On Loss-Minimal Radial Topologies in MV Systems

Anton Hinneck; Mathias Duckheim; Michael Metzger; Stefan Niessen

REVIEW 4 major objections 5 minor 16 references

On Loss-Minimal Radial Topologies in MV Systems

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Adding acyclicity constraints to an exact AC DSR model makes loss-minimal radial reconfiguration solve drastically faster than the common relaxed-radiality benchmark.

desk verdict Clean formulation and honest limitations, but the speedup claim outruns the evidence: five warm-started heuristic runs with no optimality gaps, and case 5 contradicts the abstract. read the letter →

arxiv 2506.03422 v2 pith:5GGJFJ5K submitted 2025-06-03 eess.SY cs.SY

Anton Hinneck , Mathias Duckheim , Michael Metzger , Stefan Niessen This is my paper

classification eess.SYcs.SY

keywords distributionsystemreconfigurationradialityconstraintscycleinequalitiesACoptimalpowerflowmixed-integernonlinearprogrammingspanningtreeslossminimizationmedium-voltagenetworks

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes a mixed-integer nonlinear formulation for distribution system reconfiguration (DSR) under exact AC power flow, in which radiality is enforced by requiring that every cycle of the grid graph lose at least one edge. It compares this formulation with a common relaxed-radiality benchmark that only fixes the number of energized lines. Across five benchmark medium-voltage cases, the cycle-constrained formulation reaches better loss-minimizing topologies in a fraction of the wall-clock time, while the relaxed benchmark frequently runs into a 900-second time limit. The paper also proves that only edges lying on some cycle ever need to be switchable, and that a cycle basis suffices to identify them. If the comparison holds up, adding acyclicity constraints is a cheap way to make an NP-hard switching problem tractable on real MV networks.

What carries the argument

The load-bearing object is the set of cycle inequalities in Model C-DSR: for every cycle $C_k$ of the grid graph, $\sum_{(f,t)\in E(C_k)} z_{ft} \le |E(C_k)|-1$, meaning each cycle has at least one open switch. Together with $\sum_{(f,t)\in E} z_{ft} = |V|-1$, these constraints characterize spanning trees exactly through condition 2 of Theorem 1, whereas the benchmark RR-DSR only uses the edge count and can admit disconnected forests unless the system data forces connectedness. The supporting corollaries cut the integer-variable count: a line needs a switching variable only if it lies on a cycle, and the set of all cycle edges can be built from any cycle basis rather than by enumerating all cycles.

What would settle it

Run C-DSR and RR-DSR on the same five benchmark cases with the same 900-second limit but without a warm start, and with at least one additional MINLP solver; if the relaxed formulation then solves at comparable speed or the complete formulation loses its advantage, the paper's conclusion that the acyclicity constraints themselves improve solver performance is overturned.

Watch

Extended reading notes

Core claim

The central claim is that the set of cycle inequalities (6b), one for each cycle asserting that at least one of its edges is out of service, combined with the |V|-1 edge count (6c), characterizes spanning trees and, when added to an exact AC DSR model, dramatically tightens the search space without sacrificing accuracy. In the reported experiments, the complete formulation C-DSR finds loss-minimal radial topologies with lower or equal losses in a fraction of the wall-clock time of the relaxed benchmark RR-DSR, which omits the cycle inequalities and frequently exceeds the 900-second time limit. A supporting theoretical result shows that the switchable line set can be restricted to cycle edges, and that the union of all cycle edges is obtained from any cycle basis, so full enumeration of all cycles is not needed. The paper interprets the speedup as coming from the tighter search space of the acyclicity-constrained model.

Load-bearing premise

The central speed-up claim is measured with one solver, a 900-second time limit, and the baseline topology supplied as a warm start; without a warm start several cases could not be solved at all, so the reported advantage of the acyclicity constraints may depend on that specific setup rather than on the constraints alone.

