Hybridization of Meta-Evolutionary Programming and Artificial Neural Network for predicting grid-connected photovoltaic system output

Shahril Irwan Sulaiman, Khairul Safuan Muhammad, Ismail Musirin, Sulaiman Shaari
2013 IEEE 2013 Tencon - Spring  
This paper presents the evolutionary neural networks for the prediction of energy output from a gridconnected photovoltaic (GCPV) system. Two evolutionary neural network (ENN) models have been proposed using evolutionary programming and artificial immune system (AIS) respectively. The artificial neural network (ANN) employed for these models utilized solar radiation and ambient temperature as its input whereas the kilowatt-hour energy of the GCPV system is the only targeted output. The
more » ... of ANN involves the search of the optimal number of nodes, the learning rate, the momentum rate, the transfer function and the learning algorithm of a singlehidden layer multi-layer feedforward ANN. The results showed that evolutionary programming-ANN (EPANN) outperformed artificial immune system-ANN (AISANN) in terms of correlation coefficient, R as well as computation time. In addition, EPANN had also produced better convergence of the evolving parameters compared to the AISANN. Key-Words: -artificial neural network (ANN), multi-layer feedforward neural network (MLFNN), photovoltaic (PV), grid-connected photovoltaic system (GCPV), correlation coefficient (R), evolutionary programming (EP), artificial immune system (AIS) and prediction.
doi:10.1109/tenconspring.2013.6584486 fatcat:uzmzn7wlsvdwxkkaf5rmy46fca