Tensor Network-Based MIMO Volterra Model for Lithium-Ion Batteries

• Published in: IEEE transactions on control systems technology Vol. 31; no. 4; pp. 1 - 14
• Main Authors: Hu, Yangsheng, de Callafon, Raymond A., Tian, Ning, Fang, Huazhen
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.07.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Batteries; Capacitors; Computational modeling; Data models; equivalent circuit model (ECM); Heuristic algorithms; Integrated circuit modeling; Kernel; Lithium-ion batteries; Mathematical analysis; Nonlinearity; Predictive models; Rechargeable batteries; system identification; tensor network (TN); Tensors; Volterra model
• ISSN: 1063-6536; 1558-0865
• DOI: 10.1109/TCST.2022.3232894

Accurate battery modeling is fundamental for the battery management system to function well and extract the full potential from a battery without violating constraints. In this article, a tensor network (TN)-based Volterra double-capacitor (VDC) model for lithium-ion batteries is developed to improve the prediction performance of the nonlinear double-capacitor (NDC) model. It is shown that the VDC model maintains the advantages of the NDC model to account for the rate capacity effect and the voltage recovery effect. In addition, the VDC model is capable of predicting both static and dynamic nonlinearities simultaneously in a more accurate way. To estimate the TN-cores in the VDC model, a Bond Core Sweeping Algorithm is proposed and shown to lead to a low-rank representation. A comparison based on experimental data demonstrates that the VDC model gives greater prediction accuracy than the NDC model and Thevenin model, showing significant promise to enhance future battery applications.