Study on Numerical Fourier / Laplace Transformation and Inversion

カテゴリ: 研究会(論文単位)

論文No: HV20001

グループ名: 【B】電力・エネルギー部門 高電圧研究会

発行日: 2020/01/23

タイトル(英語): Study on Numerical Fourier / Laplace Transformation and Inversion

著者名: Lu Jia(University of Manitoba),Jiang Chen(University of Manitoba),Xue Haoyan(Global Energy International Development),Ametani Akihiro(University of Mantioba),Gole Aniruddha(University of Manitoba)

著者名(英語): Jia Lu(University of Manitoba),Chen Jiang(University of Manitoba),Haoyan Xue(Global Energy International Development),Akihiro Ametani(University of Mantioba),Aniruddha Gole(University of Manitoba)

キーワード: Fourier/Laplace transforms|inverse transform|reciprocal pair|integration method|sampling methld|Fourier/Laplace transforms|inverse transform|reciprocal pair|integration method|sampling methld

要約(日本語): This paper discusses the basic characteristic of Fourier / Laplace transform. Numerical reciprocity between forward Fourier transform and its inversion is confirmed for finite discrete transform. The accuracy of the transform is quite dependent on a integration method, and the odd number sampling is found to give a very high accuracy. A comparison of CPU times between various methods shows that an independently produced code of Fourier transform by C++ is much faster than that produced within a platform of a commercial software. However, FFT / IFFT functions prepared in the commercial software are very efficient from the viewpoint of the CPU times. It should be noted that the calculated results by the commercial software require adjustment of the results by a user. Laplace transform is more accurate and stable than Fourier transform especially when poles are along the imaginary axis of a complex domain.

要約(英語): This paper discusses the basic characteristic of Fourier / Laplace transform. Numerical reciprocity between forward Fourier transform and its inversion is confirmed for finite discrete transform. The accuracy of the transform is quite dependent on a integration method, and the odd number sampling is found to give a very high accuracy. A comparison of CPU times between various methods shows that an independently produced code of Fourier transform by C++ is much faster than that produced within a platform of a commercial software. However, FFT / IFFT functions prepared in the commercial software are very efficient from the viewpoint of the CPU times. It should be noted that the calculated results by the commercial software require adjustment of the results by a user. Laplace transform is more accurate and stable than Fourier transform especially when poles are along the imaginary axis of a complex domain.

原稿種別: 英語

PDFファイルサイズ: 1,626 Kバイト