Modelling and Topology Optimisation of Medium Voltage Representative Networks for The Netherlands

Abstract

The electrification of modern-day society keeps increasing and the demand for electricity grows along with it. Technologies like EVs and heat pumps are both part of the new types of demand, while PV systems and wind farms are meant to be our main new sources of generation. Contrary to traditional resources connected to the electricity network, most of these resources tend to be placed and operated in a distributed manner, increasing the demand for transport capacity within the LV and MV grid. This demand cannot always be met however, which leads to situations of congestion.

This thesis seeks to reduce this congestion by means of optimising the grid topology in line with the seasonal variations in the supply and demand of electricity. This is done by means of implementing a two-stage reconfiguration algorithm. The benefit of network reconfiguration is that it is a short-term and low-cost solution which can be implemented by the DSO without relying on other external parties.

In the first stage of the reconfiguration algorithm, the positions of the normally open switches within the network are optimised in order to adjust the power flow therein. These optimised positions are subsequently used in the second stage to calculate the network variables. The first stage is implemented as a MILP optimisation in Python, while the second stage consists of a Newton-Raphson calculation in the commercial software PowerFactory. This distinction is made to enhance the accuracy of the final network parameters, while still being able to optimise the switch positions in a deterministic manner.

Rather than day-ahead or real-time reconfiguration, as is often considered in most literature, this thesis focuses on seasonal reconfiguration to accommodate for the manual operation of the switches present within the MV grid in the Netherlands. These manually operated switches severely reduce the frequency by which reconfiguration actions can be performed, but that does not mean that the topology of the grid cannot be enhanced.

The presented reconfiguration algorithm is able to consistently reduce congestion within the analysed network, completely removing it or reducing its severity. It also outperforms two optimisation options in PowerFactory with regards to the objective function value. Those being an iterative exploration of meshes and a genetic algorithm.