-
- Chogumaira Evans N.
- Graduate School of Science and Technology, Kumamoto University
-
- Hiyama Takashi
- Graduate School of Science and Technology, Kumamoto University
この論文をさがす
抄録
This paper presents short-term electricity load forecasting using dynamic neural networks, DNN. The proposed approach includes an assessment of the DNN's stability to ascertain continued reliability. A comparative study between three different neural network architectures, which include feedforward, Elman and the radial basis neural networks, is performed. The performance and stability of each DNN is evaluated using actual hourly load data. Stability for each of the three different networks is determined through Eigen values analysis. The neural networks weights are dynamically adapted to meet the performance and stability requirements. A new approach for adapting radial basis function (RBF) neural network weights is also proposed. Evaluation of the networks is done in terms of forecasting error, stability and the effort required in training a particular network. The results show that DNN based on the radial basis neural network architecture performs much better than the rest. Eigen value analysis also shows that the radial basis based DNN is more stable making it very reliable as the input varies.
収録刊行物
-
- 電気学会論文誌B(電力・エネルギー部門誌)
-
電気学会論文誌B(電力・エネルギー部門誌) 131 (2), 181-186, 2011
一般社団法人 電気学会
- Tweet
詳細情報 詳細情報について
-
- CRID
- 1390282679579842176
-
- NII論文ID
- 10027803382
-
- NII書誌ID
- AN10136334
-
- ISSN
- 13488147
- 03854213
-
- NDL書誌ID
- 10952168
-
- 本文言語コード
- en
-
- データソース種別
-
- JaLC
- NDL
- Crossref
- CiNii Articles
-
- 抄録ライセンスフラグ
- 使用不可