Modeling a ground-coupled heat pump system by a support vector machine

被引:158
作者
Esen, Hikmet [1 ]
Inalli, Mustafa [2 ]
Sengur, Abdulkadir [3 ]
Esen, Mehmet [1 ]
机构
[1] Firat Univ, Fac Tech Educ, Dept Mech Educ, TR-23119 Elazig, Turkey
[2] Firat Univ, Fac Engn, Dept Mech Engn, TR-23279 Elazig, Turkey
[3] Firat Univ, Fac Tech Educ, Dept Elect & Comp Sci, TR-23119 Elazig, Turkey
关键词
ground coupled heat pump performance; support vector machine; forecast; artificial neural network; adaptive neuro-fuzzy inference system;
D O I
10.1016/j.renene.2007.09.025
中图分类号
X [环境科学、安全科学];
学科分类号
08 ; 0830 ;
摘要
This paper reports on a modeling study of ground coupled heat pump (GCHP) system performance (COP) by using a support vector machine (SVM) method. A GCHP system is a multi-variable system that is hard to model by conventional methods. As regards the SVM, it has a superior capability for generalization, and this capability is independent of the dimensionality of the input data. In this study, a SVM based method was intended to adopt GCHP system for efficient modeling. The Lin-kernel SVM method was quite efficient in modeling purposes and did not require a pre-knowledge about the system. The performance of the proposed methodology was evaluated by using several statistical validation parameters. It is found that the root-mean squared (RMS) value is 0.002722, the coefficient of multiple determinations (R-2) value is 0.999999, coefficient of variation (cov) value is 0.077295, and mean error function (MEF) value is 0.507437 for the proposed Lin-kernel SVM method. The optimum parameters of the SVM method were determined by using a greedy search algorithm. This search algorithm was effective for obtaining the optimum parameters. The simulation results show that the SVM is a good method for prediction of the COP of the GCHP system. The computation of SVM model is faster compared with other machine learning techniques (artificial neural networks (ANN) and adaptive neuro-fuzzy inference system (ANFIS)); because there are fewer free parameters and only support vectors (only a fraction of all data) are used in the generalization process. (C) 2007 Elsevier Ltd. All rights reserved.
引用
收藏
页码:1814 / 1823
页数:10
相关论文
共 25 条
[1]  
ANSTETT M, 1993, ASHRAE TRAN, V99, P505
[2]   Artificial neural network analysis of heat pumps using refrigerant mixtures [J].
Arcaklioglu, E ;
Erisen, A ;
Yilmaz, R .
ENERGY CONVERSION AND MANAGEMENT, 2004, 45 (11-12) :1917-1929
[3]   Neural networks - a new approach to model vapour-compression heat pumps [J].
Bechtler, H ;
Browne, MW ;
Bansal, PK ;
Kecman, V .
INTERNATIONAL JOURNAL OF ENERGY RESEARCH, 2001, 25 (07) :591-599
[4]   Applicational aspects of support vector machines [J].
Belousov, AI ;
Verzakov, SA ;
von Frese, J .
JOURNAL OF CHEMOMETRICS, 2002, 16 (8-10) :482-489
[5]  
BERDAL VT, 2005, P 9 IBPSA C MONTR CA, P1245
[6]  
Bose J.E., 2002, P 7 INT EN AG HEAT P, V324, P313
[7]   On the modelling of nonlinear dynamic systems using support vector neural networks [J].
Chan, WC ;
Chan, CW ;
Cheung, KC ;
Harris, CJ .
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, 2001, 14 (02) :105-113
[8]  
Curtiss J F, 1994, ASHRAE T, V100, P476
[9]   Modeling hourly energy use in commercial buildings with Fourier series functional forms [J].
Dhar, A ;
Reddy, TA ;
Claridge, DE .
JOURNAL OF SOLAR ENERGY ENGINEERING-TRANSACTIONS OF THE ASME, 1998, 120 (03) :217-223
[10]   The greedy search algorithm on binary vectors [J].
Dor, A .
JOURNAL OF ALGORITHMS, 1998, 27 (01) :42-60