural network inference unit. The edge intelligence layer is responsible for real-time data processing and rapid response. All algorithms are adapted to run on resource-constrained platforms such as ARM Cortex-M4 or RISC-V processors. Specifically, the high-precision failure identification model trained in the cloud is compressed into a lightweight model with a size of less than 500KB, a single inference latency of less than 10ms, and a memory footprint of less than 128KB through knowledge distillation technology, and then deployed to the edge intelligence layer.

[0063] The third layer is the cloud platform layer, which includes a data storage and management module, a full lifecycle carbon footprint accounting engine, a retirement decision Monte Carlo simulation module, and a parameter self-learning and OTA update module. It is responsible for performing complex model training and compliant calculations. The cloud platform layer is compatible with domestic blockchain platforms and supports domestic consortium blockchains such as ChainMaker when storing carbon data.

[0064] The fourth layer is the application service layer, which provides a human-computer interaction interface and includes a carbon footprint report automatic generation module, a retirement decision support module, a trusted carbon label management module, and a digital signature public key filing library, providing differentiated services for different roles such as manufacturers, operators, and regulatory agencies.

[0065] Furthermore, based on the above four-layer architecture of "cloud-edge-device" collaboration, the power module full lifecycle carbon-health collaborative accounting system is divided into the following modules: The data acquisition and alignment module is used to acquire process energy consumption data and material data during the power module manufacturing stage, electrical and thermal status data and power supply energy type ratio data during the operation stage, and recycling and processing data during the decommissioning stage through end-side sensors and data interfaces, and to perform cross-stage time-series alignment of the data based on digital twin identifiers. The dynamic junction temperature estimation module is used to estimate the junction temperature of the power module in real time using an electrothermal network model that includes online thermal resistance correction. The online thermal resistance correction is dynamically performed based on the equivalent number of thermal cycles extracted from the shell temperature time series by a sliding window thermal cycle counting algorithm. The time complexity of the algorithm is O(1), and the memory usage is less than 50KB. The dynamic carbon emission accounting module is used to distinguish between market-based methods and location-based methods to determine the electricity carbon emission factor, and to calculate the dynamic operating carbon emission based on the electricity carbon emission factor and real-time power loss. The full life cycle carbon footprint accounting module is used to normalize the carbon emissions of the three stages of manufacturing, use, and end of life cycle into a product-level full life cycle carbon footprint value. The end of life cycle stage is calculated using the circular footprint formula, with a data quality score of less than 3.0 and an expanded uncertainty of less than ±15%. The retirement decision module is used to quantify the probability distribution of the net carbon gain of the retirement module through random simulation and output retirement decision suggestions based on a preset confidence threshold. The trusted evidence storage module is used to digitally sign key data packets, store cryptographic hash values ​​on the blockchain, and generate zero-knowledge proofs containing the hash of the data source signature public key and proof of signature validity.

[0066] In some embodiments, an electronic device is provided for performing the above-described method, including a processor and a memory. The memory stores an executable computer program, which, when executed by the processor, implements all steps of the above-described full lifecycle carbon-health co-accounting method. The memory can be volatile or non-volatile, such as dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, or other forms of solid-state memory. The processor can be a general-purpose processor, digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device. The electronic device may also include a communication interface for data interaction with external sensors, blockchain networks, and cloud platforms via wired or wireless means.

[0067] In some embodiments, a computer-readable storage medium is provided on which a computer program is stored. When the computer program is executed by a processor, it implements all the steps of the above-described whole-lifecycle carbon-health co-accounting method. The computer-readable storage medium may be any medium capable of storing program code, such as an optical disc, magnetic disk, USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), or optical disc.

[0068] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

[0069] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0070] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that the embodiments of this application described herein can be implemented in sequences other than those illustrated or described herein.

[0071] Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or apparatus.

[0072] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways, such as rotated 90 degrees or in other orientations, and the spatial relative descriptions used herein will be interpreted accordingly.

