Abstract: The thermal safety of power batteries has been widely concerned by academia and industry. The temperature information of battery system is the essential reference to make efficient thermal management strategies. Specific to cylindrical batteries, lightweight lumped-parameter temperature prediction models are developed for embedded applications. Firstly, based on an equivalent circuit model, a mass-point（MP） temperature model is formulated according to basic electrical-thermal effects by regarding the cell as an isotropic homogeneous body. On this basis, the winding and shell parts of the cell are separately studied that surface heat dissipation,heat conduction delay and reversible heat are modeled in detail and related parameters are experimentally identified, thereby deriving an improved layered-divided（LD） temperature model to predict battery kernel（internal） and shell（external） temperatures. Finally,taking an 18650 li-ion battery as the object, experimental results under different temperatures and load conditions show that the temperature prediction performance of the LD model is obviously ameliorated in contrast to the MP model; the LD model can keep high reliability subject to various working conditions. Subject to hybrid pulse excitations, the max temperature error of the LD model is about 1 ℃ after 10 minutes of open-loop simulation.

Key words: lithium-ion battery, internal temperature, external temperature, electrical-thermal coupling, temperature prediction model