Editorial extensions

If this is right

Using the cycle-constrained formulation, AC-accurate DSR becomes solvable in tens of seconds on the tested medium-voltage networks, where the relaxed formulation frequently hits the time limit without improving on the baseline.
Loss reductions from reconfiguration exceed 30 percent on several of the tested rural and commercial MV cases, with no line ratings violated and with zero voltage violations on most cases.
Only cycle edges need switching variables, and a cycle basis suffices to identify them, so the model size depends on the graph's cyclomatic structure rather than on its total line count.
Because the formulation only needs a spanning tree to exist, it is independent of system data that would otherwise be required to enforce connectedness, unlike the relaxed benchmark.
On the largest test case the solver still needs about 600 seconds for a solution, showing that the formulation alone does not remove the need for faster heuristics on larger networks.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The integer-variable reduction suggests that difficulty scales with the cyclomatic number $|E|-|V|+1$ rather than raw line count; adding extra meshes to a fixed feeder graph should make the problem harder, which is a testable prediction.
Different choices of cycle basis will produce different individual constraints even though all bases identify the same switchable edge set, so basis selection with short or sparse cycles may tighten the relaxation further than the paper's implementation.
The same spanning-tree characterization should transfer to other switching problems that require radial operating states, such as outage restoration and intentional islanding, where the cycle inequalities could play the same tightening role.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know two things about arXiv:2506.03422. First, the formulation work is clean: C-DSR adds all-cycle acyclicity constraints (6b) to an exact AC MINLP, and the two corollaries that let you restrict switching to cycle-edges are elementary but correctly argued. Second, the empirical claim in the abstract—that acyclicity constraints significantly improve solver performance—is not yet supported by the data in the paper.

What's actually new is modest. The cycle-based radiality constraint appears in earlier MILP work [8]–[10]; the contribution here is applying it to exact AC power flow and benchmarking on five Simbench instances. That's a legitimate incremental step, and I appreciate that the paper is upfront about its limitations. It explicitly says that without MIP starts several models could not be solved, and it relaxes voltage and loading limits and reports violations afterward.

The soft spots are the ones the authors don't fully advertise. All runs are warm-started from the baseline topology with Juniper, a heuristic solver, and there are no optimality gaps or dual bounds reported. So the speedup in cases 1–4 (13–27 s vs 900 s) may be an interaction between the warm start and the extra constraints rather than a general property of the formulation. More damning, case 5—the largest—shows no speedup: both models time out near 900 s. The abstract says 'significantly improve solver performance' without this caveat. Also, because several reported solutions violate the relaxed voltage or loading limits, comparing objective values between models is not apples-to-apples.

The citation pattern is fine. The base model [11] is cited as a starting point, not as self-supporting evidence, and the graph theory references are standard. The corollaries are correct.

Who should read this? Researchers working on DSR formulations and anyone benchmarking MINLP solvers for distribution networks. The paper gives a clear, reproducible methodology for adding cycle constraints to an AC model, but the speedup conclusion needs tighter benchmarking: report lower bounds or optimality gaps, test cold starts, and run at least one independent solver.

I would send it to peer review—the question is well-posed and the formulation is useful—but I'd expect major revision on the empirical section. I wouldn't cite the speedup claim in my own work until that evidence exists.

Referee Report

4 major / 5 minor

Summary. The paper proposes C-DSR, an AC mixed-integer nonlinear program for distribution system reconfiguration that enforces radiality by adding acyclicity constraints (6b) over all cycles of the network graph together with the spanning-tree edge count (6c), and compares it against RR-DSR, a benchmark formulation that omits (6b). The authors also present two corollaries intended to reduce the number of switchable lines to cycle-edges, computable from a cycle basis rather than by full cycle enumeration. Experiments on five SimBench cases are run with the Juniper heuristic under a 900 s time limit and with the baseline topology as a warm start; voltage and line-loading limits are removed from the model and their violations are evaluated ex post. The paper reports lower computation times and lower losses for C-DSR in four cases and concludes that the additional acyclicity constraints significantly improve solver performance.

