Renewable Energy System Design by Artificial Neural Network Simulation Approach

An alternative approach for optimization of renewable energy systems using artificial neural network (ANN) models and simulation is applied for standalone wind energy and photovoltaic cell. Feed forward back propagation (FFBP) and radial basis functions (RBF) are considered. Large training data are obtained by simulations in Homer software package. Several input parameters and two output parameters (number of units and cost of energy) are considered for performance analysis of the ANN network models. Irrespective of the input parameters, FFBF model fails to yield results that match with target values both for WT and PVC systems. The R-square values are observed to be scattered and lie in the range of 35 to 90 percent. On the other hand, RBF model output with the same model training data consistently matches with Homer simulation output for all input conditions. The correlation coefficients as measured by R-square are well above 95 percent in all cases. In another words, very good generalizations of the input parameters are observed for RBF model.