SOFC system state detection and health diagnosis method and system

Abstract

The invention provides a state detection and health diagnosis method and system of an SOFC system, and relates to the technical field of solid oxide battery state monitoring, and the method comprises the steps: obtaining the operation parameters of an SOFC, and optimizing the data consistency and anti-noise capability based on the fusion algorithm of dynamic Bayesian reasoning and the conditional probability distribution of the operation state of the SOFC; frequency domain features are extracted through fast Fourier transform, dimension reduction is carried out in combination with improved principal component analysis, key variables are reserved, and feature expression is optimized for SOFC high-temperature vibration features; predicting SOH (state of health) and RUL (residual life) of the SOFC based on a random forest, a generative adversarial network and an online updating mechanism; and outputting a diagnosis result and generating a dynamic maintenance scheduling plan based on the RUL prediction, and optimizing the system life by minimizing the downtime and the cost. Through integration of multi-source data fusion, frequency domain feature extraction, AI diagnosis and data enhancement and predictive maintenance, limitations in precision, efficiency and service life optimization in the prior art are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of solid oxide battery state monitoring, and in particular to a state detection and health diagnosis method and system for a SOFC system. Background Art

[0002] Solid oxide fuel cells (SOFCs) have become a key technology for distributed power generation and clean energy conversion since the late 20th century due to their high efficiency, pollution-free operation, and fuel flexibility. However, SOFCs are susceptible to performance degradation during high-temperature operation (600-1000°C) and long-term use, including electrode aging, electrolyte cracking, and fuel gas flow fluctuations. This poses significant challenges to their condition monitoring and health diagnosis.

[0003] In existing technologies, the following problems exist in SOFC status monitoring and health diagnosis: First, data fusion is limited. Existing technologies often use a single data source or simple fusion, failing to fully utilize the dynamic probabilistic characteristics of multi-source data, resulting in insufficient data consistency and noise resistance.

[0004] Second, feature extraction is insufficient. In existing technologies, feature analysis is mostly limited to time-domain statistics, without exploiting frequency-domain information, making it difficult to reflect the complex dynamic characteristics of SOFC high-temperature aging.

[0005] Third, diagnose model defects. In existing related technologies, AI models are highly complex and lack data enhancement methods, which limits their generalization capabilities in scenarios where samples are scarce or have long-term aging.

[0006] Fourth, maintenance optimization is lacking. Existing technologies mostly remain at the state detection level, without integrating lifespan prediction to optimize maintenance scheduling, and thus cannot effectively extend system lifespan or reduce costs.

[0007] The above problems limit the reliability and economy of SOFC in efficient power generation and energy conversion. Therefore, a new solution is urgently needed to at least partially solve the above problems. Summary of the Invention

[0008] In order to at least partially solve the technical problems in the related art, the present invention provides a state detection and health diagnosis method and system for a SOFC system.

[0009] In order to achieve the above object, the technical solution adopted by the present invention includes: According to a first aspect of the present invention, a method for state detection and health diagnosis of a SOFC system is provided, comprising the following steps: Step S1: Acquire SOFC operating parameters, including temperature, voltage, current, fuel flow, exhaust gas composition, and vibration signals, and optimize data consistency and noise immunity based on a fusion algorithm based on dynamic Bayesian inference and the conditional probability distribution of SOFC operating status; Step S2: Extract frequency domain features through fast Fourier transform, combine with improved principal component analysis to reduce dimensionality and retain key variables, and optimize feature expression based on the characteristics of SOFC high-temperature vibration; Step S3: Perform state classification based on random forest, use generative adversarial networks to generate synthetic operating data to enhance the training set, and combine with online update mechanism to improve the model's adaptability to dynamic aging and predict the SOFC health state (SOH) and remaining life (RUL). The calculation formulas for SOH and RUL are: Where, is the current instantaneous output power of the SOFC system, is the initial output power, is the vibration amplitude compensation factor, is the failure threshold, is the attenuation rate, which is used to reflect the performance attenuation rate of SOFC in the early stage of operation. β is the fuel flow deviation factor, which is used to express the effect of flow change on attenuation. is the fuel flow deviation, is the time accumulation factor, which is used to characterize the cumulative effect of operating time on life, and T is the operating time; Step S4: Output the diagnosis results and generate a dynamic maintenance scheduling plan based on the RUL prediction to optimize the system life by minimizing downtime and cost.

[0010] Optionally, Set to 60 to 80, the decay rate is set to 0.015 to 0.035, and β is set to 0 to 0.01. Set to 0 to 0.0005, The value range is 0.8 to 1.2.

[0011] Optionally, step S1 specifically includes: Step S1-1: obtaining the operating parameters of the SOFC; Step S1-2: The dynamic Bayesian inference algorithm updates the conditional probability distribution of the SOFC operating state in real time and combines it with the Gaussian mixture model for probability modeling: Where, is the state variable of the system at time t, is the observed variable of the system at time t, is the posterior probability, is the likelihood function, is the state transition probability, is the posterior probability at the previous moment, For observation The marginal probability of is the expected value, indicating the fusion result.

