Thermosyphon solar water heaters : validated numerical simulation and design correlations

A detailed analysis of the heat transfers and fluid flows within a direct thermosyphonic solar-energy water-heater has been undertaken. The collector energy equations when cast in a two-dimensional form enabled heat transfer and thermal capacitance effects to be simulated accurately at the small flow rates encountered commonly in such systems. An investigation of thermocline relaxation processes within the store indicated negligible mixing at the store inlet over a wide range of Richardsons numbers (43,608 < Ri < 729,016). Thermal relaxation under conditions of no flow was shown to be due predominantly to axial conduction along the store wall. The use of an appropriate non-isothermal friction factor correlation when calculating frictional losses in the collector's riser pipes, produced predicted steady-state flow rates which were corroborated experimentally to within 2%. An indoor test facility, monitored and controlled by a microcomputer, enabled "real" operating conditions to be simulated. The predicted responses of the system to identical conditions showed good agreement with the corresponding experimental observations, the predicted heat delivery being within 2.8% of the measured value. A technique for correlating the daily performances of thermosyphon solar-energy water-heaters has been developed. The five dimensionless groups which form the basis of the correlations and the functional relationships between these groups were derived from an analytical solution of a linear first-order differential transient heat balance carried out on a generic system. Thermal performance data used in the correlations was generated by the numerical simulation using representative U. K. hourly weather data and operating conditions. The minimum amount of data required to establish a characteristic curve for an individual system was found to be thirty days. Using such a curve, the total annual solar fraction agreed with that predicted by the high level model to within 3%. Two universal curves were determined in which the gradients of characteristic curves were correlated against the derived dimensionless groups. The accuracy of the resulting two-stage algorithm in determining annual solar fractions was established as ranging from 5.5% for predominantly multiple-pass systems to a mean of 10.5% for single-pass systems