Performance analysis of three ANN models using improved fast evolutionary programming for power output prediction in grid-connected photovoltaic system / Puteri Nor Ashikin Megat Yunus, Shahril Irwan Sulaiman and Ahmad Maliki Omar

This paper presents an assessment of three ANN
models using hybrid Improved Fast Evolutionary Programming
IFEP-ANN techniques for solving single objective optimization
problem. In this study, multi-layer feed forward ANN models for
the prediction of the total AC power output from a grid-connected
PV system has been chosen. The three models were developed
based on different sets of ANN inputs. It utilizes solar radiation,
ambient temperature and module temperature as its inputs.
However, all three models utilize similar output, which is total AC
power produced from the grid-connected PV system.
The mixtures of Gaussian and Cauchy are used during the
mutation process in the EP technique. The best predictive model
was selected based on the lowest root mean square error (RMSE)
and higher regression, R. Besides, the comparison between
classical ANN (without evolutionary programming) and hybrid
IFEP-ANN was compared to determine which model performs
better for single-objective optimization. The IFEP-ANN models
showed the best in having the lowest RMSE and significantly
better than ANN in terms of highest regression, R.