Advanced digital signal processing techniques for electricity distribution grid monitoring and diagnostics

Abstract

The continuous monitoring of the power grid voltage parameters, namely amplitude, frequency and phase is essential for the implementation of key power system functions such as control and protection, load shedding and restoration, assessment of power quality and impact of distributed generation to mention just few. Such monitoring requires modern digital signal processing (DSP) techniques which should be computationally efficient, accurate, fast and robust against any grid disturbances irrespective of their nature and type. Efficient low bandwidth DSP techniques for voltage monitoring will contribute to an economical realization of the smart grid vision for smart meters, phasor measurement units, and power quality analyser. The thesis contributes to the previously mentioned topic. First, it offers a comprehensive overview of the grid voltage parameters estimation techniques reported thus far in the technical literature. These techniques are the discrete Fourier Transform (DFT), Kalman filter (KF), least-squares (LS) and phase-locked loop (PLL). There are however limitations associated with these techniques. For instance, the DFT one requires periodic waveforms being sampled. While the KF and the LS offer better performance under grid disturbances, they are computationally demanding techniques for real-time implementation. The PLL is a simple technique to implement in real-time but requires a trade-off between good dynamic performance and estimation accuracy under distorted grid conditions. The thesis attempts to overcome the above mentioned limitations and proposes and documents the mathematical basis, simulation and experimental implementation of a number of DSP techniques for single-phase grid voltage fundamental frequency and/or amplitude estimation. First, the grid voltage fundamental frequency is tracked by four proposed techniques which are based on a Teager energy operator, a Newton-type algorithm and a differentiation filter, a voltage modulation, and a demodulation and a differentiation filter. In addition, the thesis reports five techniques to estimate both fundamental voltage amplitude and frequency based on a modified demodulation, the DFT and a second-order generalized integrator (DFT-SOGI), a SOGI and a differentiation filter (SOGI-DF), a cascaded delayed signal cancellation (CDSC), and a linear KF. The performance of the above nine techniques has been evaluated through simulation and experimentally. A comparison with selected competing techniques reported in the technical literature is also included for a number of cases. These cases cover at all times harmonics being present in the grid voltage under steady-state and in combination with frequency sweep, frequency step, voltage sag and voltage flicker. The results presented in the thesis confirm that the proposed techniques provide improved performance with respect to the existing techniques for the cases studied as mentioned above. All the proposed techniques can reject the negative effects caused by harmonics and can also meet the accuracy and dynamics as specified by the standard requirements. The techniques based on the voltage modulation, demodulation and differentiation filter, modified demodulation, DFT-SOGI, and SOGI-DF provide improved stability, since they do not include an interdependent loop for the feedback of the estimated fundamental frequency. On the other hand, the techniques relying on the Teager energy operator, Newton-type algorithm and differentiation filter, modified demodulation, SOGI-DF, and CDSC are relatively computationally efficient for being implemented in real-time digital signal processors. The thesis concludes with a benchmark comparison of all the proposed techniques in terms of a number of key criteria such as accuracy, dynamic performance, stability, simplicity and computational efficiency. The comparison provides a proven way for choosing a suitable technique for single-phase grid voltage fundamental amplitude and/or frequency estimation. Finally, research directions for conducting future research are discussed.