skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Prediction-Correction Algorithms for Time-Varying Constrained Optimization

Abstract

This article develops online algorithms to track solutions of time-varying constrained optimization problems. Particularly, resembling workhorse Kalman filtering-based approaches for dynamical systems, the proposed methods involve prediction-correction steps to provably track the trajectory of the optimal solutions of time-varying convex problems. The merits of existing prediction-correction methods have been shown for unconstrained problems and for setups where computing the inverse of the Hessian of the cost function is computationally affordable. This paper addresses the limitations of existing methods by tackling constrained problems and by designing first-order prediction steps that rely on the Hessian of the cost function (and do not require the computation of its inverse). In addition, the proposed methods are shown to improve the convergence speed of existing prediction-correction methods when applied to unconstrained problems. Numerical simulations corroborate the analytical results and showcase performance and benefits of the proposed algorithms. A realistic application of the proposed method to real-time control of energy resources is presented.

Authors:
 [1];  [2]
  1. IBM Research, Dublin (Ireland)
  2. National Renewable Energy Lab. (NREL), Golden, CO (United States)
Publication Date:
Research Org.:
National Renewable Energy Lab. (NREL), Golden, CO (United States)
Sponsoring Org.:
USDOE Office of Energy Efficiency and Renewable Energy (EERE); NREL Laboratory Directed Research and Development (LDRD)
OSTI Identifier:
1394904
Report Number(s):
NREL/JA-5D00-67655
Journal ID: ISSN 1053-587X
Grant/Contract Number:
AC36-08GO28308
Resource Type:
Journal Article: Accepted Manuscript
Journal Name:
IEEE Transactions on Signal Processing
Additional Journal Information:
Journal Volume: 65; Journal Issue: 20; Journal ID: ISSN 1053-587X
Publisher:
IEEE
Country of Publication:
United States
Language:
English
Subject:
97 MATHEMATICS AND COMPUTING; time-varying optimization; non-stationary optimization; parametric programming; prediction-correction methods; real-time control of energy resources

Citation Formats

Simonetto, Andrea, and Dall'Anese, Emiliano. Prediction-Correction Algorithms for Time-Varying Constrained Optimization. United States: N. p., 2017. Web. doi:10.1109/TSP.2017.2728498.
Simonetto, Andrea, & Dall'Anese, Emiliano. Prediction-Correction Algorithms for Time-Varying Constrained Optimization. United States. doi:10.1109/TSP.2017.2728498.
Simonetto, Andrea, and Dall'Anese, Emiliano. Wed . "Prediction-Correction Algorithms for Time-Varying Constrained Optimization". United States. doi:10.1109/TSP.2017.2728498.
@article{osti_1394904,
title = {Prediction-Correction Algorithms for Time-Varying Constrained Optimization},
author = {Simonetto, Andrea and Dall'Anese, Emiliano},
abstractNote = {This article develops online algorithms to track solutions of time-varying constrained optimization problems. Particularly, resembling workhorse Kalman filtering-based approaches for dynamical systems, the proposed methods involve prediction-correction steps to provably track the trajectory of the optimal solutions of time-varying convex problems. The merits of existing prediction-correction methods have been shown for unconstrained problems and for setups where computing the inverse of the Hessian of the cost function is computationally affordable. This paper addresses the limitations of existing methods by tackling constrained problems and by designing first-order prediction steps that rely on the Hessian of the cost function (and do not require the computation of its inverse). In addition, the proposed methods are shown to improve the convergence speed of existing prediction-correction methods when applied to unconstrained problems. Numerical simulations corroborate the analytical results and showcase performance and benefits of the proposed algorithms. A realistic application of the proposed method to real-time control of energy resources is presented.},
doi = {10.1109/TSP.2017.2728498},
journal = {IEEE Transactions on Signal Processing},
number = 20,
volume = 65,
place = {United States},
year = {Wed Jul 26 00:00:00 EDT 2017},
month = {Wed Jul 26 00:00:00 EDT 2017}
}

Journal Article:
Free Publicly Available Full Text
This content will become publicly available on July 26, 2018
Publisher's Version of Record

Save / Share:
  • This paper focuses on the design of online algorithms based on prediction-correction steps to track the optimal solution of a time-varying constrained problem. Existing prediction-correction methods have been shown to work well for unconstrained convex problems and for settings where obtaining the inverse of the Hessian of the cost function can be computationally affordable. The prediction-correction algorithm proposed in this paper addresses the limitations of existing methods by tackling constrained problems and by designing a first-order prediction step that relies on the Hessian of the cost function (and do not require the computation of its inverse). Analytical results are establishedmore » to quantify the tracking error. Numerical simulations corroborate the analytical results and showcase performance and benefits of the algorithms.« less
  • The authors describe a convergence theory for evolutionary pattern search algorithms (EPSAs) on a broad class of unconstrained and linearly constrained problems. EPSAs adaptively modify the step size of the mutation operator in response to the success of previous optimization steps. The design of EPSAs is inspired by recent analyses of pattern search methods. The analysis significantly extends the previous convergence theory for EPSAs. The analysis applies to a broader class of EPSAs,and it applies to problems that are nonsmooth, have unbounded objective functions, and which are linearly constrained. Further, they describe a modest change to the algorithmic framework ofmore » EPSAs for which a non-probabilistic convergence theory applies. These analyses are also noteworthy because they are considerably simpler than previous analyses of EPSAs.« less
  • Automatic data-base independent peptide identification from collision-induced dissociation tandem mass spectrometry data is made difficult by large plateaus in the fitness landscapes of scoring functions and the fuzzy nature of the constraints that is due to noise in the data. Two different scoring functions are combined into a parallel multi-objective optimization framework.
  • A numerically efficient procedure is presented for computing optimal time-varying pumping rates for remediation of contaminated groundwater described by two-dimensional numerical models. The management model combines a pollutant transport model with a constrained optimal control algorithm. The transport model simulates the unsteady fluid flow and transient contamination dispersion-advection in a two-dimensional confined aquifer. A Galerkin's finite element method coupled with a fully implicit time difference scheme is applied to solve the groundwater flow and contaminant transport equations. The constrained optimal control algorithm employs a hyperbolic penalty function. Several sample problems covering 5-15 years of remediation are given to illustrate themore » capability of the management model to solve a groundwater quality control problem with time-varying pumping policy and water quality constraints. In the example, the optimal constant pumping rates are 75% more expensive than the optimal time-varying pumping rates, a result that supports the need to develop numerically efficient optimal control-finite element algorithms for groundwater remediation. 28 refs., 7 figs., 10 tabs.« less
  • The criterion of thermodynamic ideality is introduced. It is the ratio of actual rate of entropy production to the minimal rate of entropy production. It is closely related to the exergy approach but incorporates the irreversible losses due to finite-time and nonzero rates. No explicit reference need be made to the average values of the thermodynamic potentials of the environment. The regimes with minimal entropy production for the system where chemical reaction occurs and for the set of parallel heat engines are found.