Analysis of a Solar Photovoltaic System Using Fractional-Order Neural Network

This paper implements fractional calculus (FC) to model solar photovoltaic fractional-order neural network (PV-FANN). Precisely, fractional derivatives like Caputo and Riemann-Liouville (RL) are executed in the backpropagation learning algorithms of multi-layer perceptron (MLP). The Log-sigmoid is used as an activation function, and a PV-FANN system is proposed. PV system is kept at different angles on sunny days to track maximum solar radiation. FANN models the dataset of this system. The performance of the proposed model is evaluated by parameters like training and testing mean square error (MSE) and training time for the fractional-order (FO) ( \(\alpha \) ) of fractional derivatives. The value of \(\alpha \) is varied from 0.1 to 0.9. The proposed algorithm is tested with different datasets and epochs to prove its efficiency. The proposed model outperforms conventional integer-order (IO) ANN by showing minimum MSE during training and testing at the cost of slightly increased training time.