Duality and sensitivity analysis of multistage linear stochastic programs

TL;DR

This paper explores the duality and sensitivity analysis of multistage stochastic linear programs, introduces Dual SDDP variants for dual problem solving, and demonstrates their effectiveness through numerical experiments on inventory and hydro-thermal planning problems.

Contribution

It develops Dual SDDP algorithms with penalizations and feasibility cuts for solving dual MSLPs, especially when RCR may fail, and applies these to sensitivity analysis and complex problem classes.

Findings

Formulas for derivatives of the value function with respect to parameters.

Dual SDDP with penalizations converges to the optimal value.

Numerical experiments show the effectiveness of the proposed methods.

Abstract

We investigate the dual of a Multistage Stochastic Linear Program (MSLP) to study two questions for this class of problems. The first of these questions is the study of the optimal value of the problem as a function of the involved parameters. For this sensitivity analysis problem, we provide formulas for the derivatives of the value function with respect to the parameters and illustrate their application on an inventory problem. Since these formulas involve optimal dual solutions, we need an algorithm that computes such solutions to use them. In this context, the second question we address is the study of solution methods for the dual problem. Writing Dynamic Programming (DP) equations for the dual, we can use an SDDP type method, called Dual SDDP, which solves these DP equations. However, applying this method will only be possible if the Relatively Complete Recourse (RCR) holds for the dual. Since the RCR assumption may fail to hold, we design two variants of Dual SDDP, namely Dual SDDP with penalizations and Dual SDDP with feasibility cuts. The study of this second question allows us to take a fresh look at the notoriously difficult to solve class of MSLP with interstage dependent cost coefficients. Indeed, for this class of problems, cost-to-go functions are non-convex and solution methods were so far using SDDP for a Markov chain approximation of the cost coefficients process. For these problems, we propose to apply Dual SDDP with penalizations to the cost-to-go functions of the dual which are concave. This algorithm converges to the optimal value of the problem. Finally, as a proof of concept of the tools developed, we present the results of numerical experiments computing the sensitivity of the optimal value of an inventory problem as a function of parameters of the demand process and compare Primal and Dual SDDP on the inventory and a hydro-thermal planning problems.