| Title |
Prediction of Photovoltaic Power Generation Based on LSTM Considering Daylight and Solar Radiation Data |
| Authors |
안연주(Yeon-Ju An) ; 이택기(Taeck-Kie Lee) ; 김규호(Kyu-Ho Kim) |
| DOI |
https://doi.org/10.5370/KIEE.2021.70.8.1096 |
| Keywords |
Photovoltaic Power Generation; Prediction; Long Short-Term Memory(LSTM); Deep Learning |
| Abstract |
This paper presents a method to predict the photovoltaic power generation using daylight and solar radiation data. Keras based long short-term memory(LSTM) model, a deep learning library, is used to predict the photovoltaic power generation and compared with a simple machine learning model. Based on the annual power generation, the weather parameters are selected with the highest correlation such as sunshine time and solar radiation. The prediction of Keras based LSTM model is superior to the prediction of the photovoltaic power generation using the simple machine learning model. This is because the probabilistic characteristics of actual variables are considered forecasting with actual weather parameters in the prediction of photovoltaic power generation. |