講演抄録/キーワード |
講演名 |
2021-05-27 09:25
LSTM-based Neural Network Model for Predicting Solar Power Generation ○Kundjanasith Thonglek・Kohei Ichikawa(NAIST)・Kazufumi Yuasa・Tadatoshi Babasaki(NTT-F) EE2021-2 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
Currently, the most popular renewable energy is solar power which reduces pollution consequences from using conventional fossil fuels. Solar power converts sunlight either directly or indirectly into electricity. However, using solar power generation as a stable power supply is still challenging since the amount of solar power generated in a day is difficult to be predicted. Accurately predicting solar power generation enables controlling the amount of stored electricity in batteries to produce stable electricity. This paper aims to improve controlling the amount of stored electricity in batteries by predicting future solar power generation. We designed and implemented a neural network model based on Long Short-Term Memory (LSTM) to predict the future solar power generation using the past solar power generation and weather forecasts. Moreover, stratified K-fold cross-validation is applied to eliminate learning deviation during the training process. Through hyperparameter tuning, we have built a neural network model with one LSTM layer. As a result, the proposed model has achieved an R2 score of around 0.78 with cross-validation. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Time-Series Forecasting / Long Short-Term Memory / Solar Power Generation / / / / / |
文献情報 |
信学技報, vol. 121, no. 40, EE2021-2, pp. 7-12, 2021年5月. |
資料番号 |
EE2021-2 |
発行日 |
2021-05-20 (EE) |
ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
EE2021-2 |
|