Z01
Refine
Language
- English (17)
Keywords
- Computational complexity (2)
- Gas networks (2)
- robust optimization (2)
- spheric-radial decomposition (2)
- Adaptive methods (1)
- Big-M (1)
- Bilevel optimization (1)
- Booking (1)
- Bookings (1)
- Bounding polyhedra (1)
Document Type
- Article (10)
- Preprint (6)
- Doctoral Thesis (1)
In many real-world mixed-integer optimisation problems from engineering, the side
constraints can be subdivided into two categories: constraints which describe a certain logic to model a feasible allocation of resources (such as a maximal number of available assets, working time requirements, maintenance requirements, contractual obligations, etc.),
and constraints which model physical processes and the related quantities (such as current,
pressure, temperature, etc.). While the first type of constraints can often easily be stated in
terms of a mixed-integer program (MIP), the second part may involve the incorporation of
complex non-linearities, partial differential equations or even a black-box simulation of the
involved physical process. In this work, we propose the integration of a trained tree-based
classifier – a decision-tree or a random forest, into a mixed-integer optimization model as a
possible remedy. We assume that the classifier has been trained on data points produced
by a detailed simulation of a given complex process to represent the functional relationship
between the involved physical quantities. We then derive MIP-representable reformulations
of the trained classifier such that the resulting model can be solved using state-of-the-art
solvers. At the hand of several use cases in terms of possible optimisation goals, we show
the broad applicability of our framework that is easily extendable to other tasks beyond
engineering. In a detailed real-world computational study for the design of stable direct-
current power networks, we demonstrate that our approach yields high-quality solutions
in reasonable computation times.
We propose an adaptive optimization algorithm for operating district heating networks in a stationary regime. The behavior of hot water flow in the pipe network is modeled using the incompressible Euler equations and a suitably chosen energy equation. By applying different simplifications to these equations, we derive a catalog of models. Our algorithm is based on this catalog and adaptively controls where in the network which model is used. Moreover, the granularity of the applied discretization is controlled in a similar adaptive manner. By doing so, we are able to obtain optimal solutions at low computational costs that satisfy a prescribed tolerance w.r.t. the most accurate modeling level. To adaptively control the switching between different levels and the adaptation of the discretization grids, we derive error measure formulas and a posteriori error measure estimators. Under reasonable assumptions we prove that the adaptive algorithm terminates after finitely many iterations. Our numerical results show that the algorithm is able to produce solutions for problem instances that have not been solvable before.
We study the robust maximum flow problem and the robust maximum flow over time problem where a given number of arcs Γ may fail or may be delayed. Two prominent models have been introduced for these problems: either one assigns flow to arcs fulfilling weak flow conservation in any scenario, or one assigns flow to paths where an arc failure or delay affects a whole path. We provide a unifying framework by presenting novel general models, in which we assign flow to subpaths. These models contain the known models as special cases and unify their advantages in order to obtain less conservative robust solutions.
We give a thorough analysis with respect to complexity of the general models. In particular, we show that the general models are essentially NP-hard, whereas, e.g. in the static case with Γ=1 an optimal solution can be computed in polynomial time. Further, we answer the open question about the complexity of the dynamic path model for Γ=1. We also compare the solution quality of the different models. In detail, we show that the general models have better robust optimal values than the known models and we prove bounds on these gaps.
Transport and trade of gas are decoupled after the liberalization of the European gas markets, which are now organized as so-called entry-exit systems. At the core of this market system are bookings and nominations, two special capacity-right contracts that grant traders access to the gas network. The latter is operated by a separate entity, known as the transmission system operator (TSO), who is in charge of the transport of gas from entry to exit nodes. In the mid to long term, traders sign a booking contract with the TSO to obtain injection and withdrawal capacities at entry and exit nodes, respectively. On a day-ahead basis, they then nominate within these booked capacities a balanced load flow of the planned amounts of gas to be injected into and withdrawn from the network the next day. The key property is that by signing a booking contract, the TSO is obliged to guarantee transportability for all balanced load flows in compliance with the booked capacities. To assess the feasibility of a booking, it is therefore necessary to check the feasibility of infinitely many nominations. As a result, deciding if a booking is feasible is a challenging mathematical problem, which we investigate in this dissertation.
Our results range from passive networks, consisting of pipes only, to active networks, containing controllable elements to influence gas flows. Since the study of the latter naturally leads to a bilevel framework, we first consider some more general properties of bilevel optimization. For the case of linear bilevel optimization, we consider the hardness of validating the correctness of big-Ms often used in solving these problems via a single-level reformulation. We also derive a family of valid inequalities to be used in a bilevel-tailored branch-and-cut algorithm as a big-M-free alternative.
We then turn to the study of feasible bookings. First, we present our results on passive networks, for which bilevel approaches are not required. A characterization of feasible bookings on passive networks is derived in terms of a finite set of nominations. While computing these nominations is a difficult task in general, we present polynomial complexity results for the special cases of tree-shaped or single-cycle passive networks. Finally, we consider networks with linearly modeled active elements. After obtaining a bilevel optimization model that allows us to determine the feasibility of a booking in this case, we derive various single-level reformulations to solve the problem. In addition, we obtain novel characterizations of feasible bookings on active networks, which generalize our characterization in the passive case. The performance of these various approaches is compared in a case study on two networks from the literature, one of which is a simplified version of the Greek gas network.
