An accurate remaining useful life (RUL) prediction method is significant to optimize the lithium-ion batteries' performances in an intelligent battery management system. Since the construction of battery models and the initialization of algorithms require a large amount of data, it is difficult for conventional methods to guarantee the RUL prediction accuracy when the available data are insufficient. To solve this problem, a synergy of sliding-window grey model (SGM) and particle filter (PF) is exploited to build an innovative framework for battery RUL prediction. The SGM is adopted to explore the modelling of battery capacity degradation, and it characterizes the capacity changes during the battery's life-time with a few data (eg, 8 data points). To promote the accuracy and traceability of prediction, the development coefficient of the SGM, which can dynamically reflect the capacity degradation, is extracted to update the state variables of state transition function in PF. Accordingly, the fusion of SGM and PF (SGM-PF) can extrapolate the changes of the capacity and realize RUL prediction using fewer data. Furthermore, the performances of SGM-PF are comprehensively validated using two types of batteries aged under different conditions. The RUL prediction results reveal that the SGM-PF framework can achieve precise and reliable predictions in different prediction horizons with as few as 8 data points, and it has prominent performance in accuracy and stability over contrastive methods, especially in long-term prognosis.