Reduced-order modelling of active distribution networks for large-disturbance simulations

Distribution systems are getting more and more complex owing to the increasing number of
Inverter-Based Generators (IBGs) connected at Medium-Voltage (MV) level. This makes distribution
networks more and more responsive and their influence on the whole power system dynamics
increases. Therefore, it becomes important for Transmission System Operators (TSOs) to model
those Active Distribution Networks (ADNs) in their dynamic studies.
First, this thesis deals with the derivation of reduced-order, “grey-box” models of ADNs, intended
for dynamic simulations of the transmission system. The ADNs are assumed to host IBGs as well
as static and motor loads, whose dynamic parameters are affected by uncertainty. This latter issue
is addressed using Monte-Carlo simulations. The parameters of the equivalents are adjusted to
match as closely as possible the average of the randomized responses, while their dispersion is
accounted for through the weights of the weighted least-square minimization. A procedure is used
to remove from the identification the parameters with negligible impact. To avoid over-fitting, the
equivalents are tuned for multiple large-disturbance simulations. A recursive procedure is used to
select the smallest possible subset of disturbances involved in the least-square minimization.
Next, the methodology is extended to account for changing operating conditions. This consists of
testing the accuracy of a set of previously derived equivalents, and updating the best of them if
its accuracy is not satisfactory. In order to update the equivalent with minimal effort, an approach
minimizes the number of parameters to update. In most cases, the results coincide with expectations
coming from “engineering judgment”, involving the adjustment of a (very) small subset of
parameters.
Finally, a new application for a Battery Energy Storage System (BESS), connected at distribution
level, is proposed. Its active and reactive powers are controlled such that the net power entering
the distribution network matches the response of the available equivalent to large disturbances in
the transmission system. The response of the equivalent is simulated in real time. This would
allow using with a higher guarantee of accuracy the equivalent model in dynamic simulations of
the transmission system. The BESS is supposed to be connected at the main substation of the
distribution grid and its control does not use any model of that grid. All simulations reported in the thesis have been carried out on three ADN test systems with different characteristics, one of them being derived from an existing distribution grid. The tests involve large disturbance scenarios that trigger nonlinear, discontinuous responses of IBGs and loads.