Estimation of solar radiation using artificial neural networks with different input parameters for Mediterranean region of Anatolia in Turkey

https://doi.org/10.1016/j.eswa.2011.01.085Get rights and content

Abstract

An artificial neural network (ANN) model was used to estimate the solar radiation parameters for seven cities from Mediterranean region of Anatolia in Turkey. As well known that Turkey is a bridge between Asia and Europe and it lies in a sunny belt, between 36° and 42°N latitudes. Indeed, the country has sufficient solar radiation intensities for solar applications. In order to make estimation of solar radiation, the data from the Turkish State and Meteorological Service were used. Data of 2006 were used for testing and data of 2005, 2007, and 2008 were estimated. Effects of number of input parameters were tested on solar radiation that was output layer. With this aim, number of input layer parameters changed from 2 to 6. The obtained results indicated that the method could be used by researchers or scientists to design high efficiency solar devices. It was also found that number of input parameters was the most effective parameter on estimation of future data on solar radiation.

Research highlights

► An artificial neural network (ANN) model was used in this work. ► The main originality of the study was the testing of input parameters on results and R2 values. ► The input parameters were the most effective parameters on estimation of future data on solar radiation.

Introduction

Thermal analysis of building, heating, ventilation, solar energy, thermal comfort in buildings and air conditioning system requires information on values of some meteorological parameters such as ambient temperature, relative humidity, wind velocity and solar radiation. Thus, estimation of these values is important for some applications as agriculture, environment and food industry. It is also necessary for sustainable development and environmental protection. More specifically, Forecasting of solar radiation is important parameter for studies of photovoltaic, solar energy supported drying systems, furnaces, water heaters, etc.

Turkey has an important advantage on solar energy potential because of its geographical position in the northern hemisphere. Turkey lies in a sunny belt, between 36–42°N latitudes and 26–45°E longitudes (Aras et al., 2006, Gunes, 2001, Ulgen and Hepbasli, 2004). Estimation studies of meteorological data can be classified in two main groups: The first group includes statistical indicators such as the relative percentage error (E), coefficient of determination (R2), the mean percentage error (MPE), the mean absolute percentage error (MAPE), the sum of the squares of relative errors (SSRE), the relative standard error (RSE), the mean bias error (MBE) and the root mean square error (RMSE) (Bakirci, 2009, Bilgili et al., 2007, Celiktas et al., 2009, Mondol et al., 2008, Togrul and Onat, 1999, Togrul and Onat, 2000, Togrul and Togrul, 2002, Ulgen and Hepbasli, 2009). Combination of MBE and RMSE gives another method, defines as t-statistic (t-stat) method. The second group for estimation of values is soft computing codes. Namely, artificial neural networks (ANN), adaptive-network-based fuzzy inference system (ANFIS) and support vector machines (SVM).

Fadare (2009) developed an artificial neural network (ANN) based model for prediction of solar energy potential in Nigeria (lat. 4–14°N, log. 2–15°E). He applied the standard multilayered, feed-forward, back-propagation neural networks with different architecture using neural toolbox for MATLAB and indicated that the model can be used easily for estimation of solar radiation for preliminary design of solar applications. Sozen, Arcaklioglu, Ozalp, and Kanit (2004) mapped the solar potential for Turkey using artificial neural networks. They used the scaled conjugate gradient (SCG), Pola–Ribiere conjugate gradient (CGP), and Levenberg–Marquardt (LM) learning algorithms and a logistic sigmoid transfer function in the network. They indicated that the trained and tested ANN models show greater accuracies for evaluating solar resource possibilities in regions where a network of monitoring stations has not been established. Senkal and Kaleli (2009) made a study to estimate the solar radiation in Turkey using resilient propagation (RP), scale conjugate gradient (SCG) learning algorithms and logistic sigmoid transfer function in the artificial neural network. Dombayci and Golcu (2009) used the artificial neural network model to predict daily mean ambient temperatures in Denizli, south-western Turkey. They showed that the ANN approach is a reliable model for ambient temperature prediction. Bosch, Lopez, and Batlles (2008) made an estimation study on solar radiation using ANN with digital terrain model. Then, the model was applied to daily solar irradiation estimation over a mountainous area. Jiang (2008) studied on comparison of models of artificial neural networks and empirical models to predict monthly mean daily diffuse solar radiation in China. His results of validation and comparative study indicate that the ANN-based estimation technique for solar radiation is more suitable to predict solar radiation than the empirical regression models proposed by other researchers. ANNs were used to develop prediction models for daily global solar radiation using measured sunshine duration for 40 cities covering nine major thermal climatic zones and subzones in China (Lam, Wan, & Yang, 2008).

