Abstract
This paper aims to enhance the performance of a cascade-forward neural network (CFNN) model to predict the output power of a photovoltaic (PV) module. This improvement is conducted by optimizing the number of hidden neurons using the genetic algorithm (GA). The optimization is carried out to minimize the value of the root mean square error (RMSE) between the actual and predicted PV output power. The performance of the CFNN-based GA is evaluated using five statistical term error terms; namely, RMSE, normalized RMSE (nRMSE), mean bias error (MBE), normalized MBE (nMBE), and mean absolute percentage error (MAPE). The values for these terms are 0.0025 W, 0.0131%, 0.0011 W, 0.0067% and 0.8121%, respectively. The proposed model’s performance is compared with that obtained using the classical CFNN, classical feed-forward neural network (FFNN), and FFNN-based on GA in accuracy training and testing time. The results show that the proposed model performs better than the models mentioned above based on the statistical error terms. The MAPE of the proposed model is 0.8121%, which is lower by 98.63%, 98.77% and 63.93% than that obtained by FFNF, CFNN, and FFNN-GA, respectively. Besides, it is faster than the models mentioned above based on training and testing time. In which, the proposed model is faster than FFNF, CFNN, and FFNN-GA by 89.68%, 59.50% and 53.99% in terms of training time, respectively. Moreover, it is also faster in testing than FFNN, CFNN, and FFNN-GA by 94.92%, 75.57% and 78.95%, respectively. Accordingly, the proposed model can be recommended to predict the output power of the PV modules.