Parameters identification of twelve-phase synchronous generator based on Levenberg-Marquardt algorithm

In terms of Gauss-Newton (G-N) nonlinear least-squares algorithm, if J(x) is rank-deficient, then either the Gauss-Newton method cannot work well, or the algorithm converges to a non-stationary point. To overcome the difficulty, in this paper we consider employing trust-region technique, and propose Levenberg-Marquardt (L-M) algorithm. In L-M algorithm, damping term mu_kI is added to G-N algorithm iterative method, i.e. L(x_k) = G(x_k) + mu_kI. This algorithm was applied to parameter identification of twelve-phase synchronous generator, and a better simulation identification result has been achieved, thus proving the validity of identification algorithm.