[0012] Optionally, step S2 specifically includes: Step S2-1: Extract multi-band features of vibration and voltage signals through fast Fourier transform: Where, is the frequency domain coefficient, N is the signal length, is the time domain signal, is the complex exponential factor; Step S2-2: Combine improved principal component analysis to reduce dimensionality and optimize feature sparsity: In the formula, X is the original feature matrix, Z is the dimension-reduced feature, is the Z transposed matrix, W is the projection matrix, is the transposed matrix of W, is the sparse penalty parameter, and I is the identity matrix.

[0013] Optionally, the sparsity penalty parameter is set to 0.1.

[0014] Optionally, in step S3, the random forest includes 150 decision trees, the maximum depth of each decision tree is 12 layers, the generator of the generative adversarial network is 4-layer fully connected, the discriminator is 3-layer fully connected, and the noise dimension is 10 dimensions; The online update mechanism specifically includes: adjusting the tree structure of the random forest and the weights of the generative adversarial network every 500 hours to adapt to the long-term aging of SOFC.

[0015] Optionally, step S4 specifically includes: Step S4-1: Generate SOH and RUL change curves, status classification results and three-dimensional spectrum in real time; Step S4-2: Based on the RUL prediction results, a genetic algorithm is used to optimize the maintenance schedule, with the objective function being to minimize the total cost: Where, and is the weight, For downtime, is the life loss, and the constraint condition is RUL>100 hours.

[0016] Optionally, in step S4-2, The value of is 0.6, The value of is 0.4.

[0017] According to a second aspect of the present invention, there is further provided a state detection and health diagnosis system for a SOFC system, configured to perform the state detection and health diagnosis method for a SOFC system as described in any one of the technical solutions in the first aspect of the present invention. The state detection and health diagnosis system for a SOFC system comprises: Data acquisition module, used to obtain SOFC operating parameters; Data fusion unit, based on dynamic Bayesian inference fusion algorithm, combines the conditional probability distribution of SOFC operating status to optimize data consistency and noise immunity; The feature extraction module uses fast Fourier transform to extract frequency domain features, combines improved principal component analysis to reduce dimensionality and retain key variables, and optimizes feature expression based on the characteristics of SOFC high-temperature vibration; The AI ​​diagnostic module uses random forests for state classification, uses a generative adversarial network to generate synthetic operating data to enhance the training set, and combines an online update mechanism to improve the model's adaptability to dynamic aging, predicting the SOFC state of health (SOH) and remaining life (RUL); The maintenance scheduling unit generates dynamic maintenance scheduling plans based on RUL predictions, optimizing system life by minimizing downtime and costs.

[0018] Beneficial effects: 1. Through the above technical solution, the technical effects that can be achieved by the present invention include: First, efficient data fusion and robustness. Specifically, the method of the present invention fuses multi-source data (temperature, voltage, current, fuel flow, exhaust gas composition, vibration, etc.) through dynamic Bayesian reasoning, with a fusion accuracy of 99% and a data consistency improvement of 15% (the traditional weighted average method is about 85%). Secondly, under the conditions of high-temperature operation of SOFC (600-1000°C) and fuel flow fluctuations (±10L / min), the prior probability is updated in real time and noise is filtered out (for example, when the vibration signal SNR is as low as 10 dB, the error is still less than 1%). Compared with existing related technologies, the dynamic probability modeling of the present invention is more adaptable to non-stationary characteristics and can effectively improve data quality.

[0019] Second, the efficiency and accuracy of frequency domain feature extraction. Specifically, FFT (Fast Fourier Transform) extracts multi-band features of the vibration signal (for example, 5Hz main frequency), and Sparse PCA (Principal Component Analysis) can reduce the feature dimension from 10 dimensions to 3 dimensions, retaining 98% of the effective information, and shortening the extraction time from 50ms to 20ms. The change in the main frequency amplitude directly reflects the aging trend of the electrode, and the accuracy of feature expression is improved by 25% (about 75% for traditional time domain methods). Compared with existing related technologies that rely only on time domain statistics, the present invention can mine frequency domain information and highlight the high-temperature aging characteristics of SOFC. This can improve computing efficiency by 70% and support real-time feature analysis.

[0020] Third, the accuracy and data enhancement capabilities of AI diagnosis. Specifically, the combination of RF (random forest) and GAN (generative adversarial network) achieves a diagnostic accuracy of 97% (about 85% for a single RF), and the training set size can be expanded from 500 groups to 1,500 groups, with a classification error of only 3%. Secondly, the online update mechanism can adjust the model every 500 hours, and the error accumulates over time by 15%. GAN takes about 2 seconds to generate data, and RF training takes less than 5 seconds. Compared with the single neural network method in the existing related technology, the present invention solves the problem of sample scarcity through GAN and improves generalization ability by 50%. Online updates adapt to long-term aging and are superior to static models.

[0021] Fourth, predictive maintenance scheduling and life optimization: Specifically, by optimizing maintenance based on RUL prediction and genetic algorithms, costs can be reduced by 20% (approximately $500 / kWh) and life can be extended by 25% (from 2,400 hours to 3,000 hours).

[0022] 2. Other beneficial effects or advantages of the present invention will be described in detail in the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative labor.

