Study on the Effect of the Training Period on the Accuracy of Insolation Forecasts with Artificial Neural Networks

カテゴリ: 部門大会

論文No: 151

グループ名: 【B】平成22年電気学会電力・エネルギー部門大会講演論文集

発行日: 2010/09/01

タイトル(英語): Study on the Effect of the Training Period on the Accuracy of Insolation Forecasts with Artificial Neural Networks

著者名: Joao Gari da Silva Fonseca Junior (産業技術総合研究所),大関 崇(産業技術総合研究所),高島 工(産業技術総合研究所),荻本 和彦(東京大学)

著者名(英語): Joao Gari da Silva Fonseca　Junior (AIST National Institute of Advanced Industrial Science and Technology),Takashi Oozeki(AIST National Institute of Advanced Industrial Science and Technology),Takumi Takashima(AIST National Institute of Advanced Industrial Science and Technology),Kazuhiko Ogimoto(Tokyo University, Institute of Industrial Science(IIS))

キーワード: 日射量予測|太陽光発電|人工の神経回路網|メソ数値予報モデル|時系列予測|Insolation Forecasting|Photovoltaic System|Artificial Neural Networks|Meso-scale Model|Time-series Forecasting

要約(日本語): One of the main problems to the dissemination of PV systems is the instability in their produced power. One way to approach this problem is with insolation forecasts, which can be done with multilayer artificial neural networks (ANN). Nevertheless, the ability that such ANNs have to provide good accuracy forecasts depends, among several factors, of the amount of data used in their training period. The objective of this study is to investigate the effect of the training period on 1 day insolation forecasts done with ANN. The training period varied from 1 day to 1 year, and a cloudy day and a clear sky day were forecast. The results showed that there are meaningful differences in the forecast accuracy according to the kind of day and training period. The study also shows in which conditions to increase the training period is effective to improve forecast accuracy.

PDFファイルサイズ: 2,594 Kバイト