In this paper, a novel semi-empirical model is developed for predicting the thermal conductivity of Diathermic Oil (DO)-based nanofluid for solar thermal application. First, a corrected thermal conductivity model of nanofluid is developed based on the widely used Yu-Choi model. In this corrected model, the thermal conductivity distribution of nanolayer is treated as a quadratic form instead of a constant value along with the distance from nanoparticle. The corrected model shows the predicted thermal conductivity of nanofluid is significantly influenced by nanolayer thickness. Then, the semi-analytical nanolayer thicknesses for different DO-based nanofluid are calculated utilizing the corrected model and the experimental data in literatures. It is found that the nanolayer thickness of DO-based nanofluid is not a constant value but varies with the volume fraction of nanoparticles, nanoparticle radius, and nanofluid temperature. According to the semi-analytical results, the empirical equations between the above variables and nanolayer thickness are summarized. Finally, by adopting the empirical equation of the nanolayer thickness in the corrected model, a novel semi-empirical thermal conductivity model is developed for predicting the thermal conductivity of DO-based nanofluid. The semi-empirical thermal conductivity model is validated by a large number of experimental data from literatures. (C) 2019 Elsevier Ltd. All rights reserved.