A parallel power system linear model reduction method based on extended Krylov subspace

• Published in: International journal of electrical power & energy systems Vol. 160; p. 110072
• Main Authors: Du, Zhaobin, Zhou, Weixian, Chen, Zhiying, Zhou, Ziqin, Chen, Baixi
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.09.2024; Elsevier
• Subjects: Balanced truncation; Extended Krylov subspace; Parallel singular value decomposition; Power system model reduction; Balanced truncation; Parallel singular value decomposition; Extended Krylov subspace; Power system model reduction
• ISSN: 0142-0615
• DOI: 10.1016/j.ijepes.2024.110072

•Extend the balanced truncation method to the unstable system for power system model reduction.•Enhance the convergence of the extended Krylov subspace method by adjusting the input matrix.•Boost the computational efficiency by improved Bartels-Stewart method and parallelism techniques.•Reduce the overall time consumption of the balanced truncation method, while the reduced-order models still retain the dynamic characteristics of the original high-dimensional power systems. With the ever-increasing scale of power systems, stability analysis and control usually bear heavy storage and massive calculation burdens. In view of this, the model order reduction technique proves valuable by constructing a low-dimensional approximate model of the original system, which is crucial for efficiently handling large-scale systems. Balanced truncation (BT), a famous model reduction method, confronts practical limitations as it requires the systems to be stable and cannot deal with unstable models. Therefore, a parallel linear balanced truncation method for power systems based on extended Krylov subspace (EKS) is proposed in this work. Besides extending the BT method to unstable systems by α-shift, the key contribution also lies in strategies to enhance the convergence of the EKS method, whereupon the algorithm improvements include effective and efficient techniques for solving dual Lyapunov equations, and parallel acceleration of the singular value decomposition. The results of the simulation case verify that the proposed method can effectively improve the convergence of the EKS method by increasing α-shift, and the improvement work in this paper reduces the total time consumption of the BT method by about 26 %–33 % of the original, exhibiting better calculation efficiency. In addition, the case studies show that the simplified model still retains the time-domain and frequency-domain response characteristics of the original high-dimensional model.