부산대학교 도서관

The use of PV power plants has increased signifi-cantly in recent years. The use of PV power plants has issues in the power grid system, where the PV power plant cannot generate a stable output every day due to weather changes. To solve those issues, forecasting using deep learning models emerges as the solution. This study explains how to forecast the PV power output using historical weather data. The forecast period is 24-hours ahead and employs several deep learning models, such as RNN, LSTM, BiLSTM, ConvLSTM, and LSTM-BNN. The results show that forecasting the 24-hour PV power output using historical weather data as input for the deep learning model is possible and express promising results. Based on the implementation, the LSTM-BNN models outperform other models with a small MSE and MAE metrics value of 0.0082 and 0.0847 respectively.