Experimentally-validated models for the off-design simulation of a medium-size solar organic Rankine cycle unit

Organic Rankine Cycle is an efficient and reliable technology for the thermal-to-electricity conversion of low- grade heat sources but the variability in boundary conditions often forces these systems to operate at off- design conditions. The development of reliable models for the performance prediction of organic Rankine cycle power systems under off-design conditions is therefore crucial for system-level integration and control implementation. In this paper, a mathematical model for the evaluation of the expected performance of organic Rankine cycle power units in a large range of operating conditions based on experimental data collected in a medium-size solar organic Rankine cycle power plant is presented. Two different empirical approaches for the performance prediction of heat exchangers and machines, namely, constant-efficiency and correlated-based approaches, are proposed and compared. In addition, empirical correlations based on experimental data are proposed for the preliminary assessment of the energy demanded during the start-up phase and the corre sponding duration. Results demonstrate that a good achievement in terms of accuracy of the model and reli-ability of the simulation performance can be obtained by using a constant-efficiency approach, with average errors lower than 5% and 2.5 K for the expected net power and outlet oil temperature respectively. The use of polynomial correlations leads to a more accurate estimation of the performance parameters used for evaporator and the turbine (in particular the evaporator heat effectiveness and the isentropic and electromechanical effi-ciency for the turbine), which strongly affect the main output variables of the model and, at the same time, are remarkably influenced by the operating conditions. A reduction in the average error in the prediction of the net power and outlet temperature of the heat transfer fluid to about 4% and 1.5 K respectively is therefore achieved by this approach. Average errors of 18.5% and 12.5% are achieved for the start-up time and the corresponding energy absorbed, respectively. Although the results obtained in terms of accuracy could be improved, these correlations can give an initial indication about the duration and energy required during this phase.