[0024] in: Figure 1 1 is a schematic diagram of a workflow of a method for state detection and health diagnosis of a SOFC system provided by an exemplary embodiment of the present invention; Figure 2 is a schematic diagram of a curve showing a change in SOH over operating time provided by an exemplary embodiment of the present invention. In this figure, the horizontal axis represents operating time (hours) and the vertical axis represents SOH (%), taking into account the influence of vibration; Figure 3 1 is a schematic diagram of a curve showing RUL changing with operating time according to an exemplary embodiment of the present invention. In this figure, the horizontal axis represents operating time (hours) and the vertical axis represents RUL (hours), reflecting the fuel deviation effect. Figure 4 FIG1 is a schematic diagram of a vibration signal spectrum characteristic provided by an exemplary embodiment of the present invention. In this figure, the horizontal axis is frequency (Hz) and the vertical axis is amplitude, reflecting the dynamic characteristics of SOFC operation; Figure 5 1 is a schematic diagram of state distribution probability provided by an exemplary embodiment of the present invention. In this figure, the horizontal axis is the operating time (hours) and the vertical axis is the probability, reflecting the state change trend. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0026] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0027] In order to facilitate relevant technical personnel to have a clearer and more accurate understanding of the technical solutions of the present invention, the technical problems existing in the prior art are described in more detail below with reference to examples.

[0028] Solid oxide fuel cells (SOFCs) have become a key technology for distributed power generation and clean energy conversion since the late 20th century due to their high efficiency, pollution-free operation, and fuel flexibility. However, SOFCs are susceptible to performance degradation during high-temperature operation (600-1000°C) and long-term use, such as electrode aging, electrolyte cracking, and fuel gas flow fluctuations. This poses significant challenges for SOFC condition monitoring and health diagnostics. Early SOFC monitoring relied on physical models and simple electrochemical measurements, but as application demands increase, artificial intelligence and data-driven approaches have become a research focus. The following analysis analyzes the current state of the art and its limitations, based on existing literature and patents.

[0029] The previous paper (Xu, Y., Pang, H., Ran, Zhang, C., Dai, H., Bai, Y., Hao, D., Chen, J., Zhang, B. "Intelligent health states recognition of fuel cell by cell voltage consistency under typical operating parameters.") relied on single voltage data and failed to fully utilize the probabilistic characteristics of multiple source parameters (such as temperature and vibration), resulting in insufficient diagnostic accuracy and robustness under complex operating conditions. This paper uses dynamic Bayesian inference to fuse multi-source data, updating the probability distribution in real time to adapt to non-stationary characteristics and significantly improving noise immunity.

[0030] In terms of life prediction, the literature (Shu, Qin, H., Wu, X., Peng, J., Jiang, C., Xia, Z., Wang, Y., Li, X. "Real-Time State of Health Estimation for Solid Oxide FuelCellsBased on Unscented Kalman Filter") achieves real-time SOFC health state estimation based on unscented Kalman filtering, using filtering technology to smooth data noise. However, its model is highly complex and lacks data generation capabilities, making it difficult to cope with scenarios with insufficient samples.

[0031] The paper (Sheng, C., Zheng, Y., Tian, ​​R., Xiang, Q., Deng, Z., Fu, X., Li, X. "A Comparative Study of the Kalman Filter and the LSTM Network for the Remaining Useful Life Prediction of SOFC") further compared the application of Kalman filtering and LSTM networks in SOFC remaining useful life prediction. The results showed that LSTM has certain advantages in long-term prediction, but the computational resources required are large and it does not combine frequency domain features or maintenance optimization strategies.

[0032] In addition, the paper (Shi, J., Zhang, X., Zhang, H., Wang, Q., Yan, J., Xiao, L. "Automated detection and diagnosis of leak fault considering volatility by graph deep probability learning") proposed a graph deep probability learning method for leak fault detection. Although it is innovative in probability modeling, it is targeted at general energy systems and does not optimize the high-temperature characteristics of SOFC.

[0033] In terms of patent technology, the Chinese patent document with announcement number CN112886038A discloses a fuel cell fault diagnosis method that uses voltage and current data for state classification. However, the data fusion method is limited to simple averaging, does not consider the dynamic dependencies between multi-source data, and has weak noise resistance.

[0034] A Chinese patent document with publication number CN102097636A proposes a fuel cell system fault diagnosis device and method, which determines the system status by voltage and current. This method is simple and easy to implement, but has poor adaptability to non-stationary operating conditions and a high misjudgment rate.

[0035] The Chinese patent document with announcement number CN105355945A proposed a microbial fuel cell fault diagnosis method based on wavelet transform. Although there has been some progress in feature extraction, it has not been optimized for the high-temperature vibration characteristics of SOFC and lacks predictive maintenance capabilities.

[0036] In summary, existing technologies for SOFC status monitoring and health diagnosis have the following problems: First, data fusion is limited. Existing technologies often use a single data source or simple fusion, failing to fully utilize the dynamic probabilistic characteristics of multi-source data, resulting in insufficient data consistency and noise resistance.

[0037] Second, feature extraction is insufficient. In existing technologies, feature analysis is mostly limited to time-domain statistics, without exploiting frequency-domain information, making it difficult to reflect the complex dynamic characteristics of SOFC high-temperature aging.

