A novel wind farm equivalent model for high voltage ride through analysis based on multi-view incremental transfer clustering. (February 2022)

• Record Type: Journal Article
• Authors: Han, Ji; Miao, Shihong; Chen, Zhe; Liu, Ziwen; Li, Yaowang; Yang, Weichen; Yin, Haoran; Zhang, Di
• Is Part Of: International journal of electrical power & energy systems. Volume 135(2022)
• Publication Date: 2022-02
• Subjects: WF model -- Multi-view incremental transfer fuzzy C means (MVIT-FCM) -- High voltage ride through (HVRT) -- Multiscale entropie (MSE); Electrical engineering -- Periodicals; Electric power systems -- Periodicals; Électrotechnique -- Périodiques; Réseaux électriques (Énergie) -- Périodiques; Electric power systems; Electrical engineering; Periodicals; 621.3
• DOI: 10.1016/j.ijepes.2021.107527
• Languages: English
• ISSNs: 0142-0615

Abstract

Abstract: The voltage swell (VS) in power system may require high voltage ride through (HVRT) of wind farms (WFs), and the detailed WF simulation model would need huge computational time, and thus is not be suitable for the analysis of HVRT dynamic behaviour of large-scale WFs. In this paper, a novel WF equivalent model with the required accuracy level for HVRT analysis is proposed. Firstly, multiscale entropies (MSEs) of the operational parameters in wind turbines (WTs) are calculated to represent their distinguishability in different HVRT processes, and the time series of several parameters which have obvious distinguishability are selected as the multi-view clustering indicators (CIs). To handle the multi-view CIs, a new clustering algorithm namely multi-view incremental transfer fuzzy C means (MVIT-FCM) is proposed. This algorithm integrates the transfer learning technique to increase the stability and accuracy of the WTs clustering. Also, the high-dimensionality of the time series based CIs and the consequent computational burden in clustering are considered, and an incremental technique is applied in MVIT-FCM to handle the large-scale WF modelling. A real WF system is used for case study. The results indicate that the multi-view CIs are very effective for increasing the equivalent accuracies. In addition, with the aid of incremental and transfer learning techniques, MVIT-FCM can acquire stable clustering results and handle large-scale WF accurately and efficiently.