The spatial resolving of temperature gradient is a key but challenging issue for the loading performance, aging evaluation, and safety guarantee of large format lithium battery, while the internal temperature cannot be measured directly in field. Notable temperature difference of large format battery emerges in heavy load applications. This paper tries to solve spatial-distributed temperature gradient by proposing a novel method. The multilayer thermal model and the real-time method based on this model are firstly proposed to noninvasively identify the spatial temperature gradient of the battery. And the thermal conduction resistance of the battery along aging are estimated by the extended lumped-parameter model and forgetting factor recursive least square algorithm, while the convection resistance and heat capacity can be estimated by the lumped parameter model and least square algorithm. The relation between the temperature sensor number and total thermal nodes is then derived by the experiment. Finally, the proposed method is systematically verified through the simulation and experiment. Both the spatial distribution and the highest node of the battery temperature are obtained with promising accuracy and robustness. And it can be implemented in battery management systems for various online applications such as electric vehicles and hybrid electric vehicles.