Page 521 - AWSAR 2.0

developed for the PQ monitors. A standard and commonly known method is discrete Fourier transform (DFT) for the stationary data (data in which signal properties do not change with time). In transforming the signal from time domain to frequency domain, DFT does not provide the time information at which a PQ disturbance has occurred. The application of DFT fails when the signal contains non- stationary features such as sudden changes, trends, drift, etc. The widely used transforms for non-stationary data, such as short-time Fourier transform (STFT), wavelet transform (WT), Stockwell transform (ST) and Hilbert Huang transform, allow obtaining time-frequency information from the PQ disturbance signals. Among several DSP techniques reported in the literature, ST has been a widely suggested technique. The ST is the extension of either continuous WT with phase-modified mother wavelet or STFT with a variable Gaussian window. To identify the type of
PQ disturbances with the aid of the features (PQ indices) extracted from decomposed PQ measurement data, commonly employed classification algorithms are artificial neural network, fuzzy logic, rule-based expert system, support vector machine, Bayesian classifier and kernel machines.
While formulating the
research objectives for my
Ph.D. work, I came to know
about several research gaps
that have not yet been explored
and need investigation: (1) a
three-phase approach for the detection of PQ disturbances is missed as most of the PQ analysis in the literature has been presented phase by phase separately. (2) Most of the PQ detection techniques have been applied for the single-stage disturbances. Therefore,
Mr. Rajat Kumar || 497
the efforts for multi-stage and multiple disturbance detection and classification are needed. (3) Most of the classification methods deal with the type of PQ disturbance without specifying the underlying cause. Therefore, the work can be extended towards cause- based classification instead of phenomenon- basedclassificationforabetterunderstanding of the PQ disturbances. Several underlying causes of PQ disturbances, such as faults, starting of induction motor, energizing of capacitor bank, energizing of transformer, etc., have been investigated during my Ph.D. programme. An effective and intelligent ST- based algorithm with reduced computations has been developed, which focuses on not only the classification based on the types of PQ disturbances (for example, sag, swell and flicker) but also the underlying causes of PQ disturbances. Another computationally efficient ST-based online PQ monitoring algorithm to
assess the multiple and multi- stage PQ disturbances and their respective underlying causes has also been developed. The performance of the aforementioned method has been experimentally validated using the hardware prototype developed in the laboratory as shown in Fig. 1. The experimental prototype comprises voltage sensors (LEM make LV 25-P) and current sensors (LEM make LA 55-P) for the measurement of bus voltage and current waveforms, respectively. A
compact DAQ chassis of national instruments (NI-cDAQ 9178) make has been utilized along with the NI-9215 input module for acquiring the PQ disturbances at a very high speed. Till today, what I have observed is that the major challenging task in recognizing the type of
An effective and intelligent ST-based algorithm with reduced computations has been developed, which focuses on not only the classification based on the types of PQ disturbances (for example, sag, swell and flicker) but also the underlying causes of PQ disturbances.