Modeling of solar energy systems using artificial neural network: A comprehensive review

Abstract

The development of different solar energy (SE) systems becomes one of the most important solutions to the problem of the rapid increase in energy demand. This may be achieved by optimizing the performance of solar-based devices under some operating conditions. Intelligent system-based techniques are used to optimize the performance of such systems. In present review, an attempt has been made to scrutinize the applications of artificial neural network (ANN) as an intelligent system-based method for optimizing and the prediction of different SE devices’ performance, like solar collectors, solar assisted heat pumps, solar air and water heaters, photovoltaic/thermal (PV/T) systems, solar stills, solar cookers, and solar dryers. The commonly used artificial neural network types and architectures in literature, such as multilayer perceptron neural network, a neural network using wavelet transform, Elman neural network, and radial basis function, are also briefly discussed. Different statistical criteria that used to assess the performance of artificial neural network in modeling SE systems have been introduced. Previous studies have reported that artificial neural network is a useful technique to predict and optimize the performance of different solar energy devices. Important conclusions and suggestions for future research are also presented.

Introduction

Energy demand is growing worldwide due to the rapid population growth and the incredible evolution in industry (Can Şener et al., 2018). Renewable energy resources, which depend on natural resources to generate an infinite supply of energy that is sustainable and nonpolluting, is a promising alternative to the conventional energy resources and have gained significant importance in the recent centuries to overcome the energy shortages (Guven and Sulun, 2017). Among all renewable resources of energy, such as wind (Alhmoud and Wang, 2018), tidal (Loisel et al., 2018), biomass (Barnes, 2015) and geothermal (Tomasini-Montenegro et al., 2017), SE have attracted the greatest attention in many engineering and industrial applications such as solar refrigeration (Chen et al., 2018), electricity generation (Jain et al., 2018), domestic space heating (Sharma et al., 2017), solar chimney (Kasaeian et al., 2017), solar radiation prediction (Kasaeian et al., 2016), and water desalination (Sharshir et al., 2016).

There are many devices which utilize SE to do some useful work such as solar collectors, solar assisted heat pumps, solar air and water heaters, PV/T systems, solar stills, solar cookers, and solar dryers. Numerous experimental and theoretical studies have been performed to figure out how the different operating parameters affect each of the aforementioned devices. However, the experimental studies are constrained by the number of the conducted experiments which may require high cost and consume more time. Moreover, the experimental based results are suitable only for little combinations of the investigated parameters with a little number of levels for each parameter. Theoretical based results, on the other hand, are suitable only for simplified models of the practical devices under many simplifying assumptions. It is worth mentioning that, the key reason of using ANN based models instead of analytical based methods is that the analytical methods are generally quite cumbersome (sometimes solving of high nonlinear partial differential equations is needed) (Bellos and Tzivanidis, 2018, Elsheikh et al., 2018, Salazar et al., 2017) and require many simplifying assumptions. However, ANN based methods have been proved to be a useful tool to model different engineering systems under real-world conditions without involving in solving complicated mathematical models. ANN, as an intelligent based technique which mimics the behavior of the human brain in dealing with different problems instead of solving complex mathematical models, is used as a black-box model with the ability to learn and find out the nonlinear relation between the system inputs and the system outputs. ANNs have the generalization capability as it can handle unseen data faster and simpler than other classical methods after a learning process using few measured data sets. Therefore, ANN-based methods have attracted the attention of scientists and researchers in different engineering and industrial disciplines, for instance, for modeling and identification of mechanical systems (Orlowska-Kowalska and Szabat, 2007, Singh et al., 2009). Many researchers have summarized the use of the ANNs in many applications of renewable energy systems (Kalogirou, 2001) such as solar radiation prediction (Bou-Rabee et al., 2017, Kashyap et al., 2015, Shaddel et al., 2016), grid-connected photovoltaic power system (Ferlito et al., 2017), wind energy (Karabacak and Cetin, 2014), geothermal systems (Liu et al., 2018), photovoltaic applications (Almonacid et al., 2017, Mellit and Kalogirou, 2008, Miloudi and Acheli, 2015, Yadav et al., 2018), solar dryers (Prakash et al., 2016), and solar collectors (Ghritlahre and Prasad, 2018a). However, to our best knowledge, there are no published reviews on the application of ANNs in SE systems. For these reasons, this paper sheds some light on the application of ANNs in modeling and optimization of different SE systems. The commonly used architectures of ANNs that have been used in modeling SE systems have been briefly discussed. Different statistical criteria that used to assess the performance of ANN in modeling SE systems have been introduced. Therefore, this review has long term scientific value for SE researchers and beginners.

This comprehensive review covers the following points:

• A brief discussion about the types of ANNs, such as multilayer perceptron neural network, neural networks using the wavelet, the function of the radial basis, and Elman neural network (ENN), used in different SE applications is introduced.

• Different activation functions types used in different ANNs as well as different standard statistical performance evaluation criteria used in the evaluation of ANNs performance are also briefly discussed.

• The application of ANNs in different solar devices, like solar collectors, solar assisted heat pumps, solar air and water heaters, photovoltaic/thermal (PV/T) systems, solar stills, solar cookers, and solar dryers, are summarized in detail.

• Conclusions and suggestions for future research are also presented.