Abstract—One of the challenges in the photovoltaic (PV) system integration to the utility grid is its uncertainty which may affect the grid stability. Consequently, PV generation forecasting becomes one of the most important roles in the development of the grid-connected PV systems. This paper presents a PV generation forecasting using Artificial Neural Networks (ANN) with input variables and model parameters selection algorithm for a PV system located in Jangseong-gun, South Korea. It uses the ASHRAE Clear-Sky model to solve the unavailability of the irradiance data in this area. Additionally, the weather forecasting information is considered to compensate the uncertainty of the sky condition. Pearson Correlation Coefficient is employed to decide which weather data mostly influence the amount of solar irradiance. Cross-validation technique is used to select model parameters so that the model can fit to forecast PV generation in any conditions. As a result, it is shown that the model accuracy is improved compared to the model without input variables and model parameters selection.
Index Terms—ASHRAE clear-sky model, photovoltaic generation forecasting, artificial neural networks, input variable selection.
Fauzan Hanif Jufri and Jaesung Jung are with the Department of Energy Systems Research, Ajou University, Suwon, South Korea, 16499 (e-mail: email@example.com, firstname.lastname@example.org).
Cite: Fauzan Hanif Jufri and Jaesung Jung, "Photovoltaic Generation Forecasting Using Artificial Neural Networks Model with Input Variables and Model Parameters Selection Algorithm in Korea," International Journal of Machine Learning and Computing vol. 7, no. 5, pp. 156-161, 2017.