Artificial neural networks for predicting the local concentration ratio of parabolic trough collectors
Date Issued
1996
Author(s)
Abstract
The concentrated radiant flux on receiver surfaces of parabolic trough
collectors is not uniform but exhibits a “bell” type shape as presented in this paper.
This flux is given in terms of the local concentration ratio (LCR) being in effect the
ratio of the incoming to the concentrated value of solar radiation at the periphery of
the receiver. Knowledge of the LCR is required not only at normal conditions but also
at various incident angles. This is useful to a designer who wants to calculate the
intercept factor and the optical efficiency of the concentrator at those angles. In the
work presented here actual values of LCRs, at ten degree intervals around the
receiver, for two different incident angles (15° and 60°) are used to train an artificial
neural network. Subsequently the network is used to predict the LCR values at other
incident angles.The matching of the data for the 15° and 60° is very good with an R2-
value equal to 0.9997. For the unknown data at incident angles of 0°, 30° and 45° the
R2-values are 0.9966, 0,9867 and 0.9765 respectively which indicates that the
estimation is performed with adequate accuracy. By using the predicted LCR values,
the intercept factor is estimated with a maximum deviation of 3.2% from the value
estimated with the actual LCR values which is very adequate.
collectors is not uniform but exhibits a “bell” type shape as presented in this paper.
This flux is given in terms of the local concentration ratio (LCR) being in effect the
ratio of the incoming to the concentrated value of solar radiation at the periphery of
the receiver. Knowledge of the LCR is required not only at normal conditions but also
at various incident angles. This is useful to a designer who wants to calculate the
intercept factor and the optical efficiency of the concentrator at those angles. In the
work presented here actual values of LCRs, at ten degree intervals around the
receiver, for two different incident angles (15° and 60°) are used to train an artificial
neural network. Subsequently the network is used to predict the LCR values at other
incident angles.The matching of the data for the 15° and 60° is very good with an R2-
value equal to 0.9997. For the unknown data at incident angles of 0°, 30° and 45° the
R2-values are 0.9966, 0,9867 and 0.9765 respectively which indicates that the
estimation is performed with adequate accuracy. By using the predicted LCR values,
the intercept factor is estimated with a maximum deviation of 3.2% from the value
estimated with the actual LCR values which is very adequate.
Subjects
File(s)![Thumbnail Image]()
Name
C13-EUROSUN.pdf
Size
64.15 KB
Format
Adobe PDF
Checksum (MD5)
0b3c73df5926bedb618fdc97b358d1fb