High wind power penetration in power system leads to a significant challenge in balancing power production and consumption due to the intermittence of wind. Introducing energy storage system in wind energy system can help offset the negative effects, and make the wind power controllable. However, the power spectrum density of wind power outputs shows that the fluctuations of wind energy include various components with different frequencies and amplitudes. This implies that the hybrid energy storage system is more suitable for smoothing out the wind power fluctuations effectively rather than the independent energy storage system. In this paper, we proposed a preliminary scheme for capacity allocation of hybrid energy storage system for power system peak shaving by using spectral analysis method. The unbalance power generated from load dispatch plan and wind power outputs is decomposed into four components, which are outer-day, intra-day, short-term and very short-term components, by using Discrete Fourier Transform (DFT) and spectral decomposition method. The capacity allocation can be quantified according to the information in these components. The simulation results show that the power rating and energy rating of hybrid energy storage system in partial smoothing mode decrease significantly in comparison with those in fully smoothing mode. (C) 2014 Elsevier Ltd. All rights reserved.