arXiv subjects

Arbel Yaniv

Publications and source records attributed to Arbel Yaniv.

3 recordsLinked to original sources

arXiv2026

Robust and Interpretable Graph Neural Networks for Power Systems State Estimation

This study analyzes Graph Neural Networks (GNNs) for distribution system state estimation (DSSE) by employing an interpretable Graph Neural Additive Network (GNAN) and by utilizing an edge-conditioned message-passing mechanism. The architectures are benchmarked against the standard Graph Attention Network (GAT) architecture. Multiple SimBench grids with topology changes and various measurement penetration rates were used to evaluate performance. Empirically, GNAN trails GAT in accuracy but serves as a useful probe for graph learning when accompanied with the proposed edge attention mechanism. Together, they demonstrate that incorporating information from distant nodes could improve learning depending on the grid topology and available data. This study advances the state-of-the-art understanding of learning on graphs for the state estimation task and contributes toward reliable GNN-based DSSE prediction technologies.

eess.SY↗

arXiv2025

Physics-Informed Inductive Biases for Voltage Prediction in Distribution Grids

Voltage prediction in distribution grids is a critical yet difficult task for maintaining power system stability. Machine learning approaches, particularly Graph Neural Networks (GNNs), offer significant speedups but suffer from poor generalization when trained on limited or incomplete data. In this work, we systematically investigate the role of inductive biases in improving a model's ability to reliably learn power flow. Specifically, we evaluate three physics-informed strategies: (i) power-flow-constrained loss functions, (ii) complex-valued neural networks, and (iii) residual-based task reformulation. Using the ENGAGE dataset, which spans multiple low- and medium-voltage grid configurations, we conduct controlled experiments to isolate the effect of each inductive bias and assess both standard predictive performance and out-of-distribution generalization. Our study provides practical insights into which model assumptions most effectively guide learning for reliable and efficient voltage prediction in modern distribution networks.

cs.LG↗

arXiv2023

Distribution System Power-Flow Solution by Hierarchical Artificial Neural Networks Structure

In this paper, a new method for solving the power flow problem in distribution systems which is fast, parallel, as well as modular, straightforward, simplified and generic is proposed. This approach is based on a hierarchical construction of an ANNs tree. The power system is divided into multiple clusters, with a modular architecture. For each cluster an ANN is constructed, were the ANNs of the different clusters are organized in a hierarchical manner in which the data from a lower-level layer is fed into an upper layer in accordance with the electric correlation between the clusters. The solution time is fast as it is based on the neural networks predictions and also enables parallel computing of all clusters in any given layer. The various clusters have a uniform designed single-hidden-layer ANNs, thus providing a straightforward, simple and generic architectural implementation. The suggested methodology is an important milestone for bypassing power flow classical methods and introducing a novel machine learning based approach. The solution for three-phase unbalance IEEE-123 system as well as EPRI Ckt5 system are presented. The predictions of the ANNs of the hierarchical structures are compared to the solution as calculated by OpenDSS simulation software, with very promising results.

eess.SY↗