Paper Title
A Hybrid Artificial Neural Network And Particle Swarm Optimization (ANN-PSO) Algorithm To Predict Photovoltaic Power Plant Performance- A Case Study

Abstract
Photovoltaic and solar thermal are the main sources of electricity generation from solar energy. While the solar radiation that reaches the top of the layers of the atmosphere can be easily calculated, the solar irradiance that reaches the ground level where solar collectors (thermal and photovoltaic) operate depends strongly on localized and complex atmospheric conditions. In the first part of this study, ANN-PSO technique is applied to predict daily global solar radiation (GSR) on a horizontal surface, based on meteorological variables. Daily maximum air temperature, minimum air temperature, mean air temperature, maximum relative humidity, minimum relative humidity, mean relative humidity, maximum possible sunshine duration, sunshine hours, cloud cover, evaporation and wind speed data between 2000 and 2011 provided by Safiabad station, located in Dezful city, Iran (32� 16' N, 48� 25' E), are used as inputs to neural network and daily GSR used as output. The measured data between 2000 and 2008 are used to train the neural networks and the available data (732 days) from 2009 (January) to 2011 (April) are used for testing. Sensitivity Analysis is investigated to show the effect that each of the model inputs (i.e. meteorological variables) have on the model output (i.e. GSR). Eventually, In order to show the accuracy of the proposed method (i.e. MLP-PSO), a comparison is made with Backpropagation trained Multi-Layer Perceptron Neural Networks (MLP-BP) for same study. In the second part of the paper, monthly electricity generated for a 20 MW photovoltaic (PV) power plant is considered. Keywords- Global Solar Radiation (GSR), Multi-Layer Perceptron Neural Network (MLP), Particle Swarm Optimization (PSO), Error Backpropagation (BP) Algorithm, Meteorological Variables, Modeling, RETScreen Software.