In the present study, we optimized medium components for the production of ergosterol peroxide (EP) by Paecilomyces cicadae based on a mono-factor experiment, a uniform design, and a non-linear regression analysis. A maximum EP yield of 256 μg/L was achieved, which was increased by 5-fold compared with that before the optimization. Structured Monod model, Andrews model, Contois model, and Aibe model were developed to describe the effects of viscosity inhibition, substrate, and production on biomass growth. The results showed that the Monod model could predict biomass growth, and the effects of viscosity and substrate on the EP concentration were significantly higher compared with the effect of production. Moreover, the addition of water and glycerol could decrease the viscosity inhibition and glycerol inhibition, and further increase the EP yield. The newly developed structured model was demonstrated for batch growth of P. cicadae.