Linear bilevel optimization problems are known to be strongly NP-hard and the computational techniques to solve these problems are often motivated by techniques from single-level mixed-integer optimization. Thus, during the last years and decades many branch-and-bound methods, cutting planes, or heuristics have been proposed. On the other hand, there is almost no literature on presolving linear bilevel problems although presolve is a very important ingredient in state-of-the-art mixed-integer optimization solvers. In this paper, we carry over standard presolve techniques from single-level optimization to bilevel problems and show that this needs to be done with great caution since a naive application of well-known techniques does often not lead to correctly presolved bilevel models. Our numerical study shows that presolve can also be very beneficial for bilevel problems but also highlights that these methods have a more heterogeneous effect on the solution process compared to what is known from single-level optimization. As a side result, our numerical experiments reveal that there is an urgent need for better and more heterogeneous test instance libraries to further propel the field of computational bilevel optimization.
This contribution focuses on the analysis and control of friction-dominated flow of gas in pipes. The pressure in the gas flow is governed by a partial differential equation that is a doubly nonlinear parabolic equation of p-Laplace type, where p=2/3. Such equations exhibit positive solutions, finite speed of propagation and satisfy a maximum principle. The pressure is fixed on one end (upstream), and the flow is specified on the other end (downstream). These boundary conditions determine a unique steady equilibrium flow. We present a boundary feedback flow control scheme, that ensures local exponential stability of the equilibrium in an L2-sense. The analysis is done both for the pde system and an ode system that is obtained by a suitable spatial semi-discretization. The proofs are based upon suitably chosen Lyapunov functions.
We propose a mathematical optimization model and its solution for joint chance constrained DC Optimal Power Flow. In this application, it is particularly important that there is a high probability of transmission limits being satisfied, even in the case of uncertain or fluctuating feed-in from renewable energy sources. In critical network situations where the network risks overload, renewable energy feed-in has to be curtailed by the transmission system operator (TSO). The TSO can reduce the feed-in in discrete steps at each network node. The proposed optimization model minimizes curtailment while ensuring that there is a high probability of transmission limits being maintained. The latter is modeled via (joint) chance constraints that are computationally challenging. Thus, we propose a solution approach based on the robust safe approximation of these constraints. Hereby, probabilistic constraints are replaced by robust constraints with suitably defined uncertainty sets constructed from historical data. The uncertainty sets are calculated by encompassing randomly drawn scenarios using the scenario approach proposed by Margellos et al. (IEEE Transactions on Automatic Control, 59 (2014)). The ability to discretely control the power feed-in then leads to a robust optimization problem with decision-dependent uncertainties, i.e. the uncertainty sets depend on decision variables. We propose an equivalent mixed-integer linear reformulation for box uncertainties with the exact linearization of bilinear terms. Finally, we present numerical results for different test cases from the Nesta archive, as well as for a real network. We consider the discrete curtailment of solar feed-in, for which we use real-world weather and network data. The experimental tests demonstrate the effectiveness of this method and run times are very fast. Moreover, on average the calculated robust solutions lead only to a small increase in curtailment, when compared to nominal solutions.
Solving mixed-integer nonlinear optimization problems (MINLPs) to global optimality is extremely challenging. An important step for enabling their solution consists in the design of convex relaxations of the feasible set. Known solution approaches based on spatial branch-and-bound become more effective the tighter the used relaxations are. Relaxations are commonly established by convex underestimators, where each constraint function is considered separately. Instead, a considerably tighter relaxation can be found via so-called simultaneous convexification, where convex underestimators are derived for more than one constraint function at a time. In this work, we present a global solution approach for solving mixed-integer nonlinear problems that uses simultaneous convexification. We introduce a separation method that relies on determining the convex envelope of linear combinations of the constraint functions and on solving a nonsmooth convex problem. In particular, we apply the method to quadratic absolute value functions and derive their convex envelopes. The practicality of the proposed solution approach is demonstrated on several test instances from gas network optimization, where the method outperforms standard approaches that use separate convex relaxations.
We show that the feasibility of a booking in the European entry-exit gas market can be decided in polynomial time on single-cycle networks that are passive, i.e., do not contain controllable elements. The feasibility of a booking can be characterized by solving polynomially many nonlinear potential-based flow models for computing so-called potential-difference maximizing load flow scenarios. We thus analyze the structure of these models and exploit both the cyclic graph structure as well as specific properties of potential-based flows. This enables us to solve the decision variant of the nonlinear potential-difference maximization by reducing it to a system of polynomials of constant dimension that is independent of the cycle's size. This system of fixed dimension can be handled with tools from real algebraic geometry to derive a polynomial-time algorithm. The characterization in terms of potential-difference maximizing load flow scenarios then leads to a polynomial-time algorithm for deciding the feasibility of a booking. Our theoretical results extend the existing knowledge about the complexity of deciding the feasibility of bookings from trees to single-cycle networks.
One of the most frequently used approaches to solve linear bilevel optimization problems consists in replacing the lower-level problem with its Karush-Kuhn-Tucker (KKT) conditions and by reformulating the KKT complementarity conditions using techniques from mixed-integer linear optimization. The latter step requires to determine some big-M constant in order to bound the lower level's dual feasible set such that no bilevel-optimal solution is cut off. In practice, heuristics are often used to find a big-M although it is known that these approaches may fail. In this paper, we consider the hardness of two proxies for the above mentioned concept of a bilevel-correct big-M. First, we prove that verifying that a given big-M does not cut off any feasible vertex of the lower level's dual polyhedron cannot be done in polynomial time unless P=NP. Second, we show that verifying that a given big-M does not cut off any optimal point of the lower level's dual problem (for any point in the projection of the high-point relaxation onto the leader's decision space) is as hard as solving the original bilevel problem.
