Dynamic load models are fundamental components for simulating and analyzing a power system. Due to their utmost importance, a wide range of dynamic models have been introduced within the past years, able to accurately capture the dynamic behavior of conventional loads. However, new grid technologies, such as distributed generation (DG) and controllable loads, are gradually emerging, transforming the purely load-composed distribution networks to active distribution networks (ADNs). Therefore, it is vital for Transmission System Operators (TSOs) to upgrade the existing load models and reassess the stability of their system.
Motivated by study conducted in Germany several years ago, this thesis attempts to determine the steady-state relationship between the frequency and active power using real measurement data. As a next step, the respective methodology for developing load models, which was proposed several years ago, is further extended so that it can be applied to modern distribution networks with DG. The results of the “old” and the current study are compared revealing how much the load dynamics have changed within the past years.
The thesis continues with introducing a three-stage methodology to effectively build a set of dynamic models for an ADN based on field measurements. In the first stage, the proposed clustering method identifies and then discards all the irrelevant data for the model parameter estimation. In the next stage, the remaining data are clustered into groups with similar dynamics while in the third stage, a nonlinear dynamic model is developed for each of the derived groups. It is concluded that, in spite of the large number of measurements representing a wide range of grid configurations, the general dynamic characteristics of an ADN can be accurately captured using a limited number of models.
Finally, this thesis aims at modeling the uncertainty induced by the stochastic nature of load and DG. To do so, a new probabilistic dynamic model is proposed that, except for its response, additionally yields the corresponding predictive uncertainty. This probabilistic model is further enhanced so that it can directly incorporate time and weather variables. It is shown that this model can decode the influence of this kind of exogenous variables and convert it into more accurate predictions.