Biography
Emeritus Professor Gosbell was a cadet engineer with Sydney County
Council while an undergraduate student. He obtained his Ph.D. in 1971
from the University of Sydney with work on the asynchronous operation
of turbogenerators.
In 1972 he commenced lecturing at the University of Sydney where
his research interests included model power systems, power system
stability, HVDC transmission, power electronics and variable speed
motor drives. In 1990 he moved to the University of Wollongong where
he became foundation Professor of Power Engineering. His current
research interest is power quality with an emphasis on harmonics, PQ
survey measurements, and standards.
He is a member of the Standards Australia "Power Quality" Committee,
a Fellow of the Institution of Engineers, Australia and past
Chairperson of the Australasian Committee for Power Engineering. He
was the recipient of the M.A. Sargent Medal in 2008.
Research Interests
Harmonic Behaviour of Power System Loads There is an increasing use of power electronics in lighting,
electric motor drives and many other areas for greater efficiency and
accuracy of control. These circuits all involve the use of switching
electronics and draw non-sinusoidal current from the supply,
distorting the supply voltage waveform. This project will determine
means to estimate the harmonic current drawn by non-linear loads under
different conditions. Development of harmonic connection agreements for MV and LV customers The harmonic voltage levels in a power system are a combination of
the effects from all large distorting loads. When the level of
harmonic voltage distortion is excessive, there needs to be clear
procedures for identifying which load is the major cause so that
appropriate mitigation can be taken. Techniques need to be developed
to determine a fair and reasonable allocation for each load as a basis
for connection agreements between utility and customer. Present
harmonic standards contain procedures which are applicable when all
customers are connected at the transformer output. This is inadequate
for MV and LV customers who are connected part-way along feeders and
distributors which have significant changes in fault level along their
length. The work aims to develop methods for allowing for customers
connected at points of varying fault level using models requiring data
which is easy to obtain. Detecting harmonic sources When harmonic voltages exceed maximum acceptable limits, customers
are not entirely responsible for their level of distorting current.
Situations can arise in which customers who are normally compliant find
that their current exceeds what has been specified in a connection
agreement. Practicable methods need to be developed which will allow a
clear demonstration as to which customer, if any, is exceeding their
allowance. Methods being investigated include harmonic power and
reactive power flow and observing the natural daily variation of
harmonics voltage and current levels. Various analysis and simulation
tools are being investigated for their suitability for this project. Power quality surveying methodologies Power system companies are being required to given undertakings
regarding the level of PQ disturbances at different voltage levels
within their network. These undertakings can only be given with
confidence if the utility has a process for "building" the PQ limits
into the network at the planning stage. Measurement of PQ levels at a
sample of sites is required to prove that the planning processes are
adequate. PQ monitors, a communications network and suitable PQ
database and reporting facilities are very expensive. Research is
required to determine the optimum placement of PQ monitors so that the
best use is made of a limited budget. A starting point for this is the
development of methods of estimating sources of PQ disturbances and
their propagation. Power quality reporting The data from a single PQ monitor is very large. For example,
consider harmonics which require 3 readings for each of 39 harmonics
and the THD every 10 minutes. Over a year, assuming 4 bytes/reading,
gives about 25MB/site. Detailed examination of these readings for
insights would be tedious if each 10 minute reading has to be
considered manually. Instead there need to be summary values
determined which can be assessed quickly and not overlook important
details. Research is being applied to developing summary figures
applied to the original data in several steps. A framework has been
proposed in which the data is considered to be in several layers, with
each layer being a summary of the one underneath it. Hence, if one
summary figure is excessive, the network planner can be guided into
looking at the detail in the next level down. Key concepts developed
for the discrete disturbances are the severity index for characterizing
a multi-parameter discrete event, and the combination of severity
indices into a disturbance index. For continuous disturbances,
multi-parameter disturbances such as harmonics and flicker give rise
to a single parameter via the two step process of normalization and
consolidation. Having obtained a single index for each disturbance
type, they can be combined into a single Unified PQ Index for a site. Power quality data analysis The mass of data gathered for a sample of sites for a utility
survey has the potential to reveal the good and bad influences on
power quality if an appropriate diagnostic procedure can be
determined. Several functions have been identified and it now requires
the details of analysis procedures to be developed.
- The comparison of disturbance levels with reduced targets in the
case of lightly loaded networks allows network problems to be found
that would otherwise not be revealed until the network is loaded.
- Analysis of long term trends in PQ levels might reveal
potential problems in time for inexpensive corrective action to
be taken.
- Factor analysis enables the factors which contribute to good
and bad PQ levels to be identified and considered in PQ planning.
- Data mining allows insights to be found without having to
postulate the types of input/output models required for factor analysis.
|
