Title:	Day-Ahead PV Generation Forecasting Based on Deep Learning Approach
Author(s):	D. Kothona, A. Zamanidou, I.P. Panapakidis, G.C. Christoforidis
Keywords:	Energy, Market, Power Forecast, PV, Long Short-Term Memory
Topic:	PV Systems – Modelling, Design, Operation and Performance
Subtopic:	Solar Resource and Forecasting
Event:	38th European Photovoltaic Solar Energy Conference and Exhibition
Session:	5BV.4.17
Pages:	1170 - 1175
ISBN:	3-936338-78-7
Paper DOI:	10.4229/EUPVSEC20212021-5BV.4.17
Price:	0,00 EUR
Document(s):	paper, poster

Abstract/Summary:

---

The development of robust photovoltaic (PV) power forecasting models is essential for the secure and economic integration of PV power into the intra-day and day-ahead energy markets. Since the day-ahead solar power forecasting is a more challenging procedure, at the present paper we employ a Deep Learning approach and specifically the Long-Short Term Memory (LSTM) model, due to its ability to capture long-term dependencies between the timeseries. The performance of the model is examined based on a selection of different inputs: a) historical data, b) weather predictions and c) historical data alongside with weather predictions. The results point out that the utilization of meteorological forecasts can increase the forecasting accuracy of the model from 5.3% to 29%.