The Examination of Appropriate Learning Periods to Improve the Prediction Accuracy of Solar Power Generation with a LSTM Model

The purpose of this study is to develop a PV power generation machine learning model using the LSTM model, verify its accuracy, and derive an appropriate learning period to ensure the appropriateness of the solar power generation prediction model.A comparison of short-term power generation data predicted by the machine learning model with data measured at the corresponding forecast time showed R 2 = 0.9499, indicating sufficient accuracy.Finally, the minimum learning period to secure a certain level of prediction through a review of the prediction result was reviewed using the power generation data already secured and the weather information at the same point during the period; 30 days were calculated as an appropriate learning period to ensure the accuracy of the solar power generation prediction model.