[home] [Personal Program] [Help]
tag
10:50
20 mins
Extrapolability and limitations of a semi-empirical model for characterizing volumetric expanders
Olivier Dumont, Rémi Dickes, Vincent Lemort
Session: Session 1C: Volumetric Expanders (1)
Session starts: Wednesday 13 September, 10:30
Presentation starts: 10:50
Room: Building 27 - Lecture room 03


Olivier Dumont (University of Liège)
Rémi Dickes (University of Liège)
Vincent Lemort (University of Liège)


Abstract:
Many different modelling approaches can be used to simulate the performance of expansion devices in ORC power systems. A decade ago, researchers proposed a semi-empirical modelling method that can be used to generally characterize volumetric expanders (e.g. scroll, screw, piston or vane machines) [1]. The modelling approach (as depicted in Figure 1) relies on a limited number of physically meaningful equations which decompose the expansion process into six consecutive steps. Besides of under- and over-expansion losses (due to the fixed built-in volumetric ratio of the machine), the model can account for pressure drops at the inlet and outlet ports, internal leakages, mechanical losses and heat losses to the environment. The main interests of the modelling approach are a low CPU time, extrapolability through the use of physical laws and the same formalism for all the volumetric expander technologies (scroll, screw, piston, vane…). The semi-empirical model relies on different parameters that must be properly tuned accordingly to experimental (or manufacturer) data. In practice, however, the reference database used for the model calibration (e.g. the measurements gathered on a test rig) does not necessarily cover the entire range of conditions onto which the model will be evaluated. The capability of the semi-empirical model to behave well in extrapolated conditions has therefore to be assessed. In this work, a detailed analysis of the extrapolation performance of the semi-empirical model is conducted. More specifically, the semi-empirical model behavior is analyzed after being calibrated with different ranges of reference conditions. A study of the minimum reference dataset to ensure a decent modelling accuracy is proposed. Finally, the influence of the parameters guess values and the optimization algorithm on the model calibration is assessed.