Thèse de doctorat
Résumé : The transmission system operators perform a series of time-domain simulations to assess the stability of the power system towards disturbances. Depending on the results, they can decide to take some preventive actions on the real power system. The dynamic models in use cover beyond the electrical grid they are supervising because of the existing couplings with neighboring grids. On the other hand, the same level of detail cannot be maintained across the entire power system model due to a limit of computing power. The simulation times must namely stay in line with the reaction times necessary to operate an electrical grid. This thesis tackles the problem of computational complexity by proposing a systematic procedure for constructing reduced models that meet the industrial needs. In the first step, the power system model is partitioned into a study area and a multitude of clusters. The study area preserves a full mathematical description as it forms the part of the power system where the stability is examined. The clusters can instead be simplified as long as this does not alter the dynamic behavior of the study area. The decision to modify some specific clusters is taken in the second step by evaluating the nonlinearity level displayed following a disturbance initiated in the study area. If a cluster behaves like a linear system, it has a place inside the reducible external area. Otherwise, it belongs to the buffer area that is left unchanged along with the study area. The last step of the procedure replaces the clusters of the external area with linear equivalent models. For each of them, a simple and effective model of the surrounding network is developed in order to preserve a proper dynamic behavior while limiting the computing load of the reduction process. All the techniques composing the systematic procedure are validated on an academic system and a representation of the power system in Continental Europe. In both cases, the algorithms run in the same way and the reduced models are observed to be suitable for transient stability studies. It is concluded that this thesis provides the transmission system operators with a user-friendly solution to the problem of power system model reduction.