The document discusses identifying reduced-order dynamic models of gas turbines. It compares several methods for parameter identification including deepest descent direction, BFGS direction, Gauss-Newton, and nonlinear least squares optimization. It evaluates these methods based on metrics like number of iterations, mean squared error of predictions, and autocorrelation of prediction errors. The best performing method was Gauss-Newton optimization, which accurately identified parameters for a first-order model with fewer iterations and lower prediction errors than other methods.