[0038] Third, diagnose model defects. In existing related technologies, AI models are highly complex and lack data enhancement methods, which limits their generalization capabilities in scenarios where samples are scarce or have long-term aging.

[0039] Fourth, maintenance optimization is lacking. Existing technologies mostly remain at the state detection level, without integrating lifespan prediction to optimize maintenance scheduling, and thus cannot effectively extend system lifespan or reduce costs.

[0040] These challenges limit the reliability and economic viability of SOFCs in efficient power generation and energy conversion, and new technological breakthroughs are urgently needed. This paper proposes a method and system for state detection and health diagnosis of SOFC systems. This method overcomes these shortcomings through multi-source data acquisition and Bayesian fusion, spectrum analysis and principal component feature extraction, AI diagnosis using random forests and generative adversarial networks, and predictive maintenance scheduling optimization. Its significance lies in: First, it fills a technological gap. Compared to related existing technologies, this invention is the first to integrate dynamic Bayesian fusion, spectrum analysis, GAN data enhancement, and predictive maintenance to form a closed-loop monitoring and optimization system.

[0041] Second, it improves performance and economics. It achieves high-accuracy diagnosis (97%), low-complexity feature extraction (70% efficiency improvement), a 25% life extension, and a 20% cost reduction, significantly improving SOFC operating efficiency.

[0042] Third, promote application development. Provide reliable support for the large-scale deployment of SOFC in distributed energy, and promote the commercialization and sustainable development of clean energy technology.

[0043] The core role of the present invention is to solve the key problems in SOFC monitoring through technological innovation, and provide strong technical guarantee for its long-term stable operation and economic benefits in practical applications.

[0044] The technical solution of the present invention is described in detail below with reference to the accompanying drawings.

[0045] According to the first aspect of the present invention, see Figures 1 to 5 The present invention provides a method for state detection and health diagnosis of a SOFC system, comprising the following steps: Step S1: Acquire SOFC operating parameters, including temperature, voltage, current, fuel flow, exhaust gas composition, and vibration signals, and optimize data consistency and noise immunity based on a fusion algorithm based on dynamic Bayesian inference and the conditional probability distribution of SOFC operating status; Step S2: Extract frequency domain features through fast Fourier transform, combine with improved principal component analysis to reduce dimensionality and retain key variables (e.g., intake air flow, operating temperature, operating pressure, cell voltage), and optimize feature expression for the characteristics of SOFC high-temperature vibration; Step S3: Perform state classification based on random forest, use generative adversarial networks to generate synthetic operating data to enhance the training set, and combine with online update mechanism to improve the model's adaptability to dynamic aging and predict the SOFC health state (SOH) and remaining life (RUL). The calculation formulas for SOH and RUL are: Where, is the current instantaneous output power of the SOFC system, is the initial output power, is the vibration amplitude compensation factor, is the failure threshold, is the decay rate, which is used to reflect the performance decay rate of SOFC in the early operation (especially, long-term operation, for example, 3000 hours, It can be adjusted to match the slow decay characteristics according to actual conditions, for example, can be adjusted to 0.0005), β is the fuel flow deviation factor, which is used to represent the effect of flow change on attenuation. is the fuel flow deviation (unit: L / min, positive value indicates flow increase, negative value indicates flow decrease), is the time accumulation factor, which is used to characterize the cumulative effect of operating time on life, and T is the operating time; Step S4: Output the diagnosis results and generate a dynamic maintenance scheduling plan based on the RUL prediction to optimize the system life by minimizing downtime and cost.

[0046] First, it should be noted that, in this disclosure, SOH (State of Health) represents the percentage of the SOFC system's current instantaneous output power relative to its initial power and is a key metric for assessing system performance degradation. Calculated using the power ratio and vibration compensation factor, it reflects the system's operational health and is used to assess the extent of performance degradation. RUL (Remaining Useful Life) represents the estimated remaining operating time (in hours) from the SOFC system's current state until it reaches the failure threshold (70%). It is used to predict system lifespan and guide maintenance. Calculated by subtracting the threshold from the SOH and dividing the result by the decay rate, it is used to predict system lifespan and guide maintenance planning.

[0047] Through the above-mentioned technical solution, the technical effects achieved by the present invention include: First, efficient data fusion and robustness. Specifically, the method of the present invention fuses multi-source data (temperature, voltage, current, fuel flow, exhaust gas composition, vibration, etc.) through dynamic Bayesian inference, achieving a fusion accuracy of 99% and a 15% improvement in data consistency (traditional weighted average method is approximately 85%). Secondly, under the conditions of high-temperature SOFC operation (600-1000°C) and fuel flow fluctuations (±10L / min), the prior probability is updated in real time and noise is filtered out (for example, when the vibration signal SNR is as low as 10 dB, the error is still less than 1%). Compared with existing related technologies, the dynamic probability modeling of the present invention is more adaptable to non-stationary characteristics and can effectively improve data quality.

[0048] Furthermore, in practical applications, the method of the present invention processes 6-dimensional data in less than 50 milliseconds, supporting real-time applications. Furthermore, the method of the present invention provides high-quality input for subsequent feature extraction and diagnosis, ensuring the reliability and stability of monitoring results, and is suitable for SOFC operation under complex operating conditions.

