Prediction of Daily Global Solar Radiation by Daily Temperatures and Artificial Neural Networks in Different Climates

Abstract

Introduction Global solar radiation is the sum of direct, diffuse, and reflected solar radiation. Weather forecasts, agricultural practices, and solar equipment development are three major fields that need proper information about solar radiation. Furthermore, sun in regarded as a huge source of renewable and clean energy which can be used in numerous applications to get rid of environmental impacts of non-renewable fossil fuels. Therefore, easy and fast estimation of daily global solar radiation would play an effective role is these affairs. Materials and Methods This study aimed at predicting the daily global solar radiation by means of artificial neural network (ANN) method, based on easy-to-gain weather data i.e. daily mean, minimum and maximum temperatures. Having a variety of climates with long-term valid weather data, Washington State, located at the northwestern part of USA was chosen for this purpose. It has a total number of 19 weather stations to cover all the State climates. First, a station with the largest number of valid historical weather data (Lind) was chosen to develop, validate, and test different ANN models. Three training algorithms i.e. Levenberg – Marquardt (LM), Scaled Conjugate Gradient (SCG), and Bayesian regularization (BR) were tested in one and two hidden layer networks each with up to 20 neurons to derive six best architectures. R, RMSE, MAPE, and scatter plots were considered to evaluate each network in all steps. In order to investigate the generalizability of the best six models, they were tested in other Washington State weather stations. The most accurate and general models was evaluated in an Iran sample weather station which was chosen to be Mashhad. Results and Discussion The variation of MSE for the three training functions in one hidden layer models for Lind station indicated that SCG converged weights and biases in shorter time than LM, and LM did that faster than BR. It means that SCG provided the fastest performance. However, the story for accuracies was different i.e. the BR, LM, and SCG algorithms provided the most accurate performances, respectively, both among one or two hidden layers. According to the evaluation criteria, six most accurate derived models out of 1260 tested ones for Lind station was 3-14-1 and 3-11-19-1 with LM, 3-20-1 and 3-20-19-1 with BR, and 3-9-1 and 3-20-17-1 with SCG training algorithm, and 3-20-19-1 topology with BR showed the best performance out of all architectures. Results of the evaluation of the six accurate models in the remaining 18 stations of Washington State proved that regardless of the climate, in each weather station, BR with its inherent automatic regularization, provided the most accurate models (0.87 67.41 %), and then SCG (0.90>R>0.83, 3.91>RMSEMAPE > 77.28 %). Therefore, the Bayesian neural networks, which showed the best performance among all Washington State weather stations, were evaluated for Mashhad station, as an Iran sample climate. The results proved the ability of the said networks for this climate (R=0.82, RMSE=3.92‌ MJm-2, MAPE=79.92%). Conclusions The results indicated that the Bayesian neural networks are capable of predicting global solar radiation with minimum inputs in different climates. This was concluded both in Washington State weather stations, which has a variety of climates, and also in Mashhad as an Iran sample weather station. These models would eliminate the need for complex climate-dependent mathematical relations or other models which are mostly dependent on many inputs. So, this algorithm would be a good means first in weather forecast practices, also in the design and development of solar assisted equipment, as well as in managerial practices in agriculture when monitoring crop solar-dependent processes like photosynthesis and evapotranspiration.

Keywords

• bayesian regularization
• daily temperature
• global solar radiation
• neural network