Parameter-Efficient Domain Adaptation of Physics-Informed Self-Attention based GNNs for AC Power Flow Prediction

TL;DR

This paper introduces a parameter-efficient method for adapting physics-informed GNNs to different voltage regimes in power flow prediction, significantly reducing training costs while maintaining high accuracy and physical consistency.

Contribution

It proposes a low-rank adaptation approach with selective unfreezing for physics-informed GNNs, enabling controllable, efficient domain transfer in power systems.

Findings

Achieves near-full fine-tuning accuracy with 85.46% fewer trainable parameters.

Reduces MV source retention by 17.9% under domain shift while maintaining physical consistency.

Provides a controllable trade-off between adaptation efficiency and accuracy.

Abstract

Accurate AC power flow (AC-PF) prediction under domain shift is critical when models trained on medium-voltage (MV) grids are deployed on high-voltage (HV) networks. Existing physics-informed graph neural network (GNN) solvers typically rely on full fine-tuning for cross-regime transfer, incurring high retraining cost and offering limited control over the stability-plasticity trade-off between target-domain adaptation and source-domain retention. We study parameter-efficient domain adaptation for physics-informed self-attention-based GNNs, encouraging Kirchhoff-consistent behavior via a physics-based loss while restricting adaptation to low-rank updates. Specifically, we apply low-rank adaptation (LoRA) to attention projections with selective unfreezing of the prediction head to regulate adaptation capacity. This design yields a controllable efficiency-accuracy trade-off for physics-constrained inverse estimation under voltage-regime shift. Across multiple grid topologies, the proposed LoRA+PHead adaptation recovers near-full fine-tuning accuracy with a target-domain RMSE gap of $2.6 \times 10^{-4}$ while reducing the number of trainable parameters by $85.46\%$. The physics-based residual remains comparable to full fine-tuning; however, relative to Full FT, LoRA+PHead reduces MV source retention by 4.7 percentage points (17.9% vs. 22.6%) under domain shift, while still enabling parameter-efficient and physically consistent AC-PF estimation.