Ali Ahmadian

Nowadays, the use of solar energy is very common with regard to the limitation of fossil fuels, environmental pollution standards and the advancement of solar panel technology. Due to the stochastic behavior of the Solar Irradiance (SI), the power of solar panels is uncertain. In order to improve the reliability of using solar energy in the power grid, the SI should be predicted with high precision. Artificial Neural Networks (ANNs) are very effective in predicting uncertainty phenomena. In this research, environmental data and previous values of SI are used to predict the SI. Radial Basis Function (RBF) ANNs are precise method because Gaussian functions are used in their activation function. In ANNs, connection between different layers is created through weights. In conventional ANNs initial weights are randomly determined and this can affect the performance of the ANNs. In this study, the initial weights are determined by meta-heuristic optimization algorithm such as Genetic Algorithm (GA). In this method, a GA is used during the training of the ANN and the SI is predicted over a week. To evaluate the proposed method, different error calculation methods are applied.