In this study, effects of particle shape on Al2O3-water nanofluids laminar forced convection in developing and fully developed regions of a curved square duct were investigated numerically using Eulerian-Lagrangian two-phase approach. In order to improve the accuracy of the two-phase model for laminar convective heat transfer of nanofluids containing non-spherical nanoparticles, two new nanoparticle shape descriptors, flatness and elongation, were introduced. Compared with base fluid (water), nanofluids containing platelet shaped nanoparticles has the highest heat transfer enhancement, which is followed by nanofluids containing nanoparticles with cylinder, blade, sphere and brick shapes, respectively. Non-spherical nanoparticles with a suitable shape, small size and relatively high volume fraction are beneficial for enhancement of heat transfer in laminar forced convection. In developing region, a pair of Dean vortices formed and grew along the duct axis, which affected nanoparticle concentration distribution and heat and mass transfer. In fully developed region, convective heat transfer efficiencies of nanofluids are larger than 1 and vary with nanoparticle shape, size, volume fraction and Reynolds number. Enhancement of the convective heat transfer in nanofluids was attributed to the enhancement of effective thermal conductivity and effective viscosity, change of flow structure and reduction of thermal boundary layer thickness due to the presence of nanoparticles and their shapes. New correlations of Nusselt number and fraction factor with nanoparticle shape (sphericity, flatness and elongation), size and volume fraction were developed in order to predict convective heat transfer of nanofluids containing spherical and non-spherical nanoparticles.