Remaining useful life (RUL) prediction of lithium battery is very important for the safe using of lithium batteries. Due to the capacity regeneration phenomenon and random interferences in the using process of lithium batteries, the prediction accuracy and generalization performance of a single model with a single-scale signal are relatively poor. Aiming at these problems, a new RUL prediction method based on variational mode decomposition (VMD) and integrated deep model is proposed. Firstly, VMD is used to decompose the lithium battery capacity data to obtain the global degradation trend of the signal and the local random fluctuation components in multiple scales. Then, the global degradation trend and various fluctuation components are modeled using multilayer perceptron (MLP) and long short-term memory (LSTM) neural network, respectively. Finally, the prediction results of the sub-models of various components are integrated to obtain the final remaining useful life prediction result of the lithium battery. Experiment results show that the proposed method possesses high prediction accuracy and stability.