Satellite Lithium-ion Battery Remaining Useful Life Estimation by Coyote Optimization Algorithm

The estimation of batteries remaining useful life (RUL) is a critical task of the prognostic and health monitoring of satellites. RUL works as an effective decision-making tool for operators by quantifying how much time remains until it loses its functionality. As the capacity is an important indicator for estimating RUL, this paper proposes a novel optimized regression approach for predicting the capacity based on the Coyote Optimization Algorithm (COA) with Support Vector Regression (SVR) called COA-SVR for improving the prediction accuracy of the battery capacity. COA is used for finding the optimal parameters of SVR as the parameter selection has a critical impact on the predictive accuracy of SVR. The performance of COA-SVR has been experimented using NASA's Lithiumion batteries dataset, the experimental results with different evaluation measures showed that the high efficiency of prediction with good stability and low time complexity have been achieved with the COA-SVR. In addition, the prediction accuracy of COA-SVR is compared with those of the basic SVR algorithm with randomized parameter selection and a relevance vector machine (RVM) that has been recently applied in some related work, the comparative results demonstrate that the highest prediction accuracy has been achieved with the proposed model COA-SVR.