Development of a fully deterministic simulation model for organic Rankine cycle operating under off-design conditions

•A fully deterministic simulation model is newly developed for the off-design ORC.•The component models and reality-based logics are integrated into a three-stage solver.•Computational costs are significantly reduced by employing surrogate meta-models.•Critical issues on imposing internal assumptions on the model are investigated.•The proposed model can detect operation failure and predict off-design performance. A fully deterministic simulation model that requires only external boundary conditions as input parameters is newly developed for predicting the off-design performance of an organic Rankine cycle. Accurate prediction of the evaporation and condensation pressures without using any assumptions has been of major concern. In this context, the actual pressure formation characteristics are reflected into the model to identify the high- and low-pressures. The mass balance of the system is realized by applying the proposed passive design of the liquid receiver. The sub-models of each component are integrated into a three-stage solver which iterates under the law of mass and energy conservation. 77 sets of experimental data were collected from a 1 kW scale testbed using R245fa as the working fluid. The developed model is verified by examining the thermodynamic states of the working fluid and the energy balance error of the simulation results is less than 0.3%. The simulation results are compared to the experimental results and are validated within 8% error range. Also, the computational time reduced from 412 h to 93 s after applying the meta-models with only 0.03% difference in thermal efficiency. The effects of various modelling methods are compared to each other which emphasizes the importance of the newly proposed reality-based logics. The simulation model can detect several operational failure scenarios and accurately predict the off-design performance of the system. The developed model is fully predictive without imposing internal assumptions and has the potential to be utilized in various applications without conducting excessive experiments.