Abstract
The inherent time-varying nature of the wind farm power causes undesired voltage flicker in the power network. In order to mitigate the flicker to enhance the performance of the wind power system with very fast dynamics, the static VAr compensator (SVC) is utilized. However, the SVC operates with some delays which negatively affects its performance. This persuades us to predict the reactive power of the wind farm to compensate for the real-world delay. The predicted reactive power is then utilized in the SVC. Therefore, this paper develops a novel fuzzy one-step-ahead prediction approach for the wind farm reactive power. The proposed fuzzy prediction uses a Takagi-Sugeno (TS) fuzzy representation whose unknown parameters are tuned online based on an extended Kalman filter (EKF). The wind farm is modeled as a time-varying current source which its amplitude and phase change every 0.01 s. A large set of the actual data of a wind farm in Manjil, Iran is gathered and directly utilized in the simulation process. Several flicker indices are calculated to evaluate the proposed prediction method. The obtained results show the performance enhancement and flicker mitigation of the suggested power scheme.
Highlights
In recent years, with the emerging and advance of renewable energy sources technology, renewable energy sources are considered as a suit able replacement for the conventional fossil-based energies [1,2,3]
In order to mitigate the flicker to enhance the performance of the wind power system with very fast dynamics, the static VAr compensator (SVC) is utilized
The proposed fuzzy prediction uses a Takagi-Sugeno (TS) fuzzy representation whose unknown parameters are tuned online based on an extended Kalman filter (EKF)
Summary
With the emerging and advance of renewable energy sources technology, renewable energy sources are considered as a suit able replacement for the conventional fossil-based energies [1,2,3]. In [15] the impact of the time delay on the performance of SVCs connected to Manjil wind farm has been studied and ARMA models have been used to predict the reactive power for the half cycle. II) The behavior of reactive power variations of wind farms is nonlinear and the linear ARMA model is not proper. It is expected that by combining the PSO or gradient descent with EKF, the overall modeling accuracy is improved Both approaches consider an offline procedure to partially update the fuzzy model parameters or a fully online procedure with online nonlinear optimization to completely update the parameters.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call