Precisely battery state of health estimation and remaining useful lifetime prediction are crucial factors in ensuring the reliability and safety for system operation. This paper thus focuses on the short-term battery state of health estimation and long-term battery remaining useful lifetime prediction. A novel hybrid method by fusion of partial incremental capacity and Gaussian process regression is proposed and dual Gaussian process regression models are employed to forecast battery health conditions. First, the initial incremental capacity curves are filtered by using the advanced signal process technology. Second, the important health feature variables are extracted from partial incremental capacity curves using correlation analysis method. Third, the Gaussian process regression is applied to model the short-term battery SOH estimation using the feature variables. Forth, an autoregressive long-term battery remaining useful lifetime model is established using the results of battery SOH values and previous output. The predictive capability and effectiveness of two models are demonstrated by four battery datasets under different cycling test conditions. Otherwise, the robustness of the two models is verified using four datasets with different health levels. The experimental results show that the proposed method can provide accurate battery state of health estimation and remaining useful lifetime. •Dual GPR-based models are proposed to establish battery degradation models.•The health indexes are extracted from partial IC curves as model input features.•Correlational coefficient analysis method is applied to extract feature variables.•An autoregressive RUL model is developed using the capacity vs. cycle number.•Four batteries with different initial health levels are used to verify robustness.