OCTOBER 19-20, 2012 - YMCA University of Science & Technology

Proceedings of the National Conference on
Trends and Advances in Mechanical Engineering,
YMCA University of Science & Technology, Faridabad, Haryana, Oct 19-20, 2012
An ANN model has been developed to predict diffuse fraction in hourly and daily scale [35] in the plain areas of
Egypt and compared its performance with two linear regression models. They reported that the result obtained
from ANN model is more suitable for prediction of the diffuse fraction in hourly and daily scales than the
regression models. Furthermore, a new methodology based on artificial neural networks (ANN) has been
implemented to evaluate the luminous efficacy of diffuse, direct and global solar radiation with clear sky
conditions [36]. With the help of standard statistical techniques developing a non-local model considering all
physical processes is nearly impossible. The results show that an ANN model is simpler than the SMARTS
(Simple Model of the Atmospheric Radiative Transfer) radiation model and able to accurately highlight the
variations of the three components of luminous efficacy caused due to solar zenith angle, various atmospheric
absorption and scattering processes. The developed ANN model can be used without using detailed atmospheric
information or empirical models, if radiometric measurements and perceptible water data (or temperature and
relative humidity data) are available.
Recently, model based on artificial neural network with wavelet analysis has been used for solar radiation
estimation. A hybrid model based on this were developed by Cao and Cao [37] and used for forecasting
sequences of total daily solar radiation. Cao and Lin [38] proposed a new model based on diagonal recurrent
wavelet neural network (DRWNN) and a special designed training algorithm for forecasting global solar
irradiance. Mellit et al. [39] implemented an adaptive neural-network topology with the wavelet transformation
embedded in the hidden units for forecasting daily total solar radiation. They investigated several structures
which have been beneficial in resolving the missing data problem. A hybrid model based on a neural network
with Markov chain has been proposed for generating total daily solar radiation [40] at long term and it was
implemented for Algeria. The unknown validation data set generated very accurate forecast with an RMSE error
less than 8% between the predicted and measured data. A recurrent neural network with MLP network used for
generating solar radiation synthetic series [41]. In this study, the MLP and other two models has been compared
and found that values of the annual irradiance of synthetic year estimated by MLP method were nearer to the real
data than other two methods.
For more detailed investigation regarding the application of the artificial intelligence techniques for modeling
and forecasting of the solar radiation and solar energy modeling techniques the reader can follow some of the
good reviews presented in [42, 43].
2. CONCLUSIONS
In the present study, a review of AI techniques used in solar energy systems has been carried out. From this
study, following conclusions have been drawn:
1. Use of Artificial Intelligence techniques result in achieving substantial improvement in efficiency and
predicting the optimal set of design and operating variables for the solar energy systems.
2. From this study, it is clear that after training the ANN model has the potential to predict the satisfactory
results for unknown data.
3. From figure 2, it is clear that ANN predicted more accurate optimal set of variables after the model has been
trained and validated well.
4. In solar energy systems, there is a lot of scope for using combination of AI techniques with other
optimization techniques in order to improve the performance of the system.
5. AI techniques may be applied on roughened solar air heater, which is a scope for future work.
Life cycle savings (LCS) and Life cycle assessment (LCA) has been also carried out for solar energy systems
using AI techniques.
3. REFERENCES
[1] C.M. Bishop, Neural networks for pattern recognition, Oxford University Press, Oxford, 1995.
[2] D. Saxena, S.N. Singh, K.S. Verma, Application of computational intelligence in emerging power systems,
International Journal of Engineering, Science and Technology, 2 (2010) 1-7.
[3] M.H. Ahmadi, M.A. Ahmadi, S.S.G. Aghaj, Prediction of Power in Solar Stirling Heat Engine by Evolving
Particle Swarm Optimization and Neural Network, International Journal of Computer Applications, 34 (1)
(2011) 20-24.
[4] Y. Li, Y. H, W. Wang, Optimization of Solar-powered stirling heat engine with finite-time
thermodynamics, Renewable Energy 36 (2011) 421-427.
[5] S.A. Kalogirou, Optimization of solar systems using artificial neural-networks and genetic algorithms,
Applied Energy, 77 (2004) 383-405.
[6] M. Petrakis, H. Kambezides, S. Lykoudis, A. Adamopoulos, P. Kassomenos, I. Michaelides, S. Kalogirou,
G. Roditis, I. Chrysis, A. Hadjigianni, Generation of a typical meteorological year for Nicosia, Cyprus,
Renewable Energy, 13(1998) 318-388.
192