A high-precision state of health estimation method based on data augmentation for large-capacity lithium-ion batteries

Lithium-ion batteries' state of health (SOH) is a prominent issue for consumers. However, the complex work condition renders conventional SOH estimation methods ineffective in photovoltaic-storage power stations (PVPS). This paper proposed two health indicators calculation methods and a data augmentation method based on the application law of batterie in PVPS. Firstly, the voltage-cycle frequency and voltage distribution are calculated to identify the voltage plateau period from the continuous operation data. The voltage of similar energy and stair-step voltage were separated from the plateau period. Then the coulombs were calculated as health indicators based on two voltage features. Finally, the pseudo-health indicators were predicted based on the test set and augmentation model. The pseudo-health indicators were added in the test set to restore the past state of the continuity algorithms. Experiments show that the correlation coefficients of two health indicators are greater than 0.87. It confirmed their robust aging characterization capability under the PVPS condition. The accuracy of the six RNNs has been significantly improved under different numbers of pseudo-health indicators. Especially, the optimal result shows that the mean absolute percentage error is 0.204 %, while the root mean square error is 0.265 %. Through multiple validation and comparison, the precision and versatility of this study are confirmed, which provides support for large-capacity lithium‑iron-phosphate (LFP) battery applications in PVPS.