Based on the method of integrated neural networks and particle filter, a new method is proposed for predicting the residual useful life based on the particle filter with unknown measurement noise under the condition of partially observable information. Firstly, a status-observation data set is generated based on the equipment performance degradation data, and multiple data sets are constructed by using bootstrap technique. The integrated neural network is used to train the status-observation data sets, and the optimal range of measurement noise standard deviation is obtained through derivation. Then, by embedding the measurement noise standard deviation into the framework of particle filter lifetime prediction as the unknown parameter, the residual life prediction and probability density distribution of the nonlinear system are realized. Finally, the validity and feasibility of the proposed method is verified by the simulation to the life of the lithium ion battery.