Analysis of thermosyphon heat exchangers for use in solar domestic hot water heating systems

A recent innovation in the solar industry is the use of thermosyphon heat exchangers. Determining the performance of these systems requires knowledge of how thermosyphon flow rate and heat exchanger performance vary with operating conditions. This study demonstrates that several thermosyphon heat exchanger designs operate in the laminar mixed convection regime. Empirical heat transfer and pressure drop correlations are obtained for three tube-in-shell heat exchangers (four, seven, and nine tube). Thermosyphon flow is on the shell side. Correlations are obtained with uniform heat flux on the tube walls and with a mixture of glycol and water circulating inside the tubes. Ranges of Reynolds, Prandtl, and Grashof numbers are 50 to 1800, 2.5 and 6.0, and {dollar}4times10sp5{dollar} to {dollar}1times10sp8,{dollar} respectively. Nusselt number correlations are presented in a form that combines the contributions of forced and natural convection, {dollar}rm Nusp4sb{lcub}Mixed{rcub}=Nusp4sb{lcub}Forced{rcub}+Nusp4sb{lcub}Natural{rcub}.{dollar} The Nusselt number is influenced by natural convection when the term {dollar}rm Rasb{lcub}q{rcub}sp{lcub}0.25{rcub}/(Resp{lcub}0.5{rcub}Prsp{lcub}0.33{rcub}){dollar} is greater than unity. Pressure drop through these three designs is not significantly affected by mixed convection because most pressure drop losses are at the heat exchanger inlet and outlet.; A comparison and discussion of the performance of several other heat exchanger designs (tube-in-shell and coil-in-shell designs) are presented. Generally, the coil-in-shell heat exchangers perform better than the tube-in-shell heat exchangers.; Data from all heat exchanger designs is used to develop a new one-dimensional model for thermosyphon heat exchangers in solar water heating systems. The model requires two empirically determined relationships, pressure drop as a function of water mass flow rate and the overall heat transfer coefficient-area product (UA) as a function of Reynolds, Prandtl, and Grashof number. A testing protocol is presented that describes the procedure to obtain the data for the correlations. Two new TRNSYS component models are presented, for the thermosyphon heat exchanger and thermosyphon loop. Unlike previous models, which are based on forced flow relationships, the new heat exchanger model accounts for mixed convection heat transfer and accurately predicts pressure drop in the connecting piping around the thermosyphon loop. Comparison between the model and experimental data shows excellent agreement. Daily and annual ratings for a sample thermosyphon system are presented.