For achieving decarbonization in internal combustion engines, a utilization of carbon neutral fuels from renewable energies (e-fuel) could be one option. E-fuel is expected to be implemented as blended fuels with conventional fuels, which results in more uncertainties of the fuel properties. To cope with a larger number of blended fuels compared to the existing alcohol-fuel blends, fuel aging effects resulting from longer refueling intervals in hybrid engines, and fuel blend variations in each refueling station, an optimized engine control and energy management depending on the fuel blend contents will be required. In this study, a new fuel contents estimation method for the engine control based on the external refueling information and the signals of the existing engine sensors is developed, by utilizing data assimilation. With the non-linear ensemble Kalman filter, the prediction model which predicts changes of the fuel blending rate in the tank considering the refueling and fuel aging and their uncertainties was created. Also, the compensation filter which modifies the predicted value based on the engine component signals including measurement noises during engine operations was developed. For the fuel blend prediction model, the relative increase of the heavy components of gasoline, the hydrolysis of dimethyl carbonate (DMC) as one of e-fuels, and the fuel evaporation were considered. The compensation filter is composed of the fuel observation matrix which describes the relationship between the fuel blend rate and the fuel properties, and the engine observation matrix which modelled the signals of the engine components to those fuel properties. The numerical experiment with the repetition of operation and refueling was conducted, considering the 4 fuel blends (Gasoline, Ethanol, Methanol, and DMC) and the fuel aging. The accuracy improvement of the fuel blend estimation with data assimilation was confirmed.