Short-Term Energy Yield Prediction of Dust Accumulated Standard, Half-Cut and Quarter-Cut PV Modules Using Multi-layer Neural Network Algorithm

Abstract

AbstractIn this paper, the implementation of a multi-layer artificial neural network (ANN) algorithm is discussed to predict short-term photo-generated power of dust accumulated mono-crystalline PV modules variants; PV modules with full, half-cut, and quarter-cut solar cells. The datum to train neural networks was obtained from real-time results. Experimental observations were performed under standard irradiance conditions, i.e., 1000 \({\text{W/m}}^{{2}}\) over consecutive three months. To have targeted output, 70% of the data set is utilized in the training of neural networks. Comparison of neural network predicted results and experimental outcomes is discussed based on calculations of root mean square error (RMSE) as well as coefficient of determination \(R^{2}\). The performance indices account RMSE = 0.17631 and \(R^{2}\) = 0.99923 for standard cell PV modules, RMSE = 0.83192 and \(R^{2}\) = 0.99969 for half-cut technology and RMSE = 0.90929 and \(R^{2}\) = 0.99971 for the quarter-cut solar cell-based PV module which dictates the efficacy of the ANN algorithm in the power estimation of PV modules.KeywordsStandard solar cell PV moduleHalf-cut PV moduleQuarter-cut PV moduleANNEnergy prediction