Many of researchers were used the ANN models to make prediction of solar energy values as Dorvlo et al., 2002, Elminir, 2005, Elminir et al., 2007, Reddy and Ranjan, 2003, Alam, 2006, Al-Lawati, 2003, Zhou et al., 2005, Tymvios et al., 2005, Sozen et al., 2005, Sozen, 2004, Rehmana and Mohandes, 2008, Mubiru and Banda, 2007, Mohandes et al., 1998, Mellit et al., 2005.

The main objective of this study was to estimate the solar data using artificial neural network. As given in the literature, ANNs have shown to be more suitable to predict solar radiation than other empirical regression models (Fadare, 2009, Jiang, 2008). In the study, to train the model, data obtained by The Turkish State of Meteorological Service over 2006 years were used as training data and the values of 2005, 2007 and 2008 were used as testing data.

Section snippets

Artificial neural networks

The artificial neural networks (ANNs) are widely used in various fields of mathematics, engineering, meteorology, economics and in adaptive control and robotics, in electrical and thermal load predictions and many other subjects (Gokbulut et al., 2007, Senkal and Kaleli, 2009). ANNs are information processing systems which have the ability to learn, recall and generalize from training data. In order to perform predictions, ANNs need given examples instead of conventional equations (Lin & Lee,

Application of the ANN

ANN is used for the modeling of solar radiation in Turkey. The ANN model is trained using the solar radiation data from Antalya and Burdur, and then data from Isparta, Maras, Mersin, Adana and Antakya are used for testing the model. The block diagram of the ANN model is shown in Fig. 2. In the figure, the input variables are latitude, longitude, altitude, month of the year and mean cloudiness and the output variable is the solar radiation.

Three layers are employed in the model; those are input

Results and discussion

In the study, we made an analysis to predict the solar radiation for seven cities from the Mediterranean Region of Turkey as Antalya, Burdur, Isparta, K. Maras, Mersin, Adana and Antakya. These cities are presented on the Turkey map in Fig. 3. The geographical information (longitudes, latitude and altitude) for these cities are also listed in Table 1.

The essence of this study was to investigate the feasibility of using ANN to model the non-linear relationship between solar radiation and other

Conclusions

The ANN model is a powerful tool to predicting unmeasured values for any field of engineering. In this work, the ANN model seems promising for evaluating the solar resource potential in places where there are no monitoring stations in Mediterranean region of the Turkey.

The main conclusions, which may be drawn from the results of the present study, are listed as follows:

  • The obtained results in the present work indicated that the ANN based model for solar radiation was accurate for prediction of

References (36)

Cited by (148)

  • A novel procedure for photovoltaic fault forecasting

    2024, Electric Power Systems Research
  • Prediction of hourly solar radiation in Tamil Nadu using ANN model with different learning algorithms

    2022, Energy Reports
    Citation Excerpt :

    They used various activation functions in ANN model’s hidden layer. The article established the suggested models are appropriate for assessing radiation in Turkey [16]. Fadare et al. estimated solar radiation in 195 localities in Nigeria using different models that rely on feed forward as well as multi-layered network by incorporating certain weather [17].

View all citing articles on Scopus
View full text