Surface Temperature Field Reconstruction of Lithium-Ion Batteries Toward Lumped Thermal Model-Based KF-MLP Estimation Algorithm

High-capacity and large-sized batteries are widely employed in electric vehicles and energy storage systems. The surface temperature field of these batteries is usually maldistributed and unmeasurable in practice, which brings great challenges to temperature safety monitoring. Thus, this article rebuilds the lumped thermal model and proposes a Kalman filter (KF)-multilayer perception (MLP) joint estimation algorithm to reconstruct the 2-D ST field of lithium-ion batteries (LIBs). First, an improved lumped thermal model is devised to accurately obtain multipoint temperatures with only one sensor. Then, a KF-MLP neural network is proposed to diminish the utilization of computational resources and enhance the model generalization capability. Finally, a 2-D temperature acquisition method is designed to obtain reliable experiential data. Experiments are designed to demonstrate the effectiveness of the proposed lumped thermal model-based KF-MLP estimation algorithm. Under 5 $^{\circ }$C discharge conditions for square LIBs, the maximum average error between estimated and actual temperatures on the battery surface is 0.0761 $^{\circ }$C.