Flexible method for estimating the state of health of lithium-ion batteries using partial charging segments

Batteries are crucial components of electric vehicles (EVs), necessitating the accurate estimation of their state of health (SOH). Despite numerous works focusing on SOH estimation, most assume complete battery charging data, which seldom aligns with real-world scenarios where charging rarely initiates from 0% state of charge (SOC). This study introduces a novel battery SOH estimation method tailored for partial charging segments. The proposed methodology involves introducing an Incremental Energy per SOC (IES) curve to analyze battery aging characteristics. This curve is derived by dividing incremental energy by SOC during the charging phase. Key battery health indicators (HIs), namely the maximum and standard deviation of the IES curve, are then extracted from charging data with varying initial SOCs. Subsequently, we present a Bidirectional Long Short-Term Memory with Reduction Mechanism (LSTM-reduction). This model integrates forward and reverse LSTM structures, featuring a reduction gate within the LSTM architecture. This gate filters unnecessary data, preserving valuable information during processing. The bidirectional LSTM-reduction combines both LSTM structures, effectively incorporating historical and future information for improved time series modeling. This comprehensive approach enhances sequence modeling efficiency, thereby elevating reliability. To demonstrate the effectiveness and robustness of the proposed SOH estimation method, we utilize randomly generated partial charging segments from a battery aging dataset with four distinct charging rates. In the experimental phase, the Root Mean Square Error (RMSE) of estimated SOH consistently remains below 0.8%. Moreover, a significant majority of the Coefficient of Determination (R2) values exceed 0.9, across varying initial battery SOC charging processes. These results affirm the feasibility and robustness of the proposed SOH estimation method for partial charging segments with different charging rates.