Performance Monitoring in Fuel Cell Systems

Fuel cell hybrid energy management and control system with improved accuracy and reliability by filtering sensor data and predicting fuel cell degradation. The system uses a Butterworth filter to remove noise and interference from sensor signals. This cleaned data is then fed into a support vector machine (SVM) to predict trends in fuel cell hybrid energy performance. This allows detecting if the fuel cell is degrading and taking action before major issues arise.

Source

Monitoring the health and performance of small air-cooled hydrogen fuel cells to predict faults and evaluate fuel cell operation. It uses a modified Kalman filter to accurately predict fuel cell monitoring data by considering the changing amplitudes of different types of monitoring data. The filter algorithm obtains a noise matrix from all monitoring data, determines a battery state representation value based on fluctuation differences, and updates the noise matrix for more precise predictions.

Source

Stack health monitoring for fuel cells using a multi-point amplitude phase difference detection method to improve accuracy and real-time monitoring compared to voltage-only monitoring. The method involves capturing analog signals from a fuel cell stack during normal operation, extracting differential signals between historical and current signals, and determining monitoring items based on signal changes. This allows targeted fault detection and stack health assessment using fewer items compared to full monitoring.

Source

Predicting the remaining useful life of fuel cells using dynamic health indicators that consider the complex interactions between operating conditions and factors affecting fuel cell degradation. This improves accuracy and reliability of fuel cell life prediction compared to static models. The dynamic health indicators capture the long-term dependence between operating conditions and fuel cell degradation, accounting for factors like environment and operating mode. This allows more realistic and accurate predictions of fuel cell life under dynamic operating conditions.

Source

Method for efficiently monitoring components of a fuel cell system in a vehicle like a commercial truck. The method involves comparing sensor data from the component to a reference variable to determine an event size that indicates component health. This allows targeted monitoring of costly fuel cell turbomachines. The control unit acquires sensor data, calculates a comparison variable based on that data, compares the sensor data to the comparison variable, and outputs an event size based on the comparison. This provides specific and effective monitoring of fuel cell system components like compressors and expanders.

Source

Fuel cell operation monitoring method and system to detect early warning signs of fuel cell failure or degradation in real-time during operation. The monitoring involves analyzing factors like fuel cell load, hydrogen pressure, oxygen supply, and cell balancing to identify operation risks. It calculates indices like load uniformity, hydrogen pressure risk, and oxygen supply deviation. If thresholds are exceeded, it generates warning signals indicating potential issues. This allows proactive mitigation of problems before they become critical.

Source

Obtaining and updating fuel cell engine polarization performance in real-time by measuring operating current and voltage during load current delivery and using that data to adjust the load current and update control parameters. This allows online determination of fuel cell performance curves without bench testing, which can change over time due to factors like temperature, load, and degradation. The method involves delivering load current, measuring operating current/voltage, plotting on a curve, and iteratively adjusting load current based on the measured data. This allows obtaining accurate, real-time fuel cell performance curves that can be used to optimize control parameters for the specific operating conditions.

Source

Fuel cell online health diagnosis and life prediction method, device, and system using machine learning to monitor fuel cell stack performance and predict stack degradation and failure. The method involves continuously learning stack data to predict future stack health based on time. It uses a neural network with nonlinear layers to learn intermediate features representing the stack's future performance. The network is periodically updated to correct prediction errors. Additionally, a fault diagnosis module combines predicted and real-time data to identify features predicting fault occurrence.

Source

A system and method for monitoring the cold start status of a fuel cell to improve cold start reliability. The system has a monitoring device that tracks key parameters during cold start, like stack temperature, stack pressure, and stack current, and uses them to determine the risk of ice formation. If the stack temperature is below a certain point and the stack overpressure is above a threshold, it indicates a high risk of icing. This information can be used to adjust the fuel cell operating parameters during cold start to mitigate ice formation and improve startup success.

Source

Controlling operation of multiple fuel cell systems in a vehicle to extend the overall system's service life. It involves monitoring the health of each individual fuel cell system during operation and comparing it to the expected health based on historical use conditions. If the actual health deviates significantly, indicating unexpected degradation, the other fuel cell system is used more to compensate. This prevents excessive degradation of the weaker system and extends the overall system's service life since it's based on the weakest fuel cell.

