Time series simulation for multiple wind farms based on HMMs and regular vine copulas

Abstract

The randomness and fluctuation of wind power cause difficulties in electric utilities operation and dispatch. Although the methods for simulating wind power time series have been studied by researchers for decades, methods designed for the single wind farm output or the aggregated output of multiple wind farms are inefficient in capturing the spatial dependence among regional multiple wind farms. Also, abnormal or missing data are quite common in the wind power field, which is often ignored by traditional multi-dimensional correlation modeling methods, resulting in being imprecise and unstable encountering the situation of missing data. In this paper, the discrete hidden Markov model (HMM) and regular vine copulas are applied to accurately reproduce the joint distribution of regional wind farms. The regular vine copulas can take various bivariate copulas as blocks to precisely and flexibly describe the dependence structures of multiple wind farm outputs, and using HMM, we were able to model transition probability for different dependence structures. Enough synthetic time series of multiple wind farms can be generated by the proposed method, and it is of great significance for the long-term scheduling of power systems. The effectiveness of the proposed method is tested using the datasets of five wind farms in Northwest China as a case. The simulation results prove that the proposed method can accurately model the dependence structures among multiple wind farms and statistical characteristics of power outputs, and are more robust when there are missing data in wind power records especially.