A new mathematic model for heat conduction of nanofluid is developed based on the experimental findings. Effective medium theory, nanolayer of liquid molecule around solid particle, aggregation, nano-convection due to the Brownian motion of nanoparticle, and interfacial thermal resistance are included to elucidate the heat conduction mechanism in nanofluids. The analytical result fits well with the experimental data and the maximum deviation is obtained to be 1.52% for SiO2 nanofluid. The effects of aggregate shape (i.e., ellipsoid, sphere and fiber) and its size are investigated to evaluate the thermal conductivity of the nanofluids. The prediction shows that nano-convection induced by the movement of aggregates, is leading the main contribution for thermal conductivity enhancement at a low concentration of similar to 0.1 vol%. Thermal conductivity of aggregate becomes crucial to affect the static contribution for the enhancement. In addition, it is found that the interfacial thermal resistance and nanolayer have little effect on the thermal conductivity enhancement of nanofluids at a very low concentration of nanoparticles. (c) 2014 Elsevier Ltd. All rights reserved.