Prediction accuracy is a fundamental modeling requirement. This work explores different models for a solar domestic water heating system located near Vina del Mar, Chile, in order to improve model prediction accuracy over some existing alternatives. The main approach is semi-physical modeling, which combines phenomenological modeling and system identification. The former helps to organize in a conceptual form the available system knowledge, and the latter allows to adjust that knowledge into a particular model structure for a working system. Thus, with semi-physical modeling we take advantage of a basic property of classical identification models: they are linear-in-the-parameters, but may contain nonlinear regressors. Hence, while physical knowledge suggests nonlinear data regressors, system identification adjusts linear weighting parameters. The models proposed here incorporate nonlinearities based on physical system knowledge and they include, among other inputs and disturbances, air wind speed (v) and air relative humidity (RH), signals not usually considered in these model structures. Specifically, this work shows model predictive accuracy of storage tanks temperature for three model types: semi-physical, state-space and, a combination of semi-physical and a feedforward neural network with one hidden layer and eight neurons. The best models found here, according to prediction accuracy, are of semi-physical nature, and are obtained using stepwise regressor elimination algorithms which retain disturbances such as v and RH. Additionally, final models are validated with classical statistical tests such as AIC and correlation analysis.
[1]
Soteris A. Kalogirou,et al.
Thermosiphon solar domestic water heating systems: long-term performance prediction using artificial neural networks
,
2000
.
[2]
N. Draper,et al.
Applied Regression Analysis
,
1966
.
[3]
A. D. Jones,et al.
A thermal model for photovoltaic systems
,
2001
.
[4]
Astrom.
Computer Controlled Systems
,
1990
.
[5]
Lennart Ljung,et al.
Nonlinear black-box modeling in system identification: a unified overview
,
1995,
Autom..
[6]
Lennart Ljung,et al.
System Identification: Theory for the User
,
1987
.
[7]
Keith R. Godfrey,et al.
Perturbation signals for system identification
,
1993
.
[8]
Lennart Ljung,et al.
Neural Networks in System Identification
,
1994
.
[9]
Urban Forssell,et al.
Combining Semi-Physical and Neural Network Modeling: An Example ofIts Usefulness
,
1997
.
[10]
Petre Stoica,et al.
Decentralized Control
,
2018,
The Control Systems Handbook.
[11]
Soteris A. Kalogirou,et al.
MODELING OF SOLAR DOMESTIC WATER HEATING SYSTEMS USING ARTIFICIAL NEURAL NETWORKS
,
1999
.