Abbas Ahmadi - Department of Industrial Engineering & Management Systems, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran.
Mahsa Zaman - Department of Economics, Kharazmi University, Tehran, Iran.
Siab Mamipour - Department of Economics, Kharazmi University, Tehran, Iran.

Clean solar energy is one of the best sources of energy. Solar power plants can generate electricity in Iran due to their large number of sunny days. This paper presents a short-term forecasting approach based on artificial neural networks (ANNs) for selected solar power plants in Iran and ranks the input variables of the neural network according to their importance. Two solar power plants in Hamadan province (Amirkabir and Khalij-Fars) were selected for the project. The output of solar power plants is dependent on weather conditions. Solar radiation on the horizontal plane, air temperature, air pressure, day length, number of sunny hours, cloudiness, and airborne dust particles are considered input variables in this study to predict solar power plant output. Forecasting model selection is based on considering zero and nonzero quantities of target variables. The results show that solar production forecasting based on meteorological parameters in the Khalij-Fars is more accurate than Amirkabir. The global solar radiation, air temperature, number of sunny hours, day length, airborne dust particles, cloudiness, air pressure, and dummy variables[۱] are the order of the most important inputs to solar power generation. Results show simultaneous influences of radiation and temperature on solar power plant production.   [۱]. The first half of the year is counted as one, and the second half is counted as zero.

Solar power plants, metrological parameters, Artificial Neural Networks, Iran