Comparison of artificial intelligence methods in estimation of daily global solar radiation

Abstract

Abstract Assessment of solar potential over a location of interest is introduced as an important step for the successful planning of solar energy systems (photovoltaic or thermal). Due to the absence of meteorological stations and sophisticated solar sensors, solar data may be unavailable for every point of interest. Hence, empirical and intelligence methods are developed to estimate solar irradiance data. In this study, the idea of artificial intelligence methods is employed to predict the daily global solar radiation. The developed models are: group method of data handling (GMDH) type neural network, multilayer feed-forward neural network (MLFFNN), adaptive neuro-fuzzy inference system (ANFIS), ANFIS optimized with particle swarm optimization algorithm (ANFIS-PSO), ANFIS optimized with genetic algorithm (ANFIS-GA) and ANFIS optimized with ant colony (ANFIS-ACO). The data are collected from 12 stations in different climate zones of Iran. The input variables of the models are including month, day, average air temperature, maximum air temperature, minimum air temperature, air pressure, relative humidity, wind speed, top of atmosphere insolation, latitude and longitude. The results demonstrated that although the developed models can successfully predict the global horizontal irradiance, the GMDH model outperforms the other developed models. The values of root mean square error (RMSE), determination coefficient (R2) and mean square error (MSE) for the GMDH model were 0.2466 (kWh/m2/day), 0.9886 and 0.0608 (kWh/m2/day), respectively.