Temporal pattern attention based Hammerstein model for estimating battery SOC

• Published in: Journal of energy storage Vol. 100; p. 113666
• Main Authors: Hu, Haiyang, Xie, Zengkun, Wang, Dongqing
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 20.10.2024
• Subjects: Hammerstein model; Hierarchical identification principle; N-Adam algorithm; Temporal pattern attention; Hammerstein model; Temporal pattern attention; N-Adam algorithm; Hierarchical identification principle
• ISSN: 2352-152X
• DOI: 10.1016/j.est.2024.113666

A temporal pattern attention based Hammerstein model (TPA-H) is adopted to estimate the state of charge (SOC) of lithium-ion batteries. The proposed Hammerstein system consists of a special temporal pattern attention (TPA) nonlinear block followed by a dynamic controlled auto-regressive moving average (CARMA) linear block. The content includes: 1) By sliding window based TPA, exerting different attentions on input series generates current signal vectors as the input of the linear part; 2) By the hierarchical identification principle, the system is recasted into four fictitious models, each model contains different parameter vectors/matrices; 3) By the N-Adam optimization algorithm and the hierarchical identification principle, the four model parameters are interactively optimized; 4) In the simulation, four different temperatures of 25 °C, 10 °C, 0 °C and − 10 °C under the US06 working condition are considered. The simulation results show that the TPA-H model has good performance under the US06 condition. •TPA exerting different attention on input series generates current signal vectors•By hierarchical identification principle, the system is recasted into four fictitious models•By N-Adam algorithm, four model parameters are interactively optimized•In the simulation, four different temperatures under US06 condition are considered