A Hybrid Algorithm for Synchronous Generator Parameter Estimation

This paper presents a hybrid algorithm for parameter estimation of synchronous generator. For large-residual problems (i.e., f(x) is large or f(x) is severely nonlinear), the performance of the Gauss-Newton method and Levenberg-Marquardt method is usually poor, and the slow convergence even causes iteration emergence divergence. The Quasi-Newton method can superlinearly converge, but it is not robust in the global stage of the iteration. Hybrid algorithm combining the two methods above is proved globally convergent with a high convergence speed through the example of synchronous generator parameter identification.