An adaptive excitation controller for synchronous generators: studies and experimental results at a power station

In order to guarantee a desired value for the electromechanical mode damping of the power station-network system in all the possible operating conditions, a prototype of excitation adaptive controller was developed so as to identify in an explicit way the dynamics of interest and on the basis of this explicit identification to modify the additional feedback gains from power and speed. The dynamics of interest is modelled in an adequate way with a time discrete model autoregressive and exogenous of the third order. To identify the parameters of the model, following the variations in the process dynamics, recursive least square algorithm forgetting factor was used. The system is excited by a persistent random input. This exciting signal must lead to small variations on the output variables, because these variations must be accepted for the plant operation. The estimated parameters are used to adapt the additional feedback gains in order to reach the desired damping value. These gains are computed minimizing a cost function, which represents the control effort. The deterioration of the estimation caused by a large external perturbation is avoided using an opportune blindness mechanism. The global results reached both in laboratory tests and in field trials were very positive.<>