Learning optimization proxies for large-scale Security-Constrained Economic Dispatch

Chen, Wenbo; Park, Seonho; Tanneau, Mathieu; Hentenryck, Pascal Van

doi:10.1016/j.epsr.2022.108566

Cited by 28 publications

(14 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The values are plotted in log-scaled, highlighting the skewed nature of the active powers and voltage magnitudes: indeed, most values lie on their extreme limits. This has been observed before, leading to the use of the classificationthen-regression approach [12]. However, fortunately, Figure 2 (right) shows that the distribution of the principal components is well-posed: it is more convenient to learn the regression to the principal components rather than to the original optimal solution space directly.…”

Section: Dimension Reduction Through Pcamentioning

confidence: 74%

“…This approach has attracted significant attention in power system applications recently because it holds the promise of decreasing the computation time needed to solve recurring optimization problems with reasonably small variations of the input parameters. For example, a classification-them-regression framework [12] was proposed to directly estimate the optimal solutions to the security constrained economic dispatch (SCED) problem. Because of the existence of the bound constraints, they recognized that the majority of the generators are at their maximum/minimum limits in optimal solutions.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Compact Optimization Learning for AC Optimal Power Flow

Park¹,

Chen²,

Mak³

et al. 2023

Preprint

View full text Add to dashboard Cite

This paper reconsiders end-to-end learning approaches to the Optimal Power Flow (OPF). Existing methods, which learn the input/output mapping of the OPF, suffer from scalability issues due to the high dimensionality of the output space. This paper first shows that the space of optimal solutions can be significantly compressed using principal component analysis (PCA). It then proposes COMPACT LEARNING, a new method that learns in a subspace of the principal components before translating the vectors into the original output space. This compression reduces the number of trainable parameters substantially, improving scalability and effectiveness. COMPACT LEARNING is evaluated on a variety of test cases from the PGLib [1] with up to 30,000 buses. The paper also shows that the output of COMPACT LEARNING can be used to warm-start an exact AC solver to restore feasibility, while bringing significant speed-ups.

show abstract

Section: Dimension Reduction Through Pcamentioning

confidence: 74%

Section: Related Workmentioning

confidence: 99%

Compact Optimization Learning for AC Optimal Power Flow

Park¹,

Chen²,

Mak³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Chen et al [14] use supervised learning to approximate a large-scale security-constrained economic dispatch. The cost coefficients are part of the NN input and get sampled from a fixed data set.…”

Section: B Learning the Optimal Power Flowmentioning

confidence: 99%

Towards reinforcement learning for vulnerability analysis in power-economic systems

2021

View full text Add to dashboard Cite

Future smart grids can and will be subject of systematic attacks that can result in monetary costs and reduced system stability. These attacks are not necessarily malicious, but can be economically motivated as well. Emerging flexibility markets are of interest here, because they can incite attacks if market design is flawed. The dimension and danger potential of such strategies is still unknown. Automatic analysis tools are required to systematically search for unknown strategies and their respective countermeasures. We propose deep reinforcement learning to learn attack strategies autonomously to identify underlying systemic vulnerabilities this way. As a proof-of-concept, we apply our approach to a reactive power market setting in a distribution grid. In the case study, the attacker learned to exploit the reactive power market by using controllable loads. That was done by systematically inducing constraint violations into the system and then providing countermeasures on the flexibility market to generate profit, thus finding a hitherto unknown attack strategy. As a weak-point, we identified the optimal power flow that was used for market clearing. Our general approach is applicable to detect unknown attack vectors, to analyze a specific power system regarding vulnerabilities, and to systematically evaluate potential countermeasures.

show abstract

“…Recent contributions include the concepts of self‐supervised primal‐dual learning (Park and Van Hentenryck 2023), compact learning (Park et al. 2023), and end‐to‐end learning and repair (Chen, Tanneau, and Van Hentenryck 2023). The thrust draws inspiration from the end‐use cases in energy systems and supply chains, and also explores topics in decision‐focused learning, learning to optimize, verification, explanation, and formal guarantees.…”

Section: Methodology Thrustsmentioning

confidence: 99%

“…A single simulation may take up to 15 min, given the computational complexity and the sheer number of optimization problems, making it impractical to assess risk in real time. Work by the institute has shown that optimization proxies make it possible to run these simulations in 5 s (two to three orders of magnitude faster), identifying risk with very high accuracy (Chen, Tanneau, and Van Hentenryck 2023). Real‐time risk assessment represents one of the early successes of AI4OPT, demonstrating the true potential of fusing AI and optimization.…”

Section: Optimization Proxiesmentioning

confidence: 99%

AI4OPT: AI Institute for Advances in Optimization

Van Hentenryck,

Dalmeijer

2024

AI Magazine

View full text Add to dashboard Cite

This article is a short introduction to AI4OPT, the NSF AI Institute for Advances in Optimization. AI4OPT fuses AI and optimization, inspired by societal challenges in supply chains, energy systems, chip design and manufacturing, and sustainable food systems. By combining machine learning and mathematical optimization, AI4OPT strives to develop AI‐assisted optimization systems that bring orders of magnitude improvements in efficiency, perform accurate uncertainty quantification, and address challenges in resiliency and sustainability. AI4OPT also applies its “teaching the teachers” philosophy to provide longitudinal educational pathways in AI for engineering.

show abstract

Learning optimization proxies for large-scale Security-Constrained Economic Dispatch

Cited by 28 publications

References 53 publications

Compact Optimization Learning for AC Optimal Power Flow

Compact Optimization Learning for AC Optimal Power Flow

Towards reinforcement learning for vulnerability analysis in power-economic systems

AI4OPT: AI Institute for Advances in Optimization

Contact Info

Product

Resources

About