SynTiSe: A modified multi-regime MCMC approach for generation of wind power synthetic time series

Abstract

The Markov Chain Monte Carlo (MCMC) method is widely used for generation of synthetic wind power and wind speed time series, but its application for time resolutions of less than an hour often fails to replicate the autocorrelation (ACF) and probability density (PDF) functions of the original time series. This paper presents SynTiSe, an application software that allows fitting discrete-time, multi-regime MCMC models with percentile-based state space discretization. To illustrate its capabilities we use SynTiSe with a wind power dataset from ERCOT, and measure the quality of its simulations by comparing the ACF, the PDF, and the ramp characteristics of the input time series with those of the synthetic series. Results show that the 2nd order or higher multi-regime models with a percentile-based discretization of the state-space fitted by SynTiSe are a good alternative for the generation of synthetic time series of high resolution wind power data. These models improve the fit of the ACF, and greatly improve the representation of diurnal and seasonal patterns, while maintaining or slightly improving the fit of the PDF and ramp distribution. An executable package of SynTiSe for Windows platforms is publicly available.