CNN-LSTMとアンサンブル学習を用いた電力スポット価格予測

カテゴリ: 全国大会

論文No: 6-095

グループ名: 【全国大会】令和6年電気学会全国大会論文集

発行日: 2024/03/01

タイトル(英語): Electricity Spot Price Forecasting Using a CNN-LSTM Ensemble Learning Method

著者名: 王子洋（東京大学）,前匡鴻（東京大学）,松橋隆治（東京大学）

著者名(英語): Ziyang Wang (The University of Tokyo),Masahiro Mae (The University of Tokyo),Ryuji Matsuhashi (The University of Tokyo)

キーワード: 電力スポット価格予測|再生可能エネルギー|畳み込みニューラルネットワーク長・短期記憶ニューラルネットワーク|深層学習|アンサンブル学習|Electricity spot price forecasting|Renewable energy source (RES)|CNN-LSTM|Deep learning|Ensemble learning

要約(日本語): This paper presents a novel day-ahead electricity spot price forecasting framework for Japan's Kyushu region, focusing on periods with high renewable energy source (RES) production. It introduces a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model, enhanced with an ensemble learning approach for improved robustness. The model uniquely incorporates a range of inputs including actual and forecasted electricity metrics, spot prices, meteorological, and calendar data. This multimodal approach significantly boosts forecasting accuracy. Tested throughout the year 2022, the model demonstrates superior performance, achieving high levels of precision in its predictions, as evidenced by substantial improvements in key performance indicators of RMSE, MAE, and R2.

本誌掲載ページ: 153-154 p

原稿種別: 日本語

PDFファイルサイズ: 243 Kバイト