| Summary: | 碩士 === 國立中正大學 === 電機工程所 === 94 === With the widespread use of nonlinear loads in the power system, the problems associated with harmonics are of great concern. Harmonics cause extra power losses in the transmission system and decrease the life of power devices.
The frequency-domain methods have been widely used for the signal processing because of its computational efficiency. In addition, most power meters adopt FFT-based algorithm to analyze the harmonics and to show the frequency spectra. However, the FFT-based algorithm is less accurate if the system frequency varies and the frequency resolution decreases. The analytic results will show errors caused by the leakage and picket-fence effects. Although increasing the number of sampling data can mitigate the undesired effects, this will impede the computational efficiency.
According to aforementioned facts, this thesis proposes a neural network based signal processing technique with Levenberg-Marquardt method for the weight update rule in the harmonic analysis. The proposed weight update rule is then compared with improved Widrow-Hoff delta rule. The Levengerg-Marquardt method estimates each frequency components in the measured signal. Not only is calculation time reduced, but also the result is with a better accuracy. Finally, the thesis applies Matlab/Simulink package software to simulate results that compare with those obtained by a conventional method to demonstrate the usefulness of the proposed method.
|
|---|
