Synchronous measurements and power quality monitoring.

• Main Author: Kuhlmann, Volker
• Language: en
• Published: University of Canterbury. Electrical and Electronic Engineering 2012
• Online Access: http://hdl.handle.net/10092/6158

The task of simultaneous power systems measurements at locations which can be hundreds of kilometres apart carries the problem of precise synchronisation. To introduce this topic, this thesis begins with an overview of time, time stamping and synchronised data acquisition, and reviews its application to power quality monitoring. It then discusses the requirements for such systems and how different application scenarios shift the emphasis between aspects of the requirements. The complexity of a power distribution monitoring system is most dependent on the number of channels and sites which must be analysed, and the required time stamping accuracy. For some applications, samples need to be time stamped with an accuracy of 1µs. A requirements specification template is presented which aids, for example, in purchase decisions to establish needed features. A specific example of such a system, CHART III, has been developed at the University of Canterbury (Christchurch, New Zealand), which uses a hardware time base and sample dock generation, hardware time stamping, and GPS synchronisation to achieve a time stamping accuracy of 0.5 µs. The design of the time base of this system is published in this thesis and described in detail. The CHART III system was used to gain practical experience and to establish its usability and operational limitations, and provided input for the theoretical considerations of an ideal system. Synchronised distributed data acquisition using two and three CHART III instrument was performed on two live power systems, collecting data in the frequency and time domains. A number of enhancements were made as a result, particularly to the control and analysis software in the areas of extending the handling of the GPS receiver and provision of additional system status and error information. Because the emphasis of the work is on the instrumentaton, no further analysis of the collected data is presented in this thesis. The CHART III system was connected to the internet to investigate issues of remote configuration and the consolidation of sample analysis at a single powerful computer. Limits for time domain measurements, which have a higher data rate than the system can handle continuously, were established as being a minimum of 10 seconds. In a data acquisition system, the quantisation error introduced by the ADC sets a lower limit for the noise. The effects of this quantisation noise on the recovery of harmonic magnitudes and phases were examined. Simulations were performed to model the influence of ADC width, fast Fourier transform length, harmonic amplitude and harmonic order. Both magnitude and phase errors are independent of harmonic order, decrease with the number of ADC quantisation levels, and decrease with the square root of the transform length. The magnitude error is independent of the harmonic amplitude for a sufficiently large amplitude to noise ratio. The phase error is inversely proportional to the amplitude. The accuracy of a harmonic analyser can therefore be increased by increasing the ADC width or the transform length. For accuracies likely to be required by typical power quality applications, these simulations indicate that a 12 bit ADC gives sufficiently accurate results. Finally, the effect of current trends in microprocessor technology is discussed. Power quality monitoring systems can now be built much more simply and cheaply then when CHART III was designed. The most important improvement is that a single standard CPU can now handle the data from a number of channels, eliminating the need for specialised digital signal processors and the associated cost of producing software for a second architecture. Processor performance seems to be set to increase steadily, promising future improvements in time stamping accuracy, the number of channels which can be handled by one processor, and the availability of more complex analysis functions.