Modeling Correlated Power Time Series of Multiple Wind Farms Based on Hidden Markov Model

Generating long-term correlated wind power time series for multiple wind farms is of great significance for power system planning and operation. A new method on modeling correlated power time series of multiple wind farms was proposed based on hidden Markov model (HMM). A Markov chain was adopted to model the state of time-varying correlations between wind farms, and wind power outputs at two adjacent moments were set as observations of HMM, which established the mathematical mapping model between wind power correlations and power outputs at two adjacent moments. The Baum Welch algorithm was used to estimate the parameters of HMM, which consists of the transition probability matrix and the joint probability distribution of observations. Based on the established HMM, Monte Carlo simulation method was used to generate correlated annual wind power time series of the wind farms. In case studies, the proposed method was tested on three wind farms in the northwest of China. Results show that the generated wind power time series exhibit superior performance on the annual and monthly characteristics, joint probability distribution and auto-correlation to the compared method, and the cross- correlation is also very familiar to the historical data, which verify the effectiveness of the proposed method. © 2019 Chin. Soc. for Elec. Eng.