Battery prognostics with uncertainty fusion for aerospace applications

This paper presents a hybrid data-driven approach for battery remaining useful life (RUL) estimation for aerospace applications. The prognostic method extracts a series of health indices (HIs) with on-line monitoring parameters to conduct indirect RUL prediction. As a result, in-orbit cycle life estimation for satellite can be achieved. The Relevance Vector Machine (RVM) algorithm is applied, in which an optimized AutoRegressive (AR) model is integrated to improve the long-term predicting performance. Consequently, this method constitutes a probabilistic prognostic framework with uncertainty management capability using a heterogeneous mixture distribution fusion, which provides a more comprehensive criterion for decision makers in scientific maintenance. The actual satellite lithium-ion battery data is used to evaluate and verify the proposed approach, and the experimental results prove its effectiveness.