Gaussian Process Regression for Dynamic Mode Decomposition

カテゴリ: 部門大会

論文No: P33

グループ名: 【B】平成30年電気学会電力・エネルギー部門大会

発行日: 2018/09/12

タイトル(英語): Gaussian Process Regression for Dynamic Mode Decomposition

著者名: 升田 明利(大阪府立大学),薄 良彦(大阪府立大学),Satomi Sugaya(大阪府立大学),Manel Martinez-Ramon(Central New Mexico Community College),石亀 篤司(The University of New Mexico),Andrea Mammoli(大阪府立大学)

著者名(英語): Akitoshi Masuda|Yoshihiko Susuki| Satomi Sugaya|Manel Martinez-Ramon|Atsushi Ishigame|Andrea Mammoli

キーワード: クープマンモード分解,Koopman Mode Decomposition,Gaussian Process,Dynamic Mode Decomposition

要約(日本語): Koopman Mode Decomposition (KMD) is a novel technique of nonlinear time-series analysis based on spectral properties of the Koopman operator for underlying dynamical systems. The main feature of KMD is dynamics-oriented and thus applicable to data-driven technologies of analysis and control of power system dynamics, which have been reported by the second author. Dynamic Mode Decomposition (DMD) is a well-known algorithm of numerical computation of KMD and provides a finite approximation of the Koopman operator from a finite number of time-series data via a regression technique. In this report, we propose to use the Gaussian process regression, which is a Bayesian machine learning approach, in DMD and will present its potential application to power system data analytics.

PDFファイルサイズ: 119 Kバイト