Organic photovoltaics (OPVs) have been receiving growing interest, because they can provide large area and low cost solar cells for next generation. A power conversion efficiency (PCE) of more than 10 % under AM 1.5 illumination has been reported for the OPVs, although higher level of PCE is required for large-scale commercialization. Since the maximum thickness of light absorbing layer (around 100 nm) is determined by the low carrier mobility of organic semiconductors, it is important to develop antireflection technologies that can enhance light trapping in the thin absorbing layer. A popular method for analyzing light propagation within solar cells is the finite-difference time-domain (FDTD) simulation. In the FDTD method, the electric and magnetic field are spatially discretized on the grid and the temporal evolution of the field intensities is obtained through numerical integration. An advantage of the FDTD method is that it can obtain optical response for a wide wavelength range in a single simulation by Fourier transforming the results obtained by a brief input pulse. However, the FDTD simulation suffers from a considerable increase in calculation time when there is a glass substrate in the light path, which is much thicker than the other material layers within solar cells. To overcome this problem, in this study, we explore a new technique for FDTD simulation, to significantly reduce computational cost in the presence of a thick substrate, and apply this technique to the optical design of moth eye antireflection coatings on OPVs. In the proposed method, the absorption spectrum is averaged to eliminate the interference effect of light passing through a thick glass substrate and reproduce incoherent light addition within the substrate. We compare the proposed averaging method with the other two types of averaging methods, and show that the proposed algorithm gives the most accurate solution. In addition, our method is quite robust and independent of detailed values of control parameters. The validity of the results is checked by the comparison with the results obtained by the characteristic matrix method. In addition, we apply the proposed method to optimize the geometric pattern of moth eye array and maximize light trapping in OPVs. The results show that the combination of the FDTD simulation with our averaging method is quite useful to design broadband antireflection coatings for solar cells having thick glass substrates.