The storage battery in rail transit uninterruptible power supply system directly affects the safety and reliability of the load,but its capacity is often configured far beyond normal use,leading to waste of resources. The main reasons for this problem are the overestimation of load and using constant power discharge model in the setting stage. Besides,there is a lack of research on the relationship between load type and backup power supply time. Therefore,a new method for battery capacity reduction is proposed in this paper. In order to get the maximum operating load more accurately,the loads are clustered first. Secondly,different prediction methods are used for different types of loads. The two parameter Weibull simulated load curve is adopted to predict the maximum load for fluctuating loads. The load coefficient method is directly used for the steady load with weak time series correlation. Then,the step load method is used for further reduction of battery capacity,considering the different backup time requirements of different loads. Finally,the battery capacity is reduced based on the actual data of the uninterruptible power supply system of Suzhou rail transit line 3,verifying the accuracy and effectiveness of this method.