Short-Term PV Energy Generation Forecasting using Deep Learning

Main Article Content

Noman Shabbir
Roya Ahmadiahangar
Argo Rosin
Victor Astapov
Jako Kilter


ence-to-sequence regression is performed in each iteration. Solar photovoltaic (PV) energy
generation has witnessed exponential growth in the last few years due to increasing energy
costs from fossil fuels and environmental factors. However, this PV generation energy has
intermittent characteristics due to seasonal variations and time. The future electrical grids
require stability and flexibility for smooth operation. Therefore, accurate PV energy generation
forecasting can be beneficial in this regard. In this paper, deep learning (DL) based recurrent
neural network (RNN) has been proposed for short-term PV energy generation forecasting. The
PV energy generation data was measured in rural Estonia for a 600-kW power plant for the
whole year with a one-hour time step. The weather data and the historical PV generation data
have been used in the development of this forecasting algorithm. The long short-term memory
network (RNN-LSTM) has been used for a 24-hours ahead of forecasting. The comparative
analysis of these forecasting gives a value of root mean square error (RMSE) of around 9 kW.

Article Details