Significance. If the computational claim were established, the paper would be a useful practical contribution: an exact-AC DSR formulation with explicit all-cycle radiality constraints is natural, and the cycle-basis reduction of switchable variables is attractive for small MV networks. The paper deserves credit for using an exact AC power flow, validating the implementation against PowerModels.jl, building on open benchmark data, and honestly reporting the scaling failure on the largest case. However, the central claim is not yet supported: the evidence consists of warm-started heuristic wall-clock times without optimality gaps, the comparison is partly between completed C-DSR runs and timed-out RR-DSR runs, and several reported solutions violate the relaxed voltage limits. The theoretical proof of Corollary 2 is also incomplete. With additional benchmarking and a corrected proof, the contribution could become solid, but as it stands the manuscript needs substantial revision.

major comments (4)

[Section III, Table I] The load-bearing claim that constraints (6b) 'significantly improve solver performance' rests entirely on heuristic wall-clock times. The authors state that without MIP starts several models could not be solved due to convergence issues, and Juniper is a heuristic. Therefore the observed speedups may reflect an interaction between the baseline-topology warm start and the extra constraints rather than a general property of the formulation. To support the claim, report dual bounds or optimality gaps for both models, or repeat the comparison under cold starts and with an independent solver; without such evidence, the speedup is not established as a formulation effect.
[Table I, cases 1, 3, 4, 5] For RR-DSR in cases 1, 3, 4, and 5, Δp_L = 0 indicates that no improving topology was found within the time limit, so the reported RR-DSR objective and loss values are not converged reconfiguration results. Comparing a completed C-DSR run with a timed-out RR-DSR run biases the loss and runtime comparison in favor of C-DSR. Additionally, case 5 shows no speedup at all (C-DSR 906.93 s vs RR-DSR 902.35 s), so the abstract's 'can significantly improve solver performance' is overstated. Report the best solution found within the same time limit, together with convergence evidence for both models, and qualify the performance claim in light of case 5.
[Section II, 'Safety ratings', and Table I] The models remove the voltage and loading constraints (2) and (3), and several reported solutions are voltage-infeasible: C-DSR has γv = 0.0045 p.u. in case 4 and 0.0182 p.u. in case 5, and RR-DSR has nonzero γv in cases 2, 3, 4, and 5. The paper's assertion that minimal voltage violations are operationally admissible for short durations is not quantified or demonstrated, and the optimization objective does not penalize these violations. The reported loss values therefore correspond to operating points that are infeasible under the original safety limits, and the title's 'loss-minimal radial topologies' is not established for feasible radial configurations. Enforce or penalize the safety limits, or clearly present the study as a relaxed benchmark with an explicit operational admissibility argument.
[Appendix B, proof of Corollary 2] The proof of Corollary 2 is not rigorous: after assuming an edge e in a cycle but in no fundamental cycle of a given cycle basis, it simply states that 'B is clearly no basis of the cycle space of N', which begs the question. A correct proof is needed, for example by showing that every cycle-edge lies on at least one fundamental cycle of any spanning-tree basis via a cut argument. Since Corollary 2 is a stated contribution (replacing full cycle enumeration with a cycle basis), the manuscript should provide a complete proof.

minor comments (5)

[Figures] Figure 1 is used twice with different content (topology types and cycle-edges), and Figure 3 is used twice with captions referencing MV-Comm and MV-Semiurb; the figures and captions need renumbering and correction.
[Corollary 1] The statement of Corollary 1 is typeset incorrectly: 'e ∈ S ∀k E_N^k' should be 'e ∈ ⋃_k E(C_N^k)'.
[Section III] The text says 'two different load profiles for two of them' but then describes cases 1 and 2 as differing by renewable energy sources rather than load profiles; clarify the scenario construction and define Δp_L in terms of baseline losses.
[Section III] The paper should report solver version, hardware, convergence tolerances, MIP gap settings, and whether runs were repeated; this information is necessary to interpret heuristic wall-clock times.
[PF validation] The statement that all variable values are within 9.3×10^{-9} of the reference PowerModels.jl implementation should identify the validation instance and the variables compared.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims rest on an empirical solver benchmark with externally stated graph-theoretic radiality constraints, not on fitted inputs or self-referential derivation.

