A capacity optimization of energy storage system based on discrete Fourier transform is presented to inhibit the adverse effects of photovoltaic power generation fluctuations. The method takes into account the characteristics of photovoltaic power generation and charge-discharge response time properties of energy storage system. The spectrum analysis results of photovoltaic power generation fluctuations are obtained through DFT. In different time periods, to compensate the power deviation, different control mechanisms and control capacity are in use. Based on the spectrum analysis results and combined with the economics, a capacity optimization of energy storage system can be determined. Results indicate that the root mean square error (RMSE) of photovoltaic power prediction can be reduced from 2.2994%~10.0832% to 0 by a relatively small capacity of energy storage system determined by the presented method. The method could sharply relieve the impact of photovoltaic power generation on the grid, reduce system reserve capacity requirements, and enlarge the space to accommodate photovoltaic power.