[0049] Second, the efficiency and accuracy of frequency domain feature extraction. Specifically, FFT (Fast Fourier Transform) extracts multi-band features of the vibration signal (for example, 5Hz main frequency), and Sparse PCA (Principal Component Analysis) can reduce the feature dimension from 10 dimensions to 3 dimensions, retaining 98% of the effective information, and shortening the extraction time from 50ms to 20ms. The change in the main frequency amplitude directly reflects the aging trend of the electrode, and the accuracy of feature expression is improved by 25% (about 75% for traditional time domain methods). Compared with existing related technologies that rely only on time domain statistics, the present invention can mine frequency domain information and highlight the high-temperature aging characteristics of SOFC. This can improve computing efficiency by 70% and support real-time feature analysis.

[0050] In addition, in practical applications, the method of the present invention can improve the dynamic characteristic monitoring capability of SOFC, provide more accurate feature support for health diagnosis, and is suitable for long-term operation monitoring.

[0051] Third, the accuracy and data enhancement capabilities of AI diagnosis. Specifically, the combination of RF (random forest) and GAN (generative adversarial network) achieves a diagnostic accuracy of 97% (about 85% for a single RF), and the training set size can be expanded from 500 groups to 1,500 groups, with a classification error of only 3%. Secondly, the online update mechanism can adjust the model every 500 hours, and the error accumulates over time by 15%. GAN takes about 2 seconds to generate data, and RF training takes less than 5 seconds. Compared with the single neural network method in the existing related technology, the present invention solves the problem of sample scarcity through GAN and improves generalization ability by 50%. Online updates adapt to long-term aging and are superior to static models.

[0052] In addition, in practical applications, the method of the present invention maintains high-precision diagnosis under limited sample and aging conditions, and is suitable for SOFC full life cycle management.

[0053] Fourth, predictive maintenance scheduling improves economic efficiency and optimizes service life. Specifically, optimizing maintenance based on RUL prediction and genetic algorithms reduces costs by 20% (approximately $500 / kWh) and extends service life by 25% (from 2,400 hours to 3,000 hours). For example, when the RUL is predicted to be 800 hours, maintenance is scheduled at 700 hours, resulting in four hours of downtime and an adjustment of fuel flow to 85%. Compared to scheduled maintenance in existing technologies, this invention enables dynamic scheduling to reduce unnecessary downtime and improve operational efficiency by 10%. Furthermore, the genetic algorithm achieves a 95% convergence probability after 100 iterations, with computational time of 30 seconds.

[0054] Furthermore, in practical applications, the method of the present invention can reduce maintenance costs, improve the economic efficiency of SOFC, and promote its commercial application in distributed power generation.

[0055] In general, the technical effects of the present invention include: 1. Technological breakthrough: Through the integration of multi-source data fusion, frequency domain feature extraction, AI diagnosis and data enhancement, and predictive maintenance, the limitations of existing technologies in accuracy, efficiency, and life optimization are overcome.

[0056] 2. Quantitative results: fusion accuracy of 99%, feature efficiency increased by 70%, diagnostic accuracy of 97%, lifespan extended by 25%, and cost reduced by 20%.

[0057] 3. Application prospects: Suitable for long-term operation of SOFC in efficient power generation and energy conversion, improving system reliability and economy, and promoting the development of clean energy technology.

[0058] The present invention is described below with reference to an exemplary embodiment. DETAILED DESCRIPTION

[0059] 1. Hardware configuration: 1) Temperature sensor: thermocouple (0-1000℃, ±0.5℃), sampling rate 5Hz, with high temperature shield.

[0060] 2) Voltage sensor: high-precision amplifier (0-10V, ±0.01V), anti-electromagnetic interference.

[0061] 3) Current sensor: Hall sensor (0-200A, ±0.1A), response time <1ms.

[0062] 4) Flow sensor: mass flow meter (0-100L / min, ±1%), corrosion-resistant.

[0063] 5) Exhaust gas analyzer: infrared sensor (CO, CO2 concentration, ±0.5%), real-time analysis.

[0064] 6) Vibration sensor: accelerometer (0-50m / s², ±0.1m / s²), frequency response 10Hz-1kHz.

[0065] 7) Data acquisition system: 32-bit processor, 16-bit ADC, sampling rate 10Hz, storage 2TB.

[0066] 2. Data fusion and preprocessing: 1) Dynamic Bayesian reasoning is used to update the probability distribution in combination with the SOFC operating status (such as fuel utilization).

[0067] 2) Mathematical formula: Dynamic Bayesian Update: Where, is the posterior probability, is the likelihood function, is the state transition probability, is the posterior probability at the previous moment.

[0068] Fusion value: is the expected value, indicating the fusion result.

[0069] Assume a Gaussian distribution prior (e.g., mean temperature 45°C, variance 1), update the posterior based on observations such as voltage and flow, and calculate the expected value as the fusion result.

[0070] 3. Feature extraction and spectrum analysis: 1) The main frequency and subharmonic amplitude of the vibration and voltage signals are extracted through FFT, and Sparse PCA is used to reduce the dimension and optimize the feature sparsity.

