cover
Contact Name
Eko Prianto
Contact Email
eko_prianto@uny.ac.id
Phone
-
Journal Mail Official
edukasielektro@uny.ac.id
Editorial Address
-
Location
Kab. sleman,
Daerah istimewa yogyakarta
INDONESIA
Jurnal Edukasi Elektro
ISSN : 25488252     EISSN : 25488260     DOI : -
Edukasi Elektro is a open access peer reviewed research journal that is published by Electrical Engineering Education Department - Faculty of Engineering - Yogyakarta State University. Edukasi Elektro is providing a platform for the researchers, academicians, professional, practitioners and students to impart and share knowledge in the form of high quality empirical and theoretical research papers, case studies, literature reviews and book reviews on education.
Arjuna Subject : -
Articles 161 Documents
Implementation of Neural Networks in Daily PV Power Output Prediction Using Bayesian Regularization Algorithms to Assist Energy Management Systems Mahmudah, Norma; Delfianti, Rezi; Sigit, Firman Matiinu; Putra, Dimas Panji Eka Jala; Nusyura, Fauzan
Jurnal Edukasi Elektro Vol. 9 No. 2 (2025): Jurnal Edukasi Elektro Volume 9, No. 2, November 2025
Publisher : DPTE FT UNY

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.21831/jee.v9i2.91044

Abstract

Solar power plants have several advantages, namely continuous energy production, reduced electricity demand, and low photovoltaic maintenance, so that PV power output can be optimized with reliable PV power output predictions. Implementation of Artificial Neural Network (ANN) to predict photovoltaic (PV) power output, using the Bayesian Regularization algorithm. Accurate PV power output prediction is very important in power systems. The data used are solar radiation, PV module temperature, ambient temperature, and actual PV power output, with the target being the PV power output for the next day with the PV power output output for the next day. The architecture used in this study is a Cascade Forward Neural Network (CFNN) and an Elman Neural Network (ENN). Both ANN models use daily data sets and performance evaluation using Mean Square Error (MSE). The results of the study show that ENN is more accurate than CFNN. ENN had the lowest MSE of 0.00664 at a configuration of N=8 and R of 0.9922 with a training time of 6.4 seconds, while CFNN recorded the lowest MSE of 0.024306 with N=25. ENN's ability to capture time series patterns in PV is more reliable and effective. Reliable predictions can assist in energy management systems because they help maintain supply balance, reduce the risk of failure, and improve system stability.