| Title |
Prediction of a Floating Photovoltaic Generation Utilizing KMA Weather Forecast |
| Authors |
권오극(Ogeuk Kwon) ; 홍현표(Hyunpyo Hong) ; 조현식(Hyunsik Jo) ; 차한주(Hanju Cha) |
| DOI |
https://doi.org/10.5370/KIEE.2023.72.8.904 |
| Keywords |
Recurrent Neural Network; Power Generation Prediction; Floating Photovoltaic; LSTM; AI |
| Abstract |
This paper proposes a method for predicting the hourly generation of floating photovoltaic by creating a model and inputting weather forecast data. In particular, the paper shows the process of selecting hyperparameter settings that minimize errors, which has not been presented in other papers. The prediction model uses the Long Short-Term Memory, a type of recurrent neural network in machine learning, to predict the hourly generation of floating photovoltaic in Chungju and Hapcheon. Instead of using all the 12 existing data, the model only inputs the irradiation and temperature data that are highly correlated with power generation. The prediction model is then used to forecast the hourly power generation for Chungju and Hapcheon. The results show that the predicted power generation meets the error rate of 5% or less, as set by the Korea Power Exchange, making the machine learning-based prediction of floating solar photovoltaic practical and applicable. |