The prediction of discharge capacity of lithium batteries was one of the main tasks of battery management system. The discharge capacity of lithium batteries was related with many parameters, including discharge current, voltage, temperature, and the past charge and discharge history. The prediction methods of existing lithium battery discharge capacity mostly have no learning capabilities and nonlinear prediction ability, in order to predict the discharge capacity of lithium battery more accurately, an algorithm Based on cloud neural network (CNN) was presented. On the basis of the analysis of the actual data of NASA, determine the related influence factors of discharge capacity, set up a corresponding CNN prediction model using cloud model, and use the cloud model for adaptive adjustment of the learning speed. Comparing with the traditional NN method, the simulation result demonstrates that the CNN prediction model has smaller prediction error.