Automatic Determination of Network Openings in Meshed LV Grids

Abstract

The rapid growth of renewable energy integration, alongside the tightening of safety requirements for protection against electrical shock, is forcing distribution system operators to reassess low voltage grid planning and operation. In response, Stedin has adopted a long-term strategy to de-mesh all low voltage grids by 2040. This presents a significant challenge, as more than 50% of Stedin’s low voltage grids (excluding Zeeland) were still operated in a meshed configuration in 2025.

Meshed low voltage grids interconnect secondary substations and end-users through multiple paths, providing operational flexibility. However, these configurations complicate protection coordination and increase short-circuit current levels, resulting in more complex and widespread disturbances during fault conditions. In contrast, radial low voltage grids exhibit more predictable power flow behaviour and improved controllability, making them better suited to accommodate increasing electrification and distributed generation while complying with stricter safety requirements. Consequently, identifying permanent network opening points to achieve radial configurations is essential for developing reliable, compliant, and future-proof distribution systems.

System performance after de-meshing is sensitive to the placement of network openings, making accurate modelling of electrical interactions between interconnected secondary substations a prerequisite for informed decision-making. Existing medium voltage and low voltage grid models often rely on simplified assumptions that neglect critical interactions among multiple closed loops, potentially leading to sub-optimal opening decisions. To address this, an integrated medium- and low voltage grid model is developed in DIgSILENT PowerFactory that explicitly represents electrically relevant low voltage grid sections together with their supplying medium voltage feeders.

To support a systematic transition from meshed to radial topologies while balancing multiple performance criteria, this thesis proposes a simulation–optimisation framework for the automatic identification of feasible permanent network openings. The approach integrates a genetic algorithm implemented in Python with quasi-dynamic simulations within a closed-loop framework. This enables an efficient exploration of the highly combinatorial solution space, yielding technically feasible and near-optimal configurations that remain robust under diverse operating conditions.

Applied to a realistic case study, the framework reduces annual energy losses by 6.6% and lowers maximum feeder voltage drops by 4.07 V compared to a fuse-removal based industrial approach, while achieving performance comparable to heuristic cable-disconnecting methods. The heuristic initial search point strategy, used during initialisation to obtain efficient starting solutions and enhance convergence, provides a strong low-complexity baseline and achieves near-optimal results for the studied network. By enabling globally coordinated optimisation across interacting loops, the proposed framework shows promise for decision-making in large, complex meshed low voltage grids.