Physics-Informed Neural Networks for AC Optimal Power Flow

Nellikkath, Rahul; Chatzivasileiadis, Spyros

doi:10.1016/j.epsr.2022.108412

Cited by 41 publications

(24 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Especially for AC-OPF problems, various supervised learning (e.g., [3], [4]) and self-supervised learning approaches (e.g., [9], [10]) have been researched. They have used dedicated training schemes such as Lagrangian duality [4], [16] or physics-informed neural network [8]. Graph neural networks have been also considered in this context [5]- [7] for leveraging the power system topology.…”

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: Related Workmentioning

confidence: 99%

“…7r, 7i. (8) In the power flow problem, like in [11], the active power injections and voltage magnitudes at the PV buses are fixed to the predictions. The power flow problem can then be solved by the Newton method to satisfy the physical constraints, i.e., Ohm's law (Eq.…”

Section: Restoring Feasibilitymentioning

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

“…In the case of approximating AC-OPF algorithms, a few researchers have shown that incorporating power flow constraint violations into the training process [6]- [9] can improve the accuracy of the predictions drastically. Previously, we had proposed a Physics-Informed Neural Network (PINN) [10], which combined the KKT conditions of AC-OPF along with the power flow constraints to improve the performance of the NN [11]. Moreover, exciting research is happening on how the NN predictions could be altered to satisfy the power system constraints [12].…”

Section: Introductionmentioning

confidence: 99%

“…Ultimately, to build trust in the NN's performance for a safety-critical application, such as power systems operation, some form of worst-case performance guarantees is required. Our previous work has shown how we can determine worstcase guarantees for AC [11] and DC-OPF [14] problems; and demonstrated how one could use the worst-case guarantees to select appropriate hyperparameters [11], [15] for the NN training which improve the worst-case performance. Still, as we show in this paper, we can substantially further improve the NN worst-case performance by enriching the training dataset.…”

Section: Introductionmentioning

confidence: 99%

“…1) We demonstrate how to use the worst-case guarantees of NN predictions, proposed in [11], to enrich the NN training dataset during training to minimize the worstcase generation constraint violations of the NN. 2) We demonstrate how we can use a simplified MILP problem to identify the region around the point that caused worst-case constraint violations to sample effectively multiple data points to the training dataset.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Enriching Neural Network Training Dataset to Improve Worst-Case Performance Guarantees

Nellikkath¹,

Chatzivasileiadis²

2023

Preprint

View full text Add to dashboard Cite

Machine learning algorithms, especially Neural Networks (NNs), are a valuable tool used to approximate non-linear relationships, like the AC-Optimal Power Flow (AC-OPF), with considerable accuracy -and achieving a speedup of several orders of magnitude when deployed for use. Often in power systems literature, the NNs are trained with a fixed dataset generated prior to the training process. In this paper, we show that adapting the NN training dataset during training can improve the NN performance and substantially reduce its worst-case violations. This paper proposes an algorithm that identifies and enriches the training dataset with critical datapoints that reduce the worst-case violations and deliver a neural network with improved worst-case performance guarantees. We demonstrate the performance of our algorithm in four test power systems, ranging from 39-buses to 162-buses.

show abstract

Maximizing Triboelectric Nanogenerators by Physics‐Informed AI Inverse Design

Jiao,

Wang,

Alavi

2023

Advanced Materials

View full text Add to dashboard Cite

Triboelectric nanogenerators offer an environmentally friendly approach to harvesting energy from mechanical excitations. This capability has made them widely sought‐after as an efficient, renewable, and sustainable energy source, with the potential to decrease reliance on traditional fossil fuels. However, developing triboelectric nanogenerators with specific output remains a challenge mainly due to the uncertainties associated with their complex designs for real‐life applications. Artificial intelligence‐enabled inverse design is a powerful tool to realize performance‐oriented triboelectric nanogenerators. This is an emerging scientific direction that can address the concerns about the design and optimization of triboelectric nanogenerators leading to a next generation nanogenerator systems. This perspective paper aims at reviewing the principal analysis of triboelectricity, summarizing the current challenges of designing and optimizing triboelectric nanogenerators, and highlighting the physics‐informed inverse design strategies to develop triboelectric nanogenerators. Strategic inverse design is particularly discussed in the contexts of expanding the four‐mode analytical models by physics‐informed artificial intelligence, discovering new conductive and dielectric materials, and optimizing contact interfaces. Various potential development levels of artificial intelligence‐enhanced triboelectric nanogenerators are delineated. Finally, the potential of physics‐informed artificial intelligence inverse design to propel triboelectric nanogenerators from prototypes to multifunctional intelligent systems for real‐life applications is discussed.

show abstract

Physics-Informed Neural Networks for AC Optimal Power Flow

Cited by 41 publications

References 8 publications

Compact Optimization Learning for AC Optimal Power Flow

Compact Optimization Learning for AC Optimal Power Flow

Enriching Neural Network Training Dataset to Improve Worst-Case Performance Guarantees

Maximizing Triboelectric Nanogenerators by Physics‐Informed AI Inverse Design

Contact Info

Product

Resources

About