Dual-Scale Nonstationary Representation for Degradation Tracking and Aging-Informed Monitoring of Lithium-Ion Battery System

Battery energy storage systems generally comprise numerous lithium-ion battery (LIB) cells and require reliable monitoring. In practice, cells’ degradation introduces varying data characteristics. This reduces the performance of monitoring models designed for fresh cells, necessitating an adaptive updating scheme. However, existing methods fail to effectively distinguish between online anomalies and degradation-caused data variation. Further, cells’ charging–discharging behaviors also contribute to time-varying data trends. These factors lead to multiple sources for nonstationary operational characteristics, which complicate degradation tracking. To address these challenges, we recognize the following considerations for the first time: 1) LIB cells exhibit repetitive nonstationary trends from multiple cycles of charging–discharging behaviors. Also, a gradual deviation in these trends across cycles reveals degradation-induced nonstationarity. 2) LIB operational signals consist of two distinct components—aging-relevant components, which manifest drifting with degradation, and aging-irrelevant components remaining invariant despite aging. Thereupon, we establish a dual-scale nonstationary representation architecture operating on two scales: within individual charging–discharging cycles and across cycles. It synergistically tracks the nonstationary characteristics stemming from both charging–discharging behavior and degradation. Further, a novel degradation component decomposition algorithm is proposed to separate the LIB operational signals into aging-relevant and irrelevant parts, thereby enabling explicit observation of aging phenomena for online monitoring. On that basis, the monitoring models can be adaptively updated with reliable information for differentiating normal degradation and real anomalies. The effectiveness of the proposed method is proven via the experimental verification in a real-world LIB system rig.