Source

Predicting fuel cell performance degradation in vehicles using real-world driving data rather than lab tests. The method involves retrieving vehicle operating data over a time interval, cleaning the data, extracting the fuel cell operating data, fitting curves to predict performance decay as working time increases, and using the curves to predict future performance decay. This allows continuous, flexible, and low-cost fuel cell performance degradation analysis and prediction based on actual driving conditions.

Source

Remote monitoring system and method for fuel cells that allows monitoring fuel cell performance and detecting issues without on-site access. The system uses sensors to collect data like temperature, voltage, current, and level. A local server acquires the data at adjustable frequencies and durations. It sends the data to a remote server that compares against normal ranges. If values are abnormal, an early warning is issued. By adjusting sensor data collection durations based on rate of change, the system can optimize monitoring efficiency.

Source

Method for assessing the health and predicting the life of proton exchange membrane fuel cells (PEMFCs) to improve their durability. The method involves estimating the degradation trends and remaining useful life of PEMFCs using a model that allows independent evaluation of critical parameters like open circuit voltage, exchange current density, internal resistance, and limit diffusion current. The parameters are estimated from measured voltage and current density data using a noise reduction process. By separately assessing each parameter, the method avoids the limitations of fixed or constrained parameter values and inconsistent attenuation rates.

Source

Digital twin-based fuel cell engine monitoring method and system that provides real-time fuel cell engine monitoring using a digital twin model. The method involves acquiring real-time engine operating data and environmental conditions, and using a trained neural network model to predict fuel cell engine health and faults based on the current operating conditions. This allows monitoring of engine components in real-time rather than relying on historical data. The model predicts engine life, damage levels, and failure probabilities.

Source

Diagnosing and controlling the operating state of a fuel cell system using a detailed electrochemical and physical model of the fuel cell. The model allows continuous forecasting of the fuel cell's electrical power based on the measured current and operating media variables. This provides a real-time indication of cell health and allows optimized media management based on the diagnosed state.

Source

Monitoring fuel cell systems, especially high-temperature fuel cells, to reliably determine degradation. The monitoring involves using machine learning models to predict degradation based on inputs like operating parameters. The models are trained using historical data. The degradation model is separated from a higher-level model that predicts overall system performance. The remaining residue from separating the models is the reference model. It predicts the target parameter when degradation is removed. The models are iteratively improved through alternating steps where the degradation model predicts degradation and the reference model predicts the target parameter. The iterations continue until convergence. This allows reliable degradation monitoring without needing direct degradation measurements.

Source

Assessing the health of fuel cells and fuel cell stacks in a fuel cell system by distinguishing between bad channels and weak cells, accounting for fuel cell aging, and compensating for fuel cell system operation and measurement variability. It involves tracking stack voltage, identifying bad channels with voltage measurement differences, validating data, and assessing stack health based on cell weakness.

Source

Method for predicting the degradation and remaining life of proton exchange membrane fuel cells. The method involves using a combination of temporal convolutional networks (TCN) and random vector functional link (RVFL) neural networks to predict fuel cell degradation indicators. The TCN extracts features from the input data, and the RVFL makes the final predictions. An improved transit search algorithm is used to optimize the TCN-RVFL model. The predicted remaining life is combined with observed life to estimate total life.

Source

Method for proactively recognizing stress situations of a fuel cell to prevent failures. The method involves continuously monitoring operating parameters of the fuel cell during normal operation. It compares the current operating conditions with a database of known stress situations. If a match is found, it indicates an impending failure. This allows identifying and addressing potential issues before they escalate. The method provides real-time failure prevention rather than relying on test stand experiments.

Source

Fuel cell output control system that allows customizing the amount of power generated by a fuel cell stack based on user requirements. The system involves training a neural network using fuel cell monitoring data to derive optimal design information for the unit cells. The number and area of cells in the stack are then set based on this optimized design to achieve the desired output.

Source