As a highly efficient and stable energy storage battery, NaNiCl2 can resist over-discharge to a certain extent, but frequent over-discharge will still lead to irreversible capacity decay. Therefore, accurate estimation of the battery's SOC (State of Charge) is crucial to extending its service life. This study proposes a Physics-Informed Neural Network (PINN) combined with Sparse Identification Nonlinear Dynamics (SINDy). Given that the voltage change rate and current change rate of the battery evolve dynamically over time under a specific discharge mode, SINDy can effectively capture these nonlinear changes and build the kinetic equations of voltage and current on this basis, thereby extracting the voltage change rate and current change rate with physical significance as additional input features of the neural network to expand the feature space. Experimental results on NaNiCl2 battery data show that this method can significantly reduce the SOC estimation error when the feature information is limited.

Physics-Informed Neural Network Combined with Sparse Identification Nonlinear Dynamics for Battery SOC Estimation

Guo Q.;Grimaccia F.;Niccolai A.;
2025-01-01

Abstract

As a highly efficient and stable energy storage battery, NaNiCl2 can resist over-discharge to a certain extent, but frequent over-discharge will still lead to irreversible capacity decay. Therefore, accurate estimation of the battery's SOC (State of Charge) is crucial to extending its service life. This study proposes a Physics-Informed Neural Network (PINN) combined with Sparse Identification Nonlinear Dynamics (SINDy). Given that the voltage change rate and current change rate of the battery evolve dynamically over time under a specific discharge mode, SINDy can effectively capture these nonlinear changes and build the kinetic equations of voltage and current on this basis, thereby extracting the voltage change rate and current change rate with physical significance as additional input features of the neural network to expand the feature space. Experimental results on NaNiCl2 battery data show that this method can significantly reduce the SOC estimation error when the feature information is limited.
2025
2025 IEEE International Conference on Environment and Electrical Engineering
File in questo prodotto:
File Dimensione Formato  
Physics-Informed_Neural_Network_Combined_with_Sparse_Identification_Nonlinear_Dynamics_for_Battery_SOC_Estimation.pdf

Accesso riservato

: Publisher’s version
Dimensione 931.32 kB
Formato Adobe PDF
931.32 kB Adobe PDF   Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/1299213
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? ND
  • OpenAlex 1
social impact