Abstract :
[en] The organic Rankine cycle (ORC) is among the most suited technologies to convert low-grade and low-capacity heat sources into useful work. For many reasons - including predictive control, optimal sizing or performance forecast - a proper understanding and characterization of the ORC behaviour under off-design conditions is of significant interest. In order to avoid any intrinsic state assumption, predictive models must account for both fundamental conversation laws of hermodynamics, namely the conservation of energy and the conservation of mass. Besides of modelling the energy transfers, a true off-design model must account for the constant amount of working fluid in the system and simulate its distribution among the different components. Although well-known for HVAC systems, such charge-sensitive considerations are quasi absent for the ORC technology and existing models have never been completely validated. The goal intended by this PhD thesis is to fill this gap. To begin this work, experiments are conducted to assess the fluid distribution and the impact of the charge on a real system operation. To this end, a 2 kWe ORC test rig is constructed and tested over wide range of conditions. Besides of standard thermohydraulic sensors, the fluid charge distribution is measured on-line by bending load cells and infrared imaging techniques. Following a complete experimental campaign (which includes more than 300 steady-state points), a dual reconciliation method is applied on the raw measurements to obtain a reference dataset. An extensive study of the experimental data is then conducted. Among many results, the important contribution of the heat exchangers in the charge inventory is highlighted, so as the impact of oil circulation on the ORC performance rating. Following this experimental study, a complete modelling library is developed to replicate and extrapolate the system off-design behaviour. The intended goal is to create a true performance simulator, i.e. a predictive tool able to estimate the ORC behaviour based solely on its boundary environment (without state assumption, i.e. accounting for the charge distribution in the system). In a first step, a miscibility model of R245fa/POE oil is developed to account for the presence of the lubricant in the ORC operation. Afterwards, the modelling of each system component is conducted, with a particular focus on the heat exchangers and their charge estimation. Then a global ORC model is constructed by coupling the various components sub-models accordingly to a robust resolution scheme. The ORC model predictions are ultimately confronted to the experimental measurements, both in terms of thermodynamics and charge inventory predictions, and a good fit is demonstrated for all the model outputs. Finally, the utility of such a charge- and lubricant-sensitive ORC model is highlighted for different tasks. Considering the experimental test rig as case study, the off-design modelling tool is employed (i) to fully characterize the ORC response under off-design conditions, (ii) to prevent operating conditions where pump cavitation is likely to occur, (iii) to build optimal performance mappings for full- and part-load operations, and, finally, (iv) to optimally select the charge of working fluid and to accordingly size the liquid receiver.
