∇ProvenDecisions
Ctrl K

Benders for power flows, bilevel cuts and decision-diagram intersection cuts

The day in optimization

Six new preprints appeared on Optimization Online, and most of them concern strengthening or speeding up relaxations: Benders subproblems, bilevel cuts and MINLP cuts. A smaller group covers statistical fitting (sparse regression, slope change) and benchmark instance generation. No single result dominates the day.

Theory

Decomposition and cuts. Fast Subproblem Solving in Benders Decomposition for Affine Potential-Based Flow Problems targets networks such as lossless power grids. There, flows and node potentials are tied together by an arc potential equation. The paper allows affine potential functions, switchable topology and multiple scenarios. Benders decomposition splits such a problem into a master problem and many scenario subproblems. Solving those subproblems quickly is the bottleneck the paper addresses. The abstract excerpt we have does not give the speedups, so we do not quote any.

Valid inequalities for bilevel 0-1 shortest-path problems studies a leader who makes binary choices while a follower routes along a shortest path. The authors use a dualize-and-combine reformulation. They show that fractional solutions can still appear in the continuous relaxation even when both the leader and the follower have ideal formulations. This is why extra valid inequalities are needed on top of the individual formulations.

Intersection Cuts for MINLPs via Decision-Diagram-Free Sets starts from decision diagrams, which give an outer approximation of a bounded MINLP's feasible region. It defines "DD-free sets": convex regions whose interiors contain no feasible points. Such sets are the raw material for intersection cuts, which slice off a relaxation's current solution. The idea is to get cuts for general nonlinear constraints.

Statistics-flavoured problems. Relax-and-Sparsify Approximation Algorithms for Sparse Ridge Regression solves a convex relaxation and then sparsifies its solution to meet the sparsity limit. It focuses on low-rank, high-dimensional data. The authors report a tight sparsity bound that depends on the rank.

Polynomial-Time Solvability of the Multi-Dimensional Slope Change Problem fits a multidimensional time series with continuous piecewise-linear functions that share one breakpoint. The title states polynomial-time solvability, and the abstract covers both the dense setting (every dimension may change slope) and the sparse setting.

In practice

Grid topology optimization. Canary Media reports on NewGrid's software. It optimizes transmission topology (which lines are switched in) in day-ahead and real-time operations. Per the article, simulations across whole grid-operator networks suggest congestion costs could fall by 30 to 50 percent. Those are simulation results, not measured savings from operations. This is the same kind of topology-switching problem as the Benders paper above.

Airport staff scheduling. Gurobi's Swissport case study is a vendor account, so treat the numbers as the vendor's claims. It says better planning saves Swissport over $1 million a year and meets employee scheduling preferences 95–100% of the time. We could not confirm the date of this case study, so it is not necessarily recent.

Worth watching

Try ProvenDecisions free: type, photograph or say an optimization problem; every answer is checked by a solver.

What do you want to optimize?

Type the math, describe it in words, snap a photo or say it. ProvenDecisions builds the model, solves it, checks it and walks you through it.

One constraint per line, or separated by commas. Add integer: x, y or binary: z for whole-number decisions. LaTeX works too.

Try: