Quantifying the benefits from the spatial diversification of wind power in New Zealand.

Abstract

Wind power is one of the least cost forms of electricity generation, which along with the need to reduce carbon emissions, means that wind power capacity will certainly increase. The turbulent nature of wind and the passive reaction of wind turbines ensures that wind power is variable. The variability in power increases the requirement for system reserves to ensure power quality is maintained. Integration studies are undertaken to determine what impact wind power development will have on the power system and an often reached conclusion is that spatial diversification can alleviate some of the impacts. However, many of these studies fall short of quantifying the benefit of spatial diversification. To quantify that benefit requires models of wind power that are spatially and temporally consistent and congruent with other forms of generation and demand. In this thesis a wind power model is formed, starting with wind speed time-series from the European Centre for Medium-range Weather Forecasting reanalysis which are interpolated, scaled and imputed. The imputation requires a model of turbulence and a Wavelet Multi-resolution Analysis model is developed that accounts for the heteroskedasticity of wind while enforcing the correct temporal and spatial correlations. The wind speed time-series are transformed to power using wind power plant power curves. A Low Pass Filter is developed that accounts for the effect of spatial integration performed by Wind Power Plants. To demonstrate the benefit of spatial diversification in the New Zealand power system four scenarios are developed representing 2 GW wind power portfolios. The scenarios are Compact, Disperse, Diverse, and Business As Usual (BAU). Metrics for reliability, variability, and predictability are defined that reflect the structure of the New Zealand Electricity Market. Reliability is assessed using the standard deviation of power. Variability is assessed using ramp rates over a 5 minute period which equates with the window used for reserves scheduling. Predictability is assessed using persistence forecast errors over 2 hour horizons which equates with the gate closure in the New Zealand Electricity Market. The conclusion is that a compact wind generation portfolio will exhibit lower reliability, a diverse portfolio will exhibit less variability, and a disperse portfolio will exhibit greater predictability. The BAU scenario shows that the existing portfolio of Wind Power Plants in New Zealand achieves some of the benefits of spatial diversification, however greater benefit could be achieved through careful planning. This thesis forms part of Research Aim 1.1.1 of the GREEN Grid project.