Wind power simulation using Correlated Innovation Matrix and Wavelet Multi-Resolution Analysis Approaches

Abstract

To meet carbon emissions targets, increased demand, and replace retiring plant it will be necessary to construct new electricity generation plant in New Zealand. One of the least cost, and carbon neutral methods of generating electricity is wind power. The intermittent and variable nature of wind coupled with the passive reaction of wind turbines ensures that wind power requires scheduled and spinning reserves. Reserve requirements can be alleviated, to an extent, by distributing wind farms throughout the country, exploiting the diversity of wind. However, spatial diversification may increase costs through reducing economies of scale in wind farm construction and increasing transmission requirements. Transmission expansion has long lead times, and needs good models of an uncertain future mix of generation size, type, and location. There is insufficient measured wind power or wind speed data to assess the trade-offs for envisaged wind development scenarios hence a model of wind power is required. The model must be temporally and spatially congruent with respect to wind, demand, and other generation types. In this paper specific models are developed applying wind speedtime-seriesderivedfromaNumericalWeatherPrediction model, temporal imputation methods, transformation to power using wind farm power curves, and assumptions concerning electricalandoperationalefficiencies.Thetemporalimputation, or turbulence modeling, is achieved through two methods: a Correlated Innovation Matrix approach, and a Wavelet MultiResolution Analysis approach. The models are used to simulate power time-series for seven wind farms, and subsets of these used to assess a centralised scenario and a diversified scenario. Results are compared with aggregate measured power timeseries demonstrating the benefit of spatial diversification and illustrating differences in the turbulence modeling approaches.