Generating wind power time series based on its persistence and variation characteristics

Abstract

Generation of wind power time series is an important foundational task for assisting electric power system planning and making decision. By analyzing the characteristics of wind power persistence and variation, this paper proposes an improved Markov chain Monte Carlo (MCMC) method, identified as the PV-MC method, for the direct generation of a synthetic series of wind power output. On the basis of the MCMC method, duration time and variation features are concluded in PV-MC method, gaining a more comprehensive reflection of wind power characteristics in the generated wind power time series. First, the wind power state series is generated to meet the state transition matrix based on the definition of the wind power state. Then, the time duration of each state in the series is determined by its respective duration character. Finally, the variation characteristic is used to convert the state series to a wind power time series. A significant amount of simulations are performed based on the PV-MC and MCMC methods and are then compared for 25 wind farms at 6 different locations throughout the world. The simulation results show that the PV-MC method offers an excellent fit for the time domain features (persistence and variation characteristic) while holding other statistic features (mean value, variance, autocorrelation coefficient (ACC) and probability density function (PDF)) close to the MCMC method.