An improved convection heat transfer correlation for supercritical fluids was developed based on the underlying physical mechanism of turbulent flow of working fluids with variable properties. The new correlation employs a more physically reasonable property-averaging technique, Probability Density Function (PDF)-based time-averaged properties, to account for the effect of nonlinear dependency of properties on instantaneous temperature. In addition, the buoyancy and thermally induced acceleration modification equations proposed by Jackson can also be used with the new correlation and work well in evaluating the buoyancy and thermally induced acceleration effect on the normal convection heat transfer. The new correlation was validated with a large amount of heat transfer experimental data including forced convection heat transfer data for supercritical CO2 flow in a horizontal semicircular printed circuit heat exchanger, convection heat transfer data considering buoyancy for supercritical water upward flow in a square annular channel, and convection heat transfer data considering effect of thermally induced flow acceleration for supercritical methane flow in a horizontal miniature circular tube. Comparison of experimental data with the correlation prediction results reveals that the new correlation predicts more accurate than conventional correlations for typical supercritical working fluids of several advanced energy systems. (C) 2017 Elsevier Ltd. All rights reserved.