Covariances in the weather and the influence on an Australian large-scale renewable electricity system

© 2015 Dr. Robert Gordon HuvaSometime in the future Australia’s demand for electricity will undoubtedly need to rely on significant amounts of Renewable Electricity (RE). Fossil fuels are inherently non-renewable and their dwindling supply will force investment in alternatives. There is therefore a need to research how such increased penetrations of RE resources will be managed. Installed across large areas, a National Electricity Market (NEM) with increased contributions from RE will be affected by large-scale meteorological variability. The synoptic scale (100s-1,000s of km and days-weeks) weather systems are of particular interest. This thesis examines how synoptic scale weather variability might affect a future highly RE dependent Australia using two approaches. Approach one (Part 1 of the thesis) identifies the common synoptic scale weather systems from a reanalysis data set (ERA-Interim) and then analyses the availability of wind and solar associated with each weather type. Approach two (Part 2) utilises data from a regional model (ACCESS-R) as part of an electricity model that maximises the contribution of wind and solar electricity in meeting the demand of Australia. The influence of the weather systems identified in Part 1 on the optimised electricity network is then analysed. Part 1 of this thesis utilises the European Centre for Medium Range Weather Forecasting (ECMWF) reanalysis product ERA-Interim from 1989-2009 and for the Austral region (6S,105E) → (49.5S,165E). The Self-Organising Map (SOM) technique is then used to identify commonly occurring features in the ERA-Interim Mean Sea-Level Pressure (MSLP) field. The SOM technique converts the time series of MSLP to a time series of commonly occurring MSLP patterns, or weather types. The weather types are then analysed for their tendency to be associated with either high or low wind and solar potential. Some autumnal MSLP patterns are shown to be associated with very low wind speeds and solar irradiance for most of the Australian continent, while other summertime patterns show good potential for extracting wind and solar electricity. Following this an analysis of just the wind field demonstrates that decorrelation in the wind field is reached at a distance of approximately 1,300km. Part 2 of this thesis focuses on optimising renewable electricity capacity to meet electrical demand, using data from the Bureau of Meteorology’s high resolution regional weather model, ACCESS-R, for the period 2010-2011. The ACCESS-R data, in tandem with electrical demand data from the Australian Electricity Market Operator and a Genetic Algorithm, are utilised to investigate the influence of synoptic scale variability on a largely RE-based network. Wind and solar installations from locations across the ACCESS-R grid are optimally placed to maximise their contribution to meeting the electrical demand of 2010-2011. A gas-fired back-up system is deployed to cover moments when the combination of wind and solar cannot meet demand. The gas usage is made expensive to minimise its use, yet it is found that gas is still needed throughout the 2010-2011 period to cover moments when both the wind and solar are low. An investigation is then undertaken to determine any large-scale links to the minima in renewable generation. The common weather types identified in Part 1 are utilised in Part 2. By assigning SOM weather types to the 2010-2011 period it is shown that some regimes either adversely or favourably affect the net output of the optimised system. In particular, a late autumn and a summer weather type are shown to be significantly associated with very low RE output. The persistence and re-occurrence of low RE events shows that most episodes of low RE last for less than six hours and such low RE events have a mean return period of more than a week. Increasing the cost of transmission results in installed RE capacity contracting to four large wind stations and it is shown that the NEM region exhibits four distinct wind regimes. The wind regimes are highly uncorrelated and the minimum distance between regime locations reflects an approximate, but identical to Part 1, optimisation-based decorrelation length of 1,400km. In combination, Parts 1 and 2 of this thesis illustrate some of the issues that a future high penetration RE network might need to overcome. Knowledge of the influence of detrimental/favourable weather phenomena will be critical when designing and/or maintaining a large-scale renewable electricity network for Australia—in particular, knowledge of the decorrelation length-scale in the wind field