TY - JOUR
T1 - Battery State-of-Health Estimation: A Step towards Battery Digital Twins
AU - Safavi, Vahid
AU - Bazmohammadi, Najmeh
AU - Vasquez, Juan C.
AU - Guerrero, Josep M.
PY - 2024/2
Y1 - 2024/2
N2 - For a lithium-ion (Li-ion) battery to operate safely and reliably, an accurate state of health (SOH) estimation is crucial. Data-driven models with manual feature extraction are commonly used for battery SOH estimation, requiring extensive expert knowledge to extract features. In this regard, a novel data pre-processing model is proposed in this paper to extract health-related features automatically from battery-discharging data for SOH estimation. In the proposed method, one-dimensional (1D) voltage data are converted to two-dimensional (2D) data, and a new data set is created using a 2D sliding window. Then, features are automatically extracted in the machine learning (ML) training process. Finally, the estimation of the SOH is achieved by forecasting the battery voltage in the subsequent cycle. The performance of the proposed technique is evaluated on the NASA public data set for a Li-ion battery degradation analysis in four different scenarios. The simulation results show a considerable reduction in the RMSE of battery SOH estimation. The proposed method eliminates the need for the manual extraction and evaluation of features, which is an important step toward automating the SOH estimation process and developing battery digital twins.
AB - For a lithium-ion (Li-ion) battery to operate safely and reliably, an accurate state of health (SOH) estimation is crucial. Data-driven models with manual feature extraction are commonly used for battery SOH estimation, requiring extensive expert knowledge to extract features. In this regard, a novel data pre-processing model is proposed in this paper to extract health-related features automatically from battery-discharging data for SOH estimation. In the proposed method, one-dimensional (1D) voltage data are converted to two-dimensional (2D) data, and a new data set is created using a 2D sliding window. Then, features are automatically extracted in the machine learning (ML) training process. Finally, the estimation of the SOH is achieved by forecasting the battery voltage in the subsequent cycle. The performance of the proposed technique is evaluated on the NASA public data set for a Li-ion battery degradation analysis in four different scenarios. The simulation results show a considerable reduction in the RMSE of battery SOH estimation. The proposed method eliminates the need for the manual extraction and evaluation of features, which is an important step toward automating the SOH estimation process and developing battery digital twins.
KW - CNN-LSTM
KW - Data pre-processing
KW - Digital Twin
KW - Discharging characteristics
KW - Lithium-ion batteries (LIBs)
KW - State of the health
UR - http://www.scopus.com/inward/record.url?scp=85184512010&partnerID=8YFLogxK
U2 - 10.3390/electronics13030587
DO - 10.3390/electronics13030587
M3 - Journal article
SN - 2079-9292
VL - 13
JO - Electronics
JF - Electronics
IS - 3
M1 - 587
ER -