The energy storage system is an important facility for energy conversion and storage in the energy Internet, and lithium batteries are widely used in the construction of current energy storage systems, which occasionally cause accidents and losses due to battery aging or failure. The remaining life prediction can provide a basis for the predictive maintenance of the battery and ensure the safe and stable operation of the energy system. Aiming at the problems of capacity degradation and noise information affecting the performance of the algorithm in the existing lithium battery remaining useful life (RUL) prediction, this paper proposes a lithium battery remaining useful life prediction method based on HHT-ARIMA-PF. Firstly, the feature factors are extracted from the charge-discharge curves of the battery to construct the battery capacity feature data set. The empirical mode decomposition is used to decompose the battery capacity feature data set into IMF components and trend items, and the Hilbert transform (HT) is used to determine the noisy components and the Bessel low-pass digital filter is used to process the noisy components. Then, all the internal modal function IMF and trend terms in the processed battery capacity signal were predicted using ARIMA and PF models, respectively, and the predicted values were obtained, respectively. Finally, the prediction results of the intrinsic mode function based on ARIMA and the prediction results of the trend term based on PF are summed up as the predicted value of the battery capacity, and the predicted value of the remaining lifetime RUL of the lithium battery is calculated. The experimental results show that the method proposed in this paper can realize the differentiation processing of a large amount of data with high prediction speed and prediction accuracy.