Critical discussion of the design method for horizontally arranged evaporators in supercritical energy systems.

In emerging supercritical energy systems like Organic Rankine cycle and Brayton cycle, the evaporator where supercritical fluid is heated is often horizontally placed. Accurate prediction of supercritical heat transfer coefficients is the basis of the heat exchanger design. However, existing correlations for horizontal supercritical flow were primarily for cooling condition, and though a few correlations were developed under heating condition, the thermal boundary was uniform heat flux instead of the convective boundary condition in practical applications. This work focuses on the design method of horizontally arranged supercritical evaporators, and explores the predictive capabilities as well as the limitations of existing correlations. On the basis of the general axial-segmented approach, circumferential segments were considered to account for the buoyancy effect in horizontal flow. Results show that all 8 correlations considered have certain limitations, overestimating or underestimating the heat transfer capacity depending on the heat flux and mass flux. The Krasnoshchekov and Protopopov correlation exhibited the highest predictive capability, accurately capturing 93.3 % of the 105 experimental CO 2 cases within a 5 °C error in outlet temperature. The newly introduced circumferential segments led to improved performance, especially when mean wall temperature was used to calculate average heat transfer coefficients. Finally, this work revealed the paradox between 'convergent' and 'optimal' solutions in the segmented approach, and discussed the issue of multiple solutions during the calculation of wall temperature and heat flux. • We studied the segment approach for the design of evaporator at supercritical state. • Nusselt correlations developed under various thermal boundaries were evaluated. • Correlations basing on circumferential average T w perform best against test database. • The circumferentially segmented approach improves prediction accuracy. • The issues of convergent and multiple solutions can bring great uncertainties. [ABSTRACT FROM AUTHOR]