Accurately predicting the remaining useful life (RUL) of lithium-ion batteries is of significance for improving the safety of working environment and the reliability of equipment. To improve the stability and accuracy of RUL prediction, a battery RUL prediction method based on the combination of denoising technology and hybrid data-driven model is proposed. First, the original data is decomposed by variational mode decomposition, and the noise components are filtered by the analysis of correlation. The residual error is combined with the components which have a strong correlation to complete the sequence reconstruction process. Second, with the combination of Tent chaotic mapping, sine cosine algorithm and Levy flight strategy, the sparrow search algorithm (SSA) is optimized, and the optimal weight threshold of extreme learning machine (ELM) is obtained. Finally, the improved SSA-ELM model is trained by using the smoothed denoised data, and the RUL prediction is completed. The NASA data sets are used to verify the effectiveness of the proposed method. Experimental results show that the average absolute error and root mean square error of the prediction result obtained using this method are controlled within 1.58% and 2.14%, respectively, indicating that this method has a high robustness and a high prediction accuracy. Therefore, the proposed method can be applied to battery RUL prediction.