Real-time peak-demand minimization with energy storage using competitive ratio

Yanfang Mo; Qiulin Lin; Minghua Chen; S. Joe Qin

doi:10.1016/j.automatica.2025.112687

Real-time peak-demand minimization with energy storage using competitive ratio

Yanfang Mo^*, Qiulin Lin, Minghua Chen^*, S. Joe Qin

^*Corresponding author for this work

Department of Data Science

Research output: Journal Publications and Reviews › RGC 21 - Publication in refereed journal › peer-review

Abstract

This paper studies minimizing the peak demand for large-load users with energy storage by making irrevocable storage-consuming decisions with unknown future demands in a daily on-peak horizon. We use competitive analysis to optimize online decision-making for robustness and fairness over all possible input sequences. The best competitive ratio (CR) for the online problem is computed by solving a linear number of linear-fractional programs and used to obtain an optimal online strategy, which is a static feedback policy affine over a synthetic state. Further, we generalize the concept of CR to shrinking-horizon CRs in response to real-time inputs and decisions. Accordingly, a real-time optimal online algorithm is developed to achieve the rolling-optimized shrinking-horizon CRs. This algorithm is a dynamic feedback policy, retains the global optimal worst-case performance, and improves the general-case performance. Trace-driven simulations show that the real-time optimal algorithm can significantly decrease the peak demand compared to baseline alternatives under typical settings. © 2025 Elsevier Ltd.

Original language	English
Article number	112687
Number of pages	10
Journal	Automatica
Volume	184
Online published	14 Nov 2025
DOIs	https://doi.org/10.1016/j.automatica.2025.112687
Publication status	Published - Feb 2026

Funding

The work was supported in part by the Research Grants Council (RGC) of Hong Kong under the General Research Fund (Project Nos. 16206324 , 11206821 , 11303421 , 13300525 ), the Research Impact Fund (Project No. 130272 ), and the Collaborative Research Fund (Project No. C1049-24G ); by the Laboratory for AI-Powered Financial Technologies under an InnoHK initiative of the HKSAR; by a Shenzhen-Hong Kong-Macau Science & Technology Project (Category C, Project No. SGDX20220530111203026 ); by a Start-up Research Grant (Project No. UDF01004086 ) from The Chinese University of Hong Kong, Shenzhen ; and by the grant DR25E7 from Lingnan University.

Research Keywords

Demand response
Optimization under uncertainties
Receding horizon control
Resource allocation

RGC Funding Information

RGC-funded

Access Output

10.1016/j.automatica.2025.112687

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@article{bbf4472586f34feda45f311d8d98f7a7,

title = "Real-time peak-demand minimization with energy storage using competitive ratio",

abstract = "This paper studies minimizing the peak demand for large-load users with energy storage by making irrevocable storage-consuming decisions with unknown future demands in a daily on-peak horizon. We use competitive analysis to optimize online decision-making for robustness and fairness over all possible input sequences. The best competitive ratio (CR) for the online problem is computed by solving a linear number of linear-fractional programs and used to obtain an optimal online strategy, which is a static feedback policy affine over a synthetic state. Further, we generalize the concept of CR to shrinking-horizon CRs in response to real-time inputs and decisions. Accordingly, a real-time optimal online algorithm is developed to achieve the rolling-optimized shrinking-horizon CRs. This algorithm is a dynamic feedback policy, retains the global optimal worst-case performance, and improves the general-case performance. Trace-driven simulations show that the real-time optimal algorithm can significantly decrease the peak demand compared to baseline alternatives under typical settings. {\textcopyright} 2025 Elsevier Ltd.",

keywords = "Demand response, Optimization under uncertainties, Receding horizon control, Resource allocation",

author = "Yanfang Mo and Qiulin Lin and Minghua Chen and Qin, {S. Joe}",

year = "2026",

month = feb,

doi = "10.1016/j.automatica.2025.112687",

language = "English",

volume = "184",

journal = "Automatica",

issn = "0005-1098",

publisher = "Elsevier Ltd.",

}

TY - JOUR

T1 - Real-time peak-demand minimization with energy storage using competitive ratio

AU - Mo, Yanfang

AU - Lin, Qiulin

AU - Chen, Minghua

AU - Qin, S. Joe

PY - 2026/2

Y1 - 2026/2

N2 - This paper studies minimizing the peak demand for large-load users with energy storage by making irrevocable storage-consuming decisions with unknown future demands in a daily on-peak horizon. We use competitive analysis to optimize online decision-making for robustness and fairness over all possible input sequences. The best competitive ratio (CR) for the online problem is computed by solving a linear number of linear-fractional programs and used to obtain an optimal online strategy, which is a static feedback policy affine over a synthetic state. Further, we generalize the concept of CR to shrinking-horizon CRs in response to real-time inputs and decisions. Accordingly, a real-time optimal online algorithm is developed to achieve the rolling-optimized shrinking-horizon CRs. This algorithm is a dynamic feedback policy, retains the global optimal worst-case performance, and improves the general-case performance. Trace-driven simulations show that the real-time optimal algorithm can significantly decrease the peak demand compared to baseline alternatives under typical settings. © 2025 Elsevier Ltd.

AB - This paper studies minimizing the peak demand for large-load users with energy storage by making irrevocable storage-consuming decisions with unknown future demands in a daily on-peak horizon. We use competitive analysis to optimize online decision-making for robustness and fairness over all possible input sequences. The best competitive ratio (CR) for the online problem is computed by solving a linear number of linear-fractional programs and used to obtain an optimal online strategy, which is a static feedback policy affine over a synthetic state. Further, we generalize the concept of CR to shrinking-horizon CRs in response to real-time inputs and decisions. Accordingly, a real-time optimal online algorithm is developed to achieve the rolling-optimized shrinking-horizon CRs. This algorithm is a dynamic feedback policy, retains the global optimal worst-case performance, and improves the general-case performance. Trace-driven simulations show that the real-time optimal algorithm can significantly decrease the peak demand compared to baseline alternatives under typical settings. © 2025 Elsevier Ltd.

KW - Demand response

KW - Optimization under uncertainties

KW - Receding horizon control

KW - Resource allocation

UR - http://www.scopus.com/inward/record.url?scp=105021476525&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-105021476525&origin=recordpage

U2 - 10.1016/j.automatica.2025.112687

DO - 10.1016/j.automatica.2025.112687

M3 - RGC 21 - Publication in refereed journal

SN - 0005-1098

VL - 184

JO - Automatica

JF - Automatica

M1 - 112687

ER -

Real-time peak-demand minimization with energy storage using competitive ratio

Abstract

Funding

Research Keywords

RGC Funding Information

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

CRF: Machine Learning for Reliable and Efficient Power System Operation Considering Renewable and Load Uncertainty

CRF-Sub-pj: Machine Learning for Reliable and Efficient Power System Operation Considering Renewable and Load Uncertainty

GRF: Competitive and Prediction-Aware Online Optimization for Storage-Assisted Demand Response under Load Uncertainty and Peak-Demand Charge

Cite this

Real-time peak-demand minimization with energy storage using competitive ratio

Abstract

Funding

Research Keywords

RGC Funding Information

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Projects

CRF: Machine Learning for Reliable and Efficient Power System Operation Considering Renewable and Load Uncertainty

CRF-Sub-pj: Machine Learning for Reliable and Efficient Power System Operation Considering Renewable and Load Uncertainty

GRF: Competitive and Prediction-Aware Online Optimization for Storage-Assisted Demand Response under Load Uncertainty and Peak-Demand Charge

Cite this