full rationale

The paper proposes C-DSR, which enforces radiality via the all-cycle inequalities (6b) together with the edge-count constraint (6c), and benchmarks it against RR-DSR, which omits (6b). The reduction from (6b)+(6c) to a spanning tree is an application of the standard graph-theoretic characterization in Theorem 1, cited to an external textbook [15], not a definition tailored to the paper's conclusion. Corollaries 1 and 2 are proved from Proposition 1 and the cycle-space basis property [16]; they are independent mathematical facts, not restatements of the computational results. The performance comparison in Table I is empirical: it reports wall-clock times and objective values from Juniper on the external SimBench cases [14], with no parameters fitted to the data being 'predicted'. The only self-citation, [11], is used as a loosely similar base model for R-DSR and is not load-bearing for the central claim; the actual power-flow equations are taken from the external PowerModels framework [12]. The paper does not rename a known result, smuggle an ansatz via citation, or import a uniqueness theorem from the authors' prior work. While one may question whether the speedup generalizes beyond warm-started heuristic runs, that is a correctness/robustness concern, not circularity. The derivation chain is self-contained and the computational claims are externally benchmarked, so the circularity score is 0.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The model introduces no free parameters or invented entities. Its load-bearing assumptions are the standard graph-theoretic and power-flow facts listed above, plus the modeling choice to relax safety limits.

assumptions (6)

domain assumption The distribution network graph N is connected and undirected.
Stated in Appendix A: 'graph N is assumed to be connected and undirected.' This is needed for Theorem 1 and for existence of spanning trees.
domain assumption The exact AC power flow equations (1d)-(1g) accurately model the steady-state power system.
The model rests on the exact PF formulation from [12]; no approximation errors are quantified in the optimization.
standard math Minimizing total power injection from the external grid (objective 1a) is equivalent to minimizing power losses.
Cited to [6]; standard result for fixed loads.
ad hoc to paper Temporary violations of voltage and line loading limits are operationally admissible for short durations.
The authors relax constraints (2) and (3) and evaluate violations afterward; this is a modeling choice that lets infeasible solutions be compared.
standard math Any cycle in a graph can be expressed as the symmetric difference of the elements of a cycle basis.
Used to prove Corollary 2; standard cycle space theorem.
domain assumption The solver Juniper is a suitable heuristic for comparing formulations.
Performance conclusions are drawn from Juniper runs; other solvers may behave differently.

how reviews work

0 comments

Cite this review

Pith. "Pith review of On Loss-Minimal Radial Topologies in MV Systems." pith.science (2026). https://pith.science/paper/5GGJFJ5K

@misc{pith2026250603422,
  author       = {Pith},
  title        = {Pith review of: On Loss-Minimal Radial Topologies in MV Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5GGJFJ5K}},
  note         = {Machine review of arXiv:2506.03422}
}

read the original abstract

Distribution system reconfiguration (DSR) means optimizing the topology of a distribution grid using switching actions. Switching actions are a degrees of freedom available to distribution system operators, e.g. to manage planned and unplanned outages. DSR is a NP-hard combinatorial problem. Finding good or even optimal solutions is computationally expensive. While transmission and high-voltage grids are generally operated in a meshed state, MV distribution systems are commonly operated as radial networks even though meshed operation would be supported. This improves resilience because faults can be isolated more easily keeping the rest of the system operational and minimizing impact on customers. We propose an AC DSR formulation and benchmark it against a common formulation from the literature. Our results indicate that additional acyclicity constraints can significantly improve solver performance.

Figures

Figures reproduced from arXiv: 2506.03422 by the authors.

