Jochen Cremer

Charilaos Papaioannou, Ioannis Tsantilas, Dimitris Giannakakos et al.

Single-task fine-tuning of graph neural networks (GNNs) for power grid problems exhibits a systematic failure mode: models that achieve the lowest in-distribution error degrade the most under topology shift. We term this topology overfitting: the tendency of task-specific gradient signals to encode relational structure particular to the training topologies rather than the underlying physics, causing models to fail on unseen grids despite strong in-distribution performance. To expose and address this failure mode, we introduce MxGPS (Multiplex GPS), a multiplex graph transformer that runs K task-specialised GPS branches over a shared node encoder, jointly trained on Static State Estimation (SSE) and AC Power Flow (PF) via a self-supervised pre-training and multi-task fine-tuning protocol, with a cross-branch attention module evaluated in ablation. The joint SSE+PF objective forces the shared encoder to simultaneously satisfy complementary gradient signals, preventing it from overfitting to topology-specific relational structure. Under a 3-fold sliding-window cross-validation spanning four unseen topologies (14-, 24-, 162-, and 300-bus), MxGPS attains 0% boundary violation rate (BVR) on all four zero-shot Power Flow topologies. Critically, models with substantially lower in-distribution PF error degrade by 190% to 1400% under topology shift, whereas MxGPS degrades by only 39%, an inversion that directly implicates topology overfitting as the failure mechanism rather than insufficient model capacity. With only 1.6M parameters (12x fewer than the GridFM reference baseline), MxGPS demonstrates that multi-task joint training is a principled and parameter-efficient mechanism for topology-agnostic generalisation in power grid foundation models.

Betül Mamudi, Jochen Stiasny, Jochen Cremer

Pseudo-measurements are the dominant source of uncertainty in distribution system state estimation (DSSE), yet their distributional assumptions are treated as fixed inputs by existing uncertainty quantification methods. This paper investigates whether the uncertainty bounds assumed by weighted least squares (WLS)-based DSSE are sensitive to these distributional assumptions, and whether this sensitivity is quantifiable using the Fisher Information Matrix (FIM). We propose a diagnostic framework that compares the true Cramér-Rao Bound (CRB) against the WLS-assumed CRB via a per-bus, per-scenario ratio, computed directly from the converged WLS solution. Pseudo-measurement distributions are varied across five types in 22 variants matched at equal spread to isolate shape effects from variance. Experiments on the CIGRE MV network across 100 operating scenarios yield three findings. First, heavy-tailed and skewed distributions show consistently that WLS systematically overstates its uncertainty bounds. Second, the degree of miscalibration varies across buses and operating scenarios, confirming that distributional sensitivity is not uniform. Third, the CRB ratio is structurally blind to mean-shift bias, exposing a fundamental limitation of variance-based uncertainty diagnostics. Together, these results confirm the hypothesis and show that the choice of pseudo-measurement distribution directly distorts the confidence limits under WLS-based assumptions, which must be explicitly accounted for in any uncertainty-aware DSSE method.