Aiming at the capacity optimization configuration method for the hybrid energy storage system （HESS） in microgrid, based on an improved complete ensemble empirical mode decomposition with adaptive noise （ICEEMDAN）, is proposed to address the issue of power fluctuation in the tie-line caused by the variability of renewable energy output and electricity load in grid-connected microgrids. The method decomposes the unbalanced power signals in the microgrid into high-frequency and low-frequency components, achieving power signal reconstruction. Considering the technical characteristics of different energy storage systems, sodium-sulfur batteries are used to suppress low-frequency components, and supercapacitors are used to suppress high-frequency components. Furthermore, by establishing a HESS capacity optimization configuration model aiming at minimizing initial investment and maintenance costs, the mixed energy storage configuration scheme is solved using the commercial solver GUROBI. A case study based on a certain grid-connected microgrid shows that the configured HESS can effectively smooth out the power fluctuation of the microgrid's tie-line, and the method demonstrates good economic performance. The effectiveness and feasibility of the proposed method are thus validated.