Optimization of an Artificial Neural Network using Firefly Algorithm for modeling AC power from a photovoltaic system

Abstract

Solar electricity from Grid-Connected Photovoltaic (GCPV) systems has experienced rapid growth as one the most important renewable energy technologies globally. However, due to the intermittent and fluctuating weather conditions throughout the day and year, it is often a challenge to model the output performance of a GCPV system. This paper presents a model for modeling the AC power from a GCPV system using a hybridization of Firefly Algorithm (FA) with Multi-Layer Feedforward Neural Network (MLFNN). The AC Watt-output of a GCPV system was modeled using Artificial Neural Network (ANN) based on solar irradiance, ambient temperature and operating PV module temperature as the ANN inputs. These data were collected from a GCPV system located at Green Energy Research Centre (GERC), Universiti Teknologi MARA, Malaysia. On the other hand, FA was employed as a search tool for determining the optimal number of neurons in hidden layer, the learning rate and the momentum rate during training of the MLFNN. Upon completion of training, the ANN underwent testing process to validate the training process. In both training and testing, the modeling performance was quantified using Root Mean Square Error (RMSE). The performance of the FA-MLFNN was later compared with the performance of a Classical Evolutionary Programming (CEP)-MLFNN in modeling the AC power. The results showed that the hybrid FA-MLFNN had outperformed the hybrid CEP-MLFNN by producing lower RMSE during both training and testing.