Battery Early End-Of-Life Prediction and Its Uncertainty Assessment with Empirical Mode Decomposition and Particle Filter
Résumé
The first priority of battery predictive maintenance is to estimate its end-of-life (EOL) cycle and assess the uncertainty associated with the predicted values. In this paper, a hybrid method combining empirical mode decomposition (EMD) and particle filter (PF) is applied to an open source database of NASA Ames Prognostics Center of Excellence for the early EOL prediction of four battery cells. The results show a clear decreasing trend of EOL prediction uncertainty when the prediction starts from later operation cycles. However, the distance between the true EOL and the mean predicted EOL has no obvious decrease when more operation data is available. Interestingly, the mean predicted EOL is lower than the true EOL with more available operation data, which is meaningful for reliability engineering and system safety. For instance, the final EOL prediction results from the 80-th cycle are 17 cycles, 7 cycles, 33 cycles and 16 cycles earlier than the real values, respectively.
Mots clés
Particle Filter
Empirical Mode Decomposition
Uncertainty Assessment
Decomposition Filter
Real-valued
Early Prediction
Cycling Performance
Safety Systems
Electrochemical Cell
Obvious Decrease
Open Database
Predictive Maintenance
Remaining Useful Life
Health Status
Time Series
Monte Carlo Simulation
Probability Density Function
Measurement Noise
Duty Cycle
Bayesian Estimation
Posterior Probability Density Function
Battery Degradation
Charge Discharge Cycles
Degradation Trend
Battery Management System
Battery Capacity
Lower Envelope
Bayesian Filtering
Energy Management Strategy
Discharge Cycles