**Figure 1.** This plot shows all cycle-edges in gray, identified using [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** The figure shows results for case 2, MV-Rural without [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: This figure shows all cycles in the MV-Rural test case enumerated using Gibbs’ algorithm [13]. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

16 extracted references · 15 canonical work pages

[8]

A mixed-integer linear programming approach for the computation of the minimum-losses radial configuration of electrical distribution networks,

A. Borghetti, “A mixed-integer linear programming approach for the computation of the minimum-losses radial configuration of electrical distribution networks,”IEEE Trans. Power Syst., vol. 27, no. 3, pp. 1264–1273, Aug. 2012

work page 2012
[10]

Formu- lation of radiality constraints for optimal microgrid formation,

K. Pang, C. Wang, N. D. Hatziargyriou, F. Wen, and Y . Xue, “Formu- lation of radiality constraints for optimal microgrid formation,”IEEE Trans. Power Syst., vol. 38, no. 6, pp. 5341–5355, Nov. 2023

work page 2023
[11]

Optimal transmission switching: Improving solver performance using heuristics,

A. Hinneck and D. Pozo, “Optimal transmission switching: Improving solver performance using heuristics,”IEEE Transactions on Power Systems, vol. 38, no. 4, pp. 3317–3330, 2023

work page 2023
[1]

An efficient algorithm for minimum loss re- configuration of distribution system based on sensitivity and heuristics,

G. Raju and P. R. Bijwe, “An efficient algorithm for minimum loss re- configuration of distribution system based on sensitivity and heuristics,” IEEE Trans. Power Syst., vol. 23, no. 3, pp. 1280–1287, Aug. 2008

work page 2008
[2]

A novel optimization algorithm solving network reconfiguration,

A. De Bonis, J. P. S. Catalao, A. Mazza, G. Chicco, and F. Torelli, “A novel optimization algorithm solving network reconfiguration,” in2014 Power Systems Computation Conference. IEEE, Aug. 2014

work page 2014
[3]

Imposing radiality constraints in distribution system optimization problems,

M. Lavorato, J. F. Franco, M. J. Rider, and R. Romero, “Imposing radiality constraints in distribution system optimization problems,”IEEE Trans. Power Syst., vol. 27, no. 1, pp. 172–180, Feb. 2012

work page 2012
[4]

Assessing the potential of network reconfiguration to improve dis- tributed generation hosting capacity in active distribution systems,

F. Capitanescu, L. F. Ochoa, H. Margossian, and N. D. Hatziargyriou, “Assessing the potential of network reconfiguration to improve dis- tributed generation hosting capacity in active distribution systems,”IEEE Trans. Power Syst., vol. 30, no. 1, pp. 346–356, Jan. 2015

work page 2015
[5]

Juniper: An open- source nonlinear branch-and-bound solver in julia,

O. Kr ¨oger, C. Coffrin, H. Hijazi, and H. Nagarajan, “Juniper: An open- source nonlinear branch-and-bound solver in julia,” inIntegration of Constraint Programming, Artificial Intelligence, and Operations Re- search. Springer International Publishing, 2018, pp. 377–386

work page 2018

Show all 16 references

[6]

J. A. Taylor,Convex optimization of power systems. Cambridge, England: Cambridge University Press, May 2018. 4 0 50 Bus f 1.00 0.95 vm f [p.u.] Baseline C-/RR-DSR (a) V oltage (b) Baseline topology (c) RR-/C-DSR Fig. 2: The figure shows results for case 2, MV-Rural without RES...

2018
[7]

VDE,Technische Anschlussregel Mittelspannung (VDE-AR-N 4110), 2023, Standard

2023
[9]

Repair and resource schedul- ing in unbalanced distribution systems using neighborhood search,

A. Arif, Z. Wang, C. Chen, and J. Wang, “Repair and resource schedul- ing in unbalanced distribution systems using neighborhood search,” IEEE Trans. Smart Grid, vol. 11, no. 1, pp. 673–685, Jan. 2020

2020
[12]

Powermodels. jl: An open-source framework for exploring power flow formulations,

C. Coffrin, R. Bent, K. Sundar, Y . Ng, and M. Lubin, “Powermodels. jl: An open-source framework for exploring power flow formulations,” in 2018 Power Systems Computation Conference (PSCC), 2018, pp. 1–8

2018
[13]

A cycle generation algorithm for finite undirected linear graphs,

N. E. Gibbs, “A cycle generation algorithm for finite undirected linear graphs,”J. ACM, vol. 16, no. 4, p. 564–568, oct 1969. [Online]. Available: https://doi.org/10.1145/321541.321545

1969
[14]

Simbench—a benchmark dataset of electric power systems to compare innovative solutions based on power flow analysis,

S. Meinecke, D. Sarajli ´c, S. R. Drauz, A. Klettke, L.-P. Lauven, C. Rehtanz, A. Moser, and M. Braun, “Simbench—a benchmark dataset of electric power systems to compare innovative solutions based on power flow analysis,”Energies, vol. 13, no. 12, p. 3290, Jun. 2020. [Online]....

