Abstract :
[en] The energy transition faces a bottleneck at the distribution level, where the rapid integration of distributed energy resources (DERs), such as solar photovoltaics (PVs), electric vehicles (EVs), and heat pumps (HPs), outpaces the capabilities of current distribution networks.
Low-voltage (LV) grids, historically designed as passive infrastructure, are now subjected to stress that exceeds their original design.
Consequently, distribution system operators (DSOs) must evolve into active managers, requiring new tools to operate the network efficiently.
This operational shift is also accelerated by rising regulatory pressure requiring coordination between DSOs and transmission system operators (TSOs) in managing the power system as a whole.
The DSO is therefore pressed from two directions. From below, DERs reshape LV networks that the DSO often cannot fully observe, whose existing capacity is poorly used, and whose access must now be shared among many active users. From above, regulation requires the DSO to coordinate with the TSO, so that the problem is no longer an optimization within its own network but a coordination problem between two operators.
This thesis addresses both. For the pressure from below, it proposes an operational toolkit built on three capabilities that build on one another: seeing the network, balancing its existing capacity, and exchanging access to it. A DSO cannot act on a network it cannot see; once the network is known, capacity wasted by an uneven loading of the three phases can be recovered; and the capacity that remains must then be shared among network users. For the pressure from above, it examines why TSO-DSO coordination, despite an extensive literature and numerous pilot projects, has almost never reached commercial deployment.
A DSO cannot act on a network it cannot see; once the network is known, capacity wasted by an uneven loading of the three phases can be recovered; and the capacity that remains must then be shared among network users. For the pressure from above, it examines why TSO-DSO coordination, despite an extensive literature and numerous pilot projects, has almost never reached commercial deployment.
First, network observability (*see*) is addressed through a topological path identification (TPI) methodology, which reconstructs the electrical path linking each customer to its feeder in the MV/LV substation. Using an integer linear programming (ILP) formulation, this approach assigns a physically feasible path to 88% of the customers of a real Belgian LV network from incomplete, static geographical data, without relying on advanced metering infrastructure (AMI). A robustness analysis, in which part of the customer-to-feeder records were deliberately corrupted, showed that the method avoids connecting more than 95% of the corrupted customers to a wrong feeder (fault detection rate).
Second, a phase reconfiguration (*balance*) algorithm is developed, which reassigns customers to different phases in order to reduce load unbalance, i.e., the unequal loading of the three phases of a feeder. Instead of relying on non-linear power-flow models, the method uses a linear optimization over year-long load curves, and balances the network without physical grid reinforcement. When applied to the topology of a real Belgian network under a synthetic high-DER scenario, the approach reduced load unbalance by 36% and 62% on the two test feeders and decreased total line losses by 4.4%.
Third, a continuous market mechanism (*exchange*) is introduced for dynamic operating envelopes (DOEs), i.e., time-varying limits on the power each customer is allowed to withdraw from or inject into the network. The framework allows network users to trade these limits among themselves, while a security check based on a linearized (DC) power-flow model guarantees that every accepted trade keeps the network within its operational limits. Results indicate that this limit-exchange mechanism has the potential to reduce renewable curtailment, incentivize the use of local flexibility, and extract additional value from existing infrastructure without requiring centralized control.
Finally, the thesis turns to the pressure from above and tests a common explanation for this lack of deployment: that coordination methods are not yet sophisticated enough. If the bottleneck were algorithmic, pilot projects, being the closest to deployment, would rely on more accurate models than purely theoretical studies. A multidimensional analysis comparing the literature with real-world pilots across five dimensions (modeling, operation, validation, communication, and regulation) shows that this is not the case: pilots rely on the same simplifying assumptions as theoretical works, while they differ markedly in validation, communication, and regulation. This suggests that the gaps blocking commercial deployment lie not in algorithmic complexity but in communication infrastructure, validation at scale, and regulatory alignment.
