In this study, natural convection heat transfer and entropy generation of the non-Newtonian power-law nanofluid including Al2O3 nanoparticles, inside a cylindrical annular cavity with a concentric circular heat source covered with a conductive porous layer are investigated numerically. The nanofluid is modeled using two-phase mixture model and the mixture viscosity and thermal conductivity are computed by Corcione's correlations. The effect of partial or complete filling of the enclosure with porous media for various Rayleigh (Ra = 10(4)-10(6)), Darcy (Da =10(-4)-10(-1)), power-law index (n = 0.6-1.4), effective to base fluid thermal conductivity ratio (k(eff)/k(f) =16, 4) and the porous layer thickness (R-d) are studied on heat transfer, entropy generation and the overall performance. Results are presented and compared in terms of the average Nu, non-dimeniional entropy generation, Bejan number, streamline and isotherm contours and performance coefficient (PE). Outcomes indicate that for k(eff)/k(f) = 16 the fully porous cavity is recommended, while for k(eff)/k(f) = 4, depending on the value of Ra, Da and n parameters a proper porous layer thickness should be selected. Also, shear-thinning nanofluids show higher Nu with respect to the other studied cases. Fully porous cavities have the lowest entropy generation with the highest PE value, and for shear-thinning nanofluids the highest PE corresponds to Ra = 10(4). (C) 2017 Elsevier Ltd. All rights reserved.