Remaining Useful Life Prediction of Lithium-ion Battery based on Dual Particle Filter

• Main Authors: Hou, Zongxiang, Qu, Qilin, Yang, Tao, Su, Housheng, Zheng, Ying
• Format: Conference Proceeding
• Language: English
• Subjects: Automation; capacity regeneration; Degradation; Fluctuations; Lithium-ion batteries; Manuals; Measurement errors; noise distribution; particle filter; remaining useful life; Training
• ISSN: 2688-0938
• DOI: 10.1109/CAC57257.2022.10055725

The remaining useful life (RUL) prediction of lithium-ion battery is essential for health management, which can guide the time of battery replacement. However, environmental factors, measurement errors and capacity regeneration can introduce noise during battery performance degradation, which can lead to fluctuations in the degradation curve and affect the accuracy of RUL predictions. The distribution of noise is usually given by manual experience and lacks theoretical basis and accuracy. To address the above issues, a new prediction framework including capacity regeneration detection and dual particle filter (DPF) is presented. First, the capacity regeneration point is detected by forward differential and the 3sigma principle. Secondly, the distribution of process noise and observation noise in the non capacity regeneration moments in the training set is extracted by the first particle filter. Thirdly, the second PF uses the previously extracted noise distribution and parameters to establish the state space model of the test set and then predicts RUL. The experimental results show that the capacity regeneration point detection is accurate. In comparison with other approaches, the DPF method has a better prediction precision and significantly reduces the impact of fluctuations such as capacity regeneration on the prediction precision.