[0071] 2) Mathematical formula: Fourier transform: in, is the frequency domain coefficient, N is the signal length, is the time domain signal, is a complex exponential factor, also known as a rotation factor.

[0072] Sparse PCA optimization: Among them, X is the original feature matrix, Z is the dimension-reduced feature, is the Z transposed matrix, W is the projection matrix, is the transposed matrix of W, is the sparse penalty parameter, and I is the identity matrix.

[0073] The test is based on 5,000 hours of operating data. The initial features include 10 frequency band parameters. FFT is performed on the 5,000-point vibration signal. The vibration amplitude compensation factor is calculated based on the average of the main frequency amplitudes extracted by FFT. The first five main frequency amplitudes are extracted, and the feature dimension is reduced from the initial 10 dimensions to 3 dimensions, retaining 98% of the variance.

[0074] 4. AI model training and optimization: 1) Model structure: RF (150 trees, maximum depth 12), GAN (generator 4 layers fully connected, discriminator 3 layers, noise dimension 10).

[0075] 2) Training parameters: 600 sets of data (6000 hours), RF training time <5s, GAN iteration 1000 times, learning rate 0.0005.

[0076] 3) Online update: Update RF weights with new data every 500 hours, and GAN generates 100 sets of synthetic aging data.

[0077] 4) Technical details: GAN generates voltage decay curves, and RF classification status is “normal”, “warning”, and “fault”.

[0078] 5. Result display and maintenance scheduling: 1) Output: The system generates real-time SOH and RUL curves, status classification results, and 3D spectrograms to facilitate user monitoring of SOFC operating status. Output data includes state of health percentage (SOH), remaining life (RUL, in hours), and a visual display of spectral characteristics.

[0079] 2) Scheduling optimization: Based on the RUL prediction results, a genetic algorithm is used to optimize maintenance scheduling. The objective function is to minimize the total cost: in, and is the weight, and , , is the downtime (unit: hours), is the life loss (unit: hours), and the constraint is RUL>100 hours. The optimization process finds the best maintenance time point within the 500-hour operation window.

[0080] RUL calculation example: 1) Background: When a 1kW SOFC runs for 100 hours, the system measures a SOH of 90%. The initial fuel flow rate is 100 L / min, the current flow rate is 105 L / min, and the fuel flow deviation ΔF = 5 L / min.

[0081] 2) Formula application: According to the RUL prediction formula: 3) Substitute parameters: SOH=90%, =70%, =0.025, β=0.01, =5, =0.0005, T=100 hours.

[0082] 4) Calculation results: RUL=800 hours, 5) Parameter description: attenuation rate =0.025, fuel flow deviation factor β=0.01, time accumulation factor =0.0005 is obtained by fitting a 1000-hour SOFC operation experiment and is applicable to the early to mid-stage operation of the system. The running time T is calculated based on the difference between the real-time flow rate and the initial flow rate and accumulated from the system startup.

[0083] 6) Verification: The calculated result RUL=800 hours, which is consistent with the system prediction value, with an error range of ±40 hours, verifying the accuracy of the formula.

[0084] Scheduling case: Scenario: When the RUL is predicted to be 800 hours, system analysis shows the current SOH is 90%, the operating time is 100 hours, and the fuel flow rate deviation is 5 L / min. The genetic algorithm optimization results recommend maintenance at 700 hours, limiting the downtime to 4 hours, and adjusting the fuel flow rate to 85 L / min (i.e., the deviation ΔF changes from 5 L / min to -15 L / min).

[0085] 2) Adjustment effect: After the flow rate is adjusted, the attenuation rate slows down, and the updated RUL calculation is as follows: Note: The negative value here indicates that the decay rate may be reversed to positive repair, and the actual value should be adjusted according to the experiment Assuming adjusted Reduced to 0.015, then: Taking into account the actual situation, the RUL is extended to about 850 hours after flow adjustment (based on experimental verification), and the total cost is reduced by about 20%.

[0086] The genetic algorithm iterated 100 times with a population size of 50, taking approximately 30 seconds to compute, and the probability of the optimization result converging to the global optimal solution reached 95%. After each scheduling optimization, the system updated the SOH and RUL predictions to ensure dynamic adaptability in long-term operation.

[0087] Test results: 1kW SOFC, SOH = 90% after 100 hours, fuel flow rate deviation 5L / min, RUL calculated to be 800 hours, consistent with the prediction.

[0088] Thus, through the above exemplary embodiments, the technical effects of the present invention include: First, efficient data fusion and robustness: 1) Using a dynamic Bayesian inference algorithm, this invention achieves probabilistic fusion of multi-source data (temperature, voltage, current, fuel flow, etc.), significantly improving the robustness and reliability of data processing. Under the complex conditions of SOFC high-temperature operation (600-1000°C) and fuel flow fluctuations (±10 L / min), the system can update prior probabilities in real time and filter out sensor noise. For example, even when the vibration signal's SNR (signal-to-noise ratio) drops to 10 dB, the fused data error remains within 1%.

[0089] 2) Quantitative advantage: Fusion accuracy reaches 99%, and data consistency is improved by 15% (compared to 85% of traditional weighted average), providing high-quality input for subsequent feature extraction and diagnosis.

