State of health diagnostic and remain useful life prognostic for lithium-ion battery by combining multi-kernel in Gaussian process regression

• Published in: 2021 IEEE XXVIII International Conference on Electronics, Electrical Engineering and Computing (INTERCON) pp. 1 - 4
• Main Authors: Garay, Fernando, Huaman, William, Vargas-Machuca, Juan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 05.08.2021
• Subjects: Degradation; Gaussian process regression; Gaussian processes; Lithium-ion batteries; multi-kernel; Prediction algorithms; Reliability; remaining useful life; Safety; state of health; Training data
• DOI: 10.1109/INTERCON52678.2021.9532733

The state of health and the remain useful life are key necessary indicators to optimize the operation and reliability of lithium-ion batteries, thus guaranteeing safety and minimizing maintenance costs. However, the complex physicochemical characteristics of degradation maximize these risks. An analysis of the curves of its main degradation characteristics and a Gaussian Process Regression model based on the combination of multi-kernel are proposed in this paper. The data used for the research comes from the open-source NASA Randomized Battery Usage Dataset. The proposed algorithm obtained results in the diagnostic of the state of health of 1.13% and 98.34% in the quantitative tests of mean square error and precision, respectively. In the prognostic of the remaining useful life, it was possible to predict 64 cycles with a margin of error with the real of 1.54%.