This research proposes a new methodology to estimate parameters of synchronous generators based on trajectory sensitivity analysis. This new methodology was created to overcome convergence difficulties presented by the traditional trajectory sensitivity methodology due to: (i) Low robustness with relation to initial parameter values and noisy measurements; and (ii) singularities in the algebraic equation of the model of differential-algebraic equations (DAEs) that lead to the nonexistence of solutions, especially when the parameters are far from of the true values. Although the methodology has been developed to solve the synchronous generator problem, it is general and can be used for many types of nonlinear dynamic systems. Therefore, the main contribution of this thesis is the proposal of a new methodology based on trajectory sensitivity to estimate parameters of nonlinear dynamic systems with constraints, i.e., systems modeled by DAEs. More precisely, the equality constraint of the dynamic system is relaxed by an alternative formulation based on the minimization of the algebraic function of the model of DAEs. A second contribution of the thesis is related to the model of the generator. For this intention, the selection of the state variables, inputs and outputs is fundamental for the success of the parameter estimation methodology. In this thesis, this selection allows the generator's mechanical and electrical parameters be estimated independently. In order to estimate the electrical pa-x rameters, the generator is modeled by a set of convenient DAEs to fulfill the following practical requirements: (i) estimation of the generator parameters from the disturbance measurements obtained with the machine in operation; (ii) use of easily accessible measurements; and (iii) independence of the network parameter. As a final result, an algorithm which combines the new methodology of trajectory sensitivity to constrained systems with the two-stage estimation algorithm is proposed to estimate the generator parameters. This proposed methodology is robust for parameter's initial values and fulfills the practical requirements above mentioned. In addition, the estimation of the power angle is a byproduct of the proposed methodology.

No abstract

No abstract

No abstract

No abstract