Lithium-ion battery capacity estimation — A pruned convolutional neural network approach assisted with transfer learning

Online battery capacity estimation is a critical task for management system to maintain the performance and cycling life in electric vehicles grid energy storage applications. Convolutional Neural Networks, which have shown great potentials estimation, thousands of parameters be optimized demand substantial number aging data training. However, these require massive memory while collecting large volume time-consuming costly real-world To tackle challenges, this paper proposes novel framework incorporating concepts transfer learning network pruning build compact Network models on relatively small dataset with improved performance. First, through technique, model pre-trained transferred targeted improve accuracy. Then contribution-based neuron selection method proposed prune using fast recursive algorithm, reduces size computational complexity maintaining its The capable achieving online at any time, effectiveness verified target collected from four Lithium iron phosphate cells, compared other models. test results confirm that outperforms terms accuracy efficiency, up 68.34% reduction 80.97% computation savings.