Data-Driven Real-Time Prediction of Pouch Cell Temperature Field Under Minimal Sensing

Abstract

The monitoring of temperature distribution is crucial for advanced battery thermal management. This study proposes a data-driven temperature field prediction method for the pouch cell thermal process, a typical distributed parameter system (DPS). First, empirical spatial basis functions (SBFs) that represent underlying spatial modes of the thermal system are extracted from data snapshots collected offline. Then, we apply the obtained SBFs to the time/space (T/S) separation framework and perform online nonlinear modeling using the partial-node feedback data. On this basis, a dynamics reconstruction strategy is designed for full-node temperature prediction. Experimental studies indicate that the proposed method owns encouraging accuracy and allows minimal sensing configuration. In addition, the error source of the proposed method is systematically analyzed.