Two-stage linear decision rules for multi-stage stochastic programming

Abstract

Multi-stage stochastic linear programs (MSLPs) are notoriously hard to solve in general. Linear decision rules (LDRs) yield an approximation of an MSLP by restricting the decisions at each stage to be an affine function of the observed uncertain parameters. Finding an optimal LDR is a static optimization problem that provides an upper bound on the optimal value of the MSLP, and, under certain assumptions, can be formulated as an explicit linear program. Similarly, as proposed by Kuhn, Wiesemann, and Georghiou ("Primal and dual linear decision rules in stochastic and robust optimization" Math. Program. 130, 177-209, 2011) a lower bound for an MSLP can be obtained by restricting decisions in the dual of the MSLP to follow an LDR. We propose a new approximation approach for MSLPs, two-stage LDRs. The idea is to require only the state variables in an MSLP to follow an LDR, which is sufficient to obtain an approximation of an MSLP that is a two-stage stochastic linear program (2SLP). We similarly propose to apply LDR only to a subset of the variables in the dual of the MSLP, which yields a 2SLP approximation of the dual that provides a lower bound on the optimal value of the MSLP. Although solving the corresponding 2SLP approximations exactly is intractable in general, we investigate how approximate solution approaches that have been developed for solving 2SLP can be applied to solve these approximation problems, and derive statistical upper and lower bounds on *

A Transformation-Proximal Bundle Algorithm for Solving Multistage Adaptive Robust Optimization Problems

“…A MARMILP in its general form is shown as follows. (Bodur and Luedtke 2017). Also note that a large class of multistage ARO problems can be reformulated in this form through the introduction of additional variables and constraints (Zou, Ahmed and Sun 2018).…”

“…This paper proposes a novel transformation-proximal bundle algorithm to solve multistage adaptive robust mixed-integer linear programs (MARMILPs). In a multistage decision-making setting, decision variables can be partitioned into two different groups, namely state decision variables and local decision variables (Bodur and Luedtke 2017, Zou et al 2018). We first propose a novel multi-to-two transformation scheme that converts the multistage ARO problem into an equivalent two-stage counterpart.…”

A Scenario Approach for Chance-Constrained Short-Term Scheduling With Affine Rules

“…Finally, Bodur et al [61] applied a different approach to multistage problems, where the affine decision rules are applied only to the state variables. This modelling allows a multistage problem to be treated as a two-stage problem, which drastically reduces the computational burden.…”

A Regularized Benders Decomposition With Multiple Master Problems to Solve the Hydrothermal Generation Expansion Problem

“…A nonanticipative operational policy is a rule defining decision variables of a given period t based on previously reveled information, i.e., without assuming access to the information of uncertainty factors after t. The linear decision rule (LDR) methodology defines an implementable nonanticipative policy based on an optimized linear combination of functions applied to the previously revealed uncertainty scenario. [44] first proposed the LDR for reservoir management, and this approach is gaining more and more attention in the literature (see [45][46][47][48] and [49]) and will be used in this work to propose a new stochastic hydrothermal GEP with multistage (nonanticipative) dispatch policy based on linear decision rules.…”

“…We assume a discrete and finite sample space Ω = {1, .., ω, ...}, in which each scenario ω is assumed to have a known conditional probability p ω . For the operational variables, we use a LDR to consider a monthly nonanticipative multistage operational policy under uncertainty of inflows [49,50] for the reservoirs.…”