Multiple Wind Power Time Series Modeling Method Considering Correlation

Abstract

As growing penetration of wind power integrated into power system, effective model is demanded to capture the characteristics of wind power not only in statistics but in time dependency and spatial dependency. This paper proposes a novel model that integrating pattern recognition and Markov Chain Monte Carlo (MCMC) method. In order to embody the correlated variation of daily wind power at different sites, typical scenarios are obtained by historical multiple wind power data and clustering algorithm. A single-variable MCMC model is then established to describe the scenarios transition process. Next, a multi-variable MCMC models are established to describe the correlation existed in the daily time series of multiple wind farms. The typical scenario Markov chain and daily wind power sequences for each typical scenario state are simulated successively and then generated a complete multiple wind power sequences. The effectiveness test shows that the wind power time series generated by the proposed models show higher accuracy on the statistical characteristic, autocorrelation and crosscorrelation, compared with Copula model.