[Paper Review] Decision-Oriented Learning for Future Power System Decision-Making under Uncertainty

Ran Li, Haipeng Zhang+5 more|arXiv (Cornell University)|Jan 8, 2024

Energy Load and Power ForecastingEngineering87 references5 citations

TL;DR

This paper surveys decision-oriented learning (DOL) for power systems, showing how end-to-end decision losses can outperform traditional forecast-focused learning and reviewing techniques, applications, and challenges.

ABSTRACT

Better forecasts may not lead to better decision-making. To address this challenge, decision-oriented learning (DOL) has been proposed as a new branch of machine learning that replaces traditional statistical loss with a decision loss to form an end-to-end model. Applications of DOL in power systems have been developed in recent years. For renewable-rich power systems, uncertainties propagate through sequential tasks, where traditional statistical-based approaches focus on minimizing statistical errors at intermediate stages but may fail to provide optimal decisions at the final stage. This paper first elaborates on the mismatch between more accurate forecasts and more optimal decisions in the power system caused by statistical-based learning (SBL) and explains how DOL resolves this problem. Secondly, this paper extensively reviews DOL techniques and their applications in power systems while highlighting their pros and cons in relation to SBL. Finally, this paper identifies the challenges to adopt DOL in the energy sector and presents future research directions.

Motivation & Objective

• Explain the mismatch between forecast accuracy and decision quality in renewable-rich power systems.
• Introduce and motivate decision-oriented learning (DOL) as an end-to-end alternative to statistical-based learning.
• Review DOL techniques and their applications in power systems.
• Compare DOL with statistical-based learning (SBL) and discuss pros and cons.
• Identify challenges for adopting DOL in the energy sector and propose future research directions.

Proposed method

• Provide a conceptual analysis of DOL in the context of power systems.
• Conduct a literature review of DOL techniques and their applications in power systems.
• Discuss the advantages and disadvantages of DOL relative to SBL.
• Organize and synthesize existing applications, categorizations, and outcomes.
• Identify practical challenges for adoption and outline future research directions.

Experimental results

Research questions

• RQ1 What is the mismatch between forecast accuracy and optimal decision quality in power system decision-making under uncertainty?
• RQ2 How can decision-oriented learning (DOL) resolve the mismatch between forecasts and decisions in power systems?
• RQ3 What are the advantages and limitations of DOL compared with traditional statistical-based learning (SBL) in renewable-rich systems?
• RQ4 What challenges hinder the adoption of DOL in the energy sector, and what future research directions are proposed?

Key findings

• DOL replaces traditional statistical losses with a decision-focused loss to align learning with final decision quality.
• DOL has been applied to power-system problems with benefits and trade-offs compared to SBL.
• The paper highlights the mismatch between forecast accuracy and decision optimality as a core motivation for DOL.
• It surveys techniques and applications, outlining pros and cons of DOL in renewable-rich settings.
• It identifies practical challenges for adoption and suggests directions for future research.