[0073] In the detailed description above, reference has been made to the accompanying drawings, which form part of this document. In the drawings, similar symbols typically identify similar parts unless the context otherwise indicates otherwise. The illustrated embodiments described in the detailed specification, drawings, and claims are not intended to be limiting. Other embodiments may be used and other changes may be made without departing from the spirit or scope of the subject matter presented herein.

[0074] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for carbon-health co-accounting throughout the entire lifecycle of a power module, characterized in that, include: S1: Collect data from the power module manufacturing, operation and decommissioning stages through end-side sensors and data interfaces, and perform cross-stage data association and alignment based on digital twin identifiers; The junction temperature of the power module is estimated in real time using an electrothermal network model that includes online thermal resistance correction. The online thermal resistance correction is dynamically performed based on the equivalent number of thermal cycles extracted from the case temperature data using a sliding window thermal cycle counting algorithm. S2: Determine the power carbon emission factor based on the energy type supplied by the power module, and differentiate between the market-based method and the location-based method. The calculation method of the carbon emission factor should be consistent within the same accounting period. Calculate the dynamic operating carbon emissions based on the carbon emission factor and real-time power loss. S3: Based on energy consumption and material data during the manufacturing stage, dynamic carbon emissions during operation, and recycling and processing data during the decommissioning stage, calculate the carbon emissions during the manufacturing, use, and end-of-life stages, and normalize them into a product-level full life-cycle carbon footprint value. S4: Based on the digital twin of the decommissioned power module and the expected energy structure of the cascade utilization scenario, random sampling is performed on the remaining lifespan and future green electricity penetration rate through random simulation method to calculate the probability distribution of net carbon gain from cascade utilization, and decommissioning decision suggestions are output according to the preset confidence threshold. S5: Calculate cryptographic hash values ​​for manufacturing parameters, operational data summaries, decommissioning decision data, and full lifecycle carbon footprint values ​​respectively; digitally sign key data packets; store the hash values ​​and signature information on the blockchain; and generate zero-knowledge proofs. The inputs to the zero-knowledge proofs include the public key hash of the digital signature of the data source and proof of signature validity.

2. The power module full life cycle carbon-health co-accounting method according to claim 1, characterized in that, The digital twin identifier uses an internationally unique identifier, which includes the manufacturer code, product type code, batch number and serial number, and is written to the module's non-volatile memory or QR code; the digital twin model encapsulates manufacturing, operation and decommissioning data into sub-models and associates them with unique identifiers.

3. The power module full life cycle carbon-health co-accounting method according to claim 1, characterized in that, The sliding window thermal cycle counting algorithm uses a fixed-length window to cache historical shell temperature data and applies simplified rainflow counting to extract closed thermal cycles. The time complexity is O(1), and the memory usage is less than 50KB. The online thermal resistance correction formula is: R_th(jc)(t) = R_th0 × (1 + α·N_eq^β); Where R_th(jc)(t) is the dynamic junction thermal resistance corrected at time t, R_th0 is the factory thermal resistance baseline value, N_eq is the equivalent thermal cycle number, α and β are thermal resistance aging parameters, and the value of β ranges from 0.85 to 1.

15.

4. The power module full life cycle carbon-health co-accounting method according to claim 3, characterized in that, The thermal resistance aging parameters α and β were calibrated through accelerated aging experiments. The training group was fitted using the nonlinear least squares method, and the average relative error of the validation group was less than 5%. The parameters were calibrated separately for different packaging processes and could be updated remotely in the cloud.

5. The power module full life cycle carbon-health co-accounting method according to claim 1, characterized in that, The electrothermal network model is a Foster-Cauer hybrid model, and the junction temperature is estimated using a sliding mode observer. The observer equation is: x_hat_dot = A·x_hat + B·u + L·sign(y - C·x_hat) Where x_hat is the state vector estimate, A, B, and C are the system matrices, u is the power loss input, y is the measured case temperature, and L is the observer gain matrix determined by pole placement; under standard pulse load test conditions, the junction temperature estimation error margin is less than ±5℃.