2020 doi
[15]

J. L. Gross, J. Yellen, and M. Anderson,Graph theory and its applica- tions, 3rd ed., ser. Textbooks in Mathematics. London, England: CRC Press, Jan. 2023

2023
[16]

Deo,Graph theory with applications to engineering and computer science, 1st ed

N. Deo,Graph theory with applications to engineering and computer science, 1st ed. Mineola, NY: Dover Publications, Aug. 2016. 0 50 100 Bus f 1.00 0.95 vm f [p.u.] Baseline C-DSR (a) V oltage (b) Baseline topology (c) C-DSR Fig. 3: The figure shows results for case 4, MV-Semiu...

2016

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [8]

A mixed-integer linear programming approach for the computation of the minimum-losses radial configuration of electrical distribution networks,

A. Borghetti, “A mixed-integer linear programming approach for the computation of the minimum-losses radial configuration of electrical distribution networks,”IEEE Trans. Power Syst., vol. 27, no. 3, pp. 1264–1273, Aug. 2012

work page 2012

[2] [10]

Formu- lation of radiality constraints for optimal microgrid formation,

K. Pang, C. Wang, N. D. Hatziargyriou, F. Wen, and Y . Xue, “Formu- lation of radiality constraints for optimal microgrid formation,”IEEE Trans. Power Syst., vol. 38, no. 6, pp. 5341–5355, Nov. 2023

work page 2023

[3] [11]

Optimal transmission switching: Improving solver performance using heuristics,

A. Hinneck and D. Pozo, “Optimal transmission switching: Improving solver performance using heuristics,”IEEE Transactions on Power Systems, vol. 38, no. 4, pp. 3317–3330, 2023

work page 2023

[4] [1]

An efficient algorithm for minimum loss re- configuration of distribution system based on sensitivity and heuristics,

G. Raju and P. R. Bijwe, “An efficient algorithm for minimum loss re- configuration of distribution system based on sensitivity and heuristics,” IEEE Trans. Power Syst., vol. 23, no. 3, pp. 1280–1287, Aug. 2008

work page 2008

[5] [2]

A novel optimization algorithm solving network reconfiguration,

A. De Bonis, J. P. S. Catalao, A. Mazza, G. Chicco, and F. Torelli, “A novel optimization algorithm solving network reconfiguration,” in2014 Power Systems Computation Conference. IEEE, Aug. 2014

work page 2014

[6] [3]

Imposing radiality constraints in distribution system optimization problems,

M. Lavorato, J. F. Franco, M. J. Rider, and R. Romero, “Imposing radiality constraints in distribution system optimization problems,”IEEE Trans. Power Syst., vol. 27, no. 1, pp. 172–180, Feb. 2012

work page 2012

[7] [4]

Assessing the potential of network reconfiguration to improve dis- tributed generation hosting capacity in active distribution systems,

F. Capitanescu, L. F. Ochoa, H. Margossian, and N. D. Hatziargyriou, “Assessing the potential of network reconfiguration to improve dis- tributed generation hosting capacity in active distribution systems,”IEEE Trans. Power Syst., vol. 30, no. 1, pp. 346–356, Jan. 2015

work page 2015

[8] [5]

Juniper: An open- source nonlinear branch-and-bound solver in julia,

O. Kr ¨oger, C. Coffrin, H. Hijazi, and H. Nagarajan, “Juniper: An open- source nonlinear branch-and-bound solver in julia,” inIntegration of Constraint Programming, Artificial Intelligence, and Operations Re- search. Springer International Publishing, 2018, pp. 377–386

work page 2018