Robustness and adaptability analysis for equivalent model of doubly fed induction generator wind farm using measured data. (1st March 2020)

• Record Type: Journal Article
• Authors: Zhang, Jian; Cui, Mingjian; He, Yigang
• Is Part Of: Applied energy. Volume 261(2020)
• Publication Date: 2020-03-01
• Subjects: Doubly fed induction generators (DFIG) wind farm -- Equivalent model of wind farm -- Trajectory sensitivity -- Parameters identification -- Genetic learning particle swarm optimization (GLPSO) hybrid algorithm; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2019.114362
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: The initialization method to avoid convergence problems is proposed. The impedance of the equivalent distribution network is identified. Parameters are identified by the genetic learning particle Swarm optimization algorithm. The convergence rate is much higher than the genetic and particle swarm algorithm. The precision of parameter identification results is much higher than previous work. Abstract: As many large wind farms connected to the power grid, it is necessary to develop a robust and adaptable dynamic equivalent model of the wind farm for system analysis and control. In this paper, the trajectory sensitivity of time-varying parameters of the equivalent model is analyzed. Then the non-time- varying parameters of the equivalent model are fixed as aggregated values, while the time-varying parameters are identified using the genetic learning particle swarm optimization based on phasor measurement unit data at the point of interconnection. The robustness and adaptability of the equivalent model under different scenarios are analyzed. The simulation results using the Western Electricity Coordinating Council benchmark test system show that the global searching capability to find the optimal point of the proposed method is higher than canonical particle swarm optimization and genetic algorithm by 2 orders. Further, the biggest mismatch between the identification results of the proposed method and the true values is within 10% for parameters with high sensitivity which is much better than previous work.