[0090] Second, the efficiency and accuracy of frequency domain feature extraction: 1) Using FFT to extract multi-band features from vibration signals and optimizing feature selection through Sparse PCA, the system efficiently captures the dynamic characteristics of SOFC operation. During 5,000 hours of testing, the feature dimension was reduced from 10 to 3, retaining 98% of the effective information. Feature extraction time was shortened from approximately 50ms using traditional methods to 20ms. Changes in the amplitude of the dominant frequency feature (e.g., 5Hz) directly reflect electrode aging trends.

[0091] 2) Quantitative advantage. Feature extraction efficiency is increased by 70%, and feature expression accuracy is improved by 25% (compared to 75% of the time-domain statistical method), significantly enhancing the ability to monitor the SOFC aging process.

[0092] Third, the accuracy and data enhancement capabilities of AI diagnosis: 1) Combining the efficient classification of RF with the data generation capabilities of GAN, the system maintains high diagnostic accuracy even with limited sample sizes. Synthetic data generated by GAN (such as aging trend curves) expands the training set size from 500 to 1500 groups. RF rapidly classifies states using 150 trees, achieving a diagnosis time of less than 5 seconds. An online update mechanism adjusts the model every 500 hours to ensure long-term stability. For example, after 3000 hours of operation, the state classification error was only 3%.

[0093] 2) Quantitative advantage: The diagnostic accuracy reached 97%, an improvement of 12 percentage points over the single RF model (approximately 85%).

[0094] Fourth, the economic efficiency and life optimization of predictive maintenance scheduling: 1) Based on RUL prediction and genetic algorithms, this invention optimizes maintenance scheduling, reducing unnecessary downtime and costs. In a 1kW SOFC test, when the RUL was predicted to be 800 hours, the system scheduled maintenance at 700 hours, limiting downtime to 4 hours and adjusting the fuel flow rate to 85%. This prevented performance losses caused by premature or delayed maintenance. Test results showed that the optimized system life was extended from 2400 hours to 3000 hours.

[0095] 2) Quantified Advantages: Maintenance costs are reduced by 20% (saving approximately $500 / kWh), lifespan is extended by 25% (approximately 600 hours), and operating efficiency is improved by 10% compared to traditional scheduled maintenance (lifespan approximately 2,400 hours).

[0096] In one embodiment of the present invention, the present invention Can be set from 60 to 80, the decay rate can be set from 0.015 to 0.035, and β can be set from 0 to 0.01. Can be set from 0 to 0.0005, The value range of is 0.8 to 1.2. The above value ranges are for reference only and can be adjusted according to the actual characteristics of the SOFC system.

[0097] In this embodiment, The calculation based on the rate of change of the main frequency amplitude in the spectrum analysis is positively correlated with the degree of electrode aging, and the range of 0.8-1.2 is determined by the 5000-hour operation experiment.

[0098] In one embodiment of the present invention, step S1 of the present invention may specifically include: Step S1-1: obtaining the operating parameters of the SOFC; Step S1-2: The dynamic Bayesian inference algorithm updates the conditional probability distribution of the SOFC operating state in real time and combines it with the Gaussian mixture model for probability modeling: Where, is the state variable of the system at time t, is the observed variable of the system at time t, is the posterior probability, is the likelihood function, is the state transition probability, is the posterior probability at the previous moment, For observation The marginal probability of is the expected value, indicating the fusion result.

[0099] In one embodiment of the present invention, step S2 of the present invention may specifically include: Step S2-1: Extract multi-band features of vibration and voltage signals through fast Fourier transform: Where, is the frequency domain coefficient, N is the signal length, is the time domain signal, is a complex exponential factor, also known as a rotation factor; Step S2-2: Combine improved principal component analysis to reduce dimensionality and optimize feature sparsity: In the formula, X is the original feature matrix, Z is the dimension-reduced feature, is the Z transposed matrix, W is the projection matrix, is the transposed matrix of W, is the sparse penalty parameter, which can be set to 0.1, and I is the identity matrix.

[0100] According to a second aspect of the present invention, a state detection and health diagnosis system for a SOFC system is provided, which is configured to perform the state detection and health diagnosis method for a SOFC system according to any one of the technical solutions in the first aspect of the present invention. The state detection and health diagnosis system for a SOFC system comprises: Data acquisition module, used to obtain SOFC operating parameters; Data fusion unit, based on dynamic Bayesian inference fusion algorithm, combines the conditional probability distribution of SOFC operating status to optimize data consistency and noise resistance The feature extraction module uses fast Fourier transform to extract frequency domain features, combines improved principal component analysis to reduce dimensionality and retain key variables, and optimizes feature expression based on the characteristics of SOFC high-temperature vibration; The AI ​​diagnostic module uses random forests for state classification, uses a generative adversarial network to generate synthetic operating data to enhance the training set, and combines an online update mechanism to improve the model's adaptability to dynamic aging, predicting the SOFC state of health (SOH) and remaining life (RUL); The maintenance scheduling unit generates dynamic maintenance scheduling plans based on RUL predictions, optimizing system life by minimizing downtime and costs.

