Abstract: The current data-driven remaining useful life (RUL) prognostics methods generally have the limitation that cannot be well adapted to the prediction for different batteries, along with low prediction accuracy caused by the redundancy or deficiency of health indictors (HIs).To solve the problem, integrating the PCA-based feature fusion method and NARX neural network, an indirect RUL prognostics framework was proposed for lithium-ion battery. Firstly, multiple measurable parameters that could reflect the performance degradations of lithium-ion battery were selected as candidate HIs, and then the HI was extracted based on PCA to eliminate the redundancy. Furthermore, both health factor and capacity prediction models based on NARX-NN were established using a set of battery life data. Taking the HIs of early stage as the input, the RUL of different batteries with the same type could be predicted indirectly. Finally, sufficient experiments were carried out to validate the high efficiency and adaptability of the proposed method for the same type lithium-ion battery.