A system and method of predicting variations in a capacity of a battery, the system including an ADFM management device including an experimental data collector to collect at least one first piece of data about the capacity of the battery, an ADF optimizer to optimize a first calculation equation, and a virtual data generator to generate at least one second piece of data; and a server including a training unit to train an artificial intelligence model for outputting the relative capacity variation value by using the at least one first piece of data and the at least one second piece of data as training data, and a prediction unit to obtain a relative capacity variation prediction value, which is output from the artificial intelligence model when the number of charge and discharge cycles and the charge and discharge conditions of the battery are input to the artificial intelligence model.