[0101] The above are only specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for state detection and health diagnosis of a SOFC system, characterized in that: The steps include: Step S1: Acquire SOFC operating parameters, including temperature, voltage, current, fuel flow, exhaust gas composition, and vibration signals, and optimize data consistency and noise immunity based on a fusion algorithm based on dynamic Bayesian inference and the conditional probability distribution of SOFC operating status; Step S2: Extract frequency domain features through fast Fourier transform, combine with improved principal component analysis to reduce dimensionality and retain key variables, and optimize feature expression based on the characteristics of SOFC high-temperature vibration; Step S3: Perform state classification based on random forest, use generative adversarial networks to generate synthetic operating data to enhance the training set, and combine with online update mechanism to improve the model's adaptability to dynamic aging and predict the SOFC health state (SOH) and remaining life (RUL). The calculation formulas for SOH and RUL are: Where, is the current instantaneous output power of the SOFC system, is the initial output power, is the vibration amplitude compensation factor, is the failure threshold, is the attenuation rate, which is used to reflect the performance attenuation rate of SOFC in the early stage of operation. β is the fuel flow deviation factor, which is used to express the effect of flow change on attenuation. is the fuel flow deviation, is the time accumulation factor, which is used to characterize the cumulative effect of operating time on life, and T is the operating time; Step S4: Output the diagnosis results and generate a dynamic maintenance scheduling plan based on the RUL prediction to optimize the system life by minimizing downtime and cost.

2. The method for state detection and health diagnosis of a SOFC system according to claim 1, characterized in that: Set to 60 to 80, the decay rate is set to 0.015 to 0.035, and β is set to 0 to 0.

01. Set to 0 to 0.0005, The value range is 0.8 to 1.

2.

3. The method for state detection and health diagnosis of a SOFC system according to claim 1, wherein: The step S1 specifically includes: Step S1-1: obtaining the operating parameters of the SOFC; Step S1-2: The dynamic Bayesian inference algorithm updates the conditional probability distribution of the SOFC operating state in real time and combines it with the Gaussian mixture model for probability modeling: Where, is the state variable of the system at time t, is the observed variable of the system at time t, is the posterior probability, is the likelihood function, is the state transition probability, is the posterior probability at the previous moment, For observation The marginal probability of is the expected value, indicating the fusion result.

4. The method for state detection and health diagnosis of a SOFC system according to claim 1, wherein: The step S2 specifically includes: Step S2-1: Extract multi-band features of vibration and voltage signals through fast Fourier transform: Where, is the frequency domain coefficient, N is the signal length, is the time domain signal, is the complex exponential factor; Step S2-2: Combine improved principal component analysis to reduce dimensionality and optimize feature sparsity: In the formula, X is the original feature matrix, Z is the dimension-reduced feature, is the Z transposed matrix, W is the projection matrix, is the transposed matrix of W, is the sparse penalty parameter, and I is the identity matrix.

5. The method for state detection and health diagnosis of a SOFC system according to claim 4, characterized in that: The sparsity penalty parameter is set to 0.

1.

6. The method for state detection and health diagnosis of a SOFC system according to claim 1, wherein: In step S3, the random forest includes 150 decision trees, the maximum depth of each decision tree is 12 layers, the generator of the generative adversarial network is 4-layer fully connected, the discriminator is 3-layer fully connected, and the noise dimension is 10 dimensions; The online update mechanism specifically includes: adjusting the tree structure of the random forest and the weights of the generative adversarial network every 500 hours to adapt to the long-term aging of SOFC.

7. The method for state detection and health diagnosis of a SOFC system according to claim 1, wherein: The step S4 specifically includes: Step S4-1: Generate SOH and RUL change curves, status classification results and three-dimensional spectrum in real time; Step S4-2: Based on the RUL prediction results, a genetic algorithm is used to optimize the maintenance schedule, with the objective function being to minimize the total cost: Where, and is the weight, For downtime, is the life loss, and the constraint condition is RUL>100 hours.

8. The method for state detection and health diagnosis of a SOFC system according to claim 7, wherein: In the step S4-2, The value of is 0.6, The value of is 0.

4.

9. A state detection and health diagnosis system for a SOFC system, characterized in that: For executing the method for detecting and diagnosing the state of the SOFC system according to any one of claims 1 to 8, the state detection and health diagnosis system of the SOFC system comprises: Data acquisition module, used to obtain SOFC operating parameters; Data fusion unit, based on dynamic Bayesian inference fusion algorithm, combines the conditional probability distribution of SOFC operating status to optimize data consistency and noise immunity; The feature extraction module uses fast Fourier transform to extract frequency domain features, combines improved principal component analysis to reduce dimensionality and retain key variables, and optimizes feature expression based on the characteristics of SOFC high-temperature vibration; The AI ​​diagnostic module uses random forests for state classification, uses a generative adversarial network to generate synthetic operating data to enhance the training set, and combines an online update mechanism to improve the model's adaptability to dynamic aging, predicting the SOFC state of health (SOH) and remaining life (RUL); The maintenance scheduling unit generates dynamic maintenance scheduling plans based on RUL predictions, optimizing system life by minimizing downtime and costs.

Images