Polyspectral signal analysis techniques for interharmonics in shipboard power systems
MetadataShow full item record
In this dissertation, we present the theory and application of polyspectral signal analysis techniques for interharmonics in shipboard power systems. Interharmonics are generated from various kinds of adjustable speed drives (ASD) in such power systems. ASDs are highly nonlinear devices due to the use of rectifiers and inverters. Since interharmonics can seriously hamper the normal operation of electric ships in many different ways (e.g., excitation of undesirable electrical and/or mechanical resonances, misoperation of control devices, and light flicker), the detection and analysis of interharmonic-related events is a critical issue in assessing power quality in an all-electric ship. Standard signal analysis techniques for regular harmonics are not immediately applicable to interharmonics due to their small amplitude and uncertain frequency of occurrence. Hence, we propose the use of alternative polyspectral analysis techniques such as higher-order spectra (the cross bispectrum/bicoherence) for the detection and analysis of the ASD-generated interharmonics. First, we develop the interharmonic application specific definitions of the cross bispectrum and the cross bicoherence. The statistical characteristics and frequency domain symmetries are also investigated. We apply the modified cross bispectrum to interharmonic detection problems. Due to their small amplitudes, the detection of interharmonics is sensitive to many undesirable factors such as spectral leakage and measurement error. Our analysis results demonstrate that the detection performance of the conventional DFT-based method is seriously degraded in the presence of noise. Hence, we develop a constant false alarm rate (CFAR) interharmonic detector based on the modified cross bispectrum. Our analysis and experimental results show that our method can provide more robust detection performance than conventional methods in the presence of noise. We also develop an ASD condition monitoring method based on the cross bicoherence. The key idea is to diagnose the status of the load side of an ASD from observations made at the source side. In this dissertation, we apply our method to detection and analysis of phase imbalance at the load side of the ASD. Our experimental results demonstrate that the proposed method provides a unique interharmonic signature for detection and classification of asymmetric impedance associated with the phase imbalance. Furthermore, the proposed method shows a more sensitive detection performance compared to the conventional imbalance measurement method, which enables prognosis of potential faults. A novel quadratic phase coupling detector for a single data record with coherent interharmonics is developed. The traditional bicoherence definition fails when its ’phase randomization’ assumption is not satisfied. This assumption is not appropriate for certain applications such as continuous monitoring of rotating machines. Therefore, we propose a novel quadratic phase coupling detector and compare it with previous techniques. It is shown that our detector is superior to previous detectors at high SNRs, and can also address partially coherent cases which previous approaches could not properly address. Flicker issues related to interharmonics are also discussed. We present a newly found limitation of the current IEC flickermeter regarding detecting flicker caused by low frequency interharmonics. We also present observation results of flicker responses of various lamps including light-emitting-diode (LED) lamps. Our observation results confirm that compact fluorescent and LED lamps are sensitive to high frequency interharmonics, although the IEC flickmeter can not detect flicker caused by such interharmonics. Hence, we develop an alternative flicker detection method based on down-up sampling. Our experiment results show that our method can detect flicker regardless of the value of the interharmonic frequencies. Independent of interharmonic topics, we also present our additional achievement involving application of wavelet denoising techniques to network congestion monitoring problems. This was a collaboration with researchers at the Department of Computer Sciences in the University of Texas at Austin, and mainly completed before becoming engaged in the electric ship project. By applying wavelet techniques, we could drastically enhance shared congestion detection performance over previously proposed methods.