6. The power module full life cycle carbon-health co-accounting method according to claim 1, characterized in that, The determination method for the electricity carbon emission factor is as follows: When using a market-based approach, the residual electricity portfolio emission factor or a specific emission factor agreed upon in a power purchase agreement is used, in conjunction with an energy attribute certificate or power purchase agreement; when using a location-based approach, the officially published regional average electricity portfolio factor is used; the calculation formula for the dynamic operating carbon emissions is as follows: C_rate(t) = P_loss(t) × EF_electricity(t) Where P_loss(t) is the real-time power loss at time t, EF_electricity(t) is the carbon emission factor of electricity at time t, and the carbon emission during the operation phase is calculated by integrating this rate over time. For renewable energy direct connection scenarios, the green electricity ratio is obtained in real time through the communication interface of the electricity meter or inverter; for ordinary grid access points, the green electricity ratio is allocated on an hourly basis.

7. The power module full life cycle carbon-health co-accounting method according to claim 1, characterized in that, The formula for calculating the life-cycle carbon footprint as described in S3 is: LCACF = (C_manufacturing + C_use + C_EoL) / (P_rated × T_lifetime) Where P_rated is the module's rated power; T_lifetime is the expected service life; and LCACF is measured in g CO2-eq / kW·h, which complies with the functional unit requirements of ISO 14067. The formula for calculating carbon emissions during the manufacturing stage is as follows: C_manufacturing=Σ(E_process,i×EF_electricity)+Σ(M_material,j×EF_material,j)+ C_overhead; Where E_process,i is the energy consumption of the i-th process, M_material,j is the amount of the j-th material, EF_material,j is the corresponding carbon emission factor, and C_overhead is the allocation of public energy consumption. The carbon emissions at the end of the life cycle stage are calculated using the circular footprint formula: C_EoL = E_recycling - R_2 × V_sub × A_sub Where R_2 is the output recycling rate, E_recycling is the energy consumption for recycling, V_sub is the carbon gain from replacing virgin materials, and A_sub is the replacement ratio; Data quality was assessed using the Pedigree Matrix, with a score of less than 3.

0. Uncertainty was also quantified using Monte Carlo simulation, with expanded uncertainty less than ±15%.

8. The power module full life cycle carbon-health co-accounting method according to claim 1, characterized in that, The stochastic simulation method described in S4 embeds the cyclic footprint formula into the simulation process. The input variables include the probability distribution of remaining lifetime, future value of recycled materials, and future green electricity penetration rate. The output is the probability that the net carbon gain is greater than zero. When the probability exceeds a preset confidence threshold and the remaining lifetime meets the minimum requirements, a tiered utilization suggestion is output. For existing power modules without historical operating data, a prior distribution is constructed based on modules in the same batch that have historical operating data, and updated in real time in conjunction with the currently collected data under a Bayesian framework.

9. The power module full life cycle carbon-health co-accounting method according to claim 1, characterized in that, The digital signature described in S5 is executed by a hardware security module built into the key metering equipment, supporting the national cryptographic algorithms SM2 / SM3 / SM4 and internationally recognized algorithms; the cryptographic hash value of the key data packet is constructed into a Merkle tree, and the root hash value is uploaded to the blockchain for evidence storage; the zero-knowledge proof enables third-party auditors to verify the compliance of the carbon footprint statement without obtaining the original process data.

10. A power module lifecycle carbon-health co-accounting device employing the power module lifecycle carbon-health co-accounting method according to any one of claims 1 to 9, characterized in that, include: The data acquisition and alignment module is used to collect cross-stage data and perform time-series alignment based on digital twin identifiers; The dynamic junction temperature estimation module estimates the junction temperature based on the electrothermal network model and online thermal resistance correction. The dynamic carbon emission accounting module distinguishes between market-based and location-based carbon emission factors to calculate operational carbon emissions. The full life cycle carbon footprint accounting module calculates carbon emissions during the manufacturing, use, and end-of-life stages and normalizes them into product-level carbon footprint values. The retirement decision-making module uses random simulation to calculate the probability of net carbon gain from tiered utilization and outputs decision recommendations. The trusted evidence storage module digitally signs key data packets, stores them on the blockchain, and generates zero-knowledge proofs.

Images