Modeling of thermodynamic properties of refrigerant/absorbent couples using data mining process

• Published in: Energy conversion and management Vol. 48; no. 2; pp. 470 - 480
• Main Author: SENCAN, Arzu
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.02.2007; Elsevier
• Subjects: Applied sciences; Back propagation neural network; Devices using thermal energy; Energy; Energy. Thermal use of fuels; Exact sciences and technology; Heat pumps; M5 model tree; M5’Rules; Methanol/LiBr; Methanol/LiCl; Pace regression; SMO; Thermodynamic properties; Methanol/LiCl; Pace regression; SMO; Methanol/LiBr; M5 model tree; M5’Rules; Back propagation neural network; Thermodynamic properties; Absorbent; Methanol/LiCI; Linear regression; Neural network; Data mining; Modeling; Optimization; Absorption heat pump; M5'Rules; Refrigerant fluid; Methanol
• ISSN: 0196-8904; 1879-2227
• DOI: 10.1016/j.enconman.2006.06.018

In this paper, in order to determine thermodynamic properties of two alternative refrigerant/absorbent couples (methanol/LiBr and methanol/LiCl), a data mining process was used. These fluid couples can be used in absorption heat pump systems, and their main advantage is that they do not cause ozone depletion. In order to train the network, limited experimental measurements were used as training and test data. In the present study, linear regression (LR), pace regression (PR), sequential minimal optimization (SMO), M5 model tree, M5’Rules and back propagation neural network (BPNN) models are applied within the data mining process for determining the specific volume of the methanol/LiBr and methanol/LiCl fluid couples. The best result was obtained by using the back propagation model. A new formulation is presented for determination of the specific volumes of the two refrigerant/absorbent couples. The use of this new formulation, which can be employed with any programming language or spreadsheet program for estimation of the specific volumes of fluid couples, as described in this paper, may make the use of dedicated BPNN software unnecessary.