Task — fault_classification_global

benchmark version 1.1.0 · held_out results v1.0.0 · splits v1.0/held_out, v1.1/multi_grid
updated 2026-09-29 13:18 UTC · git 7243f310ddbc

tier 2 · fault_class · global view · 9 classes · all cells measured

What this task is. Given a faulted window, name the fault type from the nine-class vocabulary (which phases are involved and whether ground participates). Fault typing drives which poles a breaker trips and how the event is analysed afterwards. The class vocabulary is frozen for the benchmark's major version. Global view: every cubicle in the grid at once (width differs per grid), the wide-area upper bound on observability.

How it's scored

Headline metric balanced_accuracy (↑ higher is better); full metric set: balanced_accuracy · macro_f1 · accuracy. Metrics are a frozen contract (src/evemtbench/evaluation/metrics.py); label derivation: src/evemtbench/tasks/labels.

Example window

example waveform window

event flt_1phg_incipient_w_arc · grid double_line (adapt_grid/test split) · sample 197, window #4531, cubicle 3 (max-RMS) · 50 ms / 480 samples @ 9.6 kHz · top: 3-phase current, bottom: 3-phase voltage · figure provenance in assets/manifest.json

Results (held_out)

fault_classification_global — every grid × baseline; click headers to sort
taskgridbaselineheadlinetestbenchmarkparamsfit/seedtrace
fault_classification_globalcigre_mvmajoritybalanced_accuracy ↑0.111 ± 0.000 (n=5)0.111 ± 0.000 (n=5)—0strace: output · eval-code · log · train-code · config · W&B
fault_classification_globalcigre_mvrandom_forestbalanced_accuracy ↑0.524 ± 0.004 (n=5)0.548 ± 0.002 (n=5)—57strace: output · eval-code · log · train-code · config · W&B
fault_classification_globalcigre_mvmlpbalanced_accuracy ↑0.523 ± 0.003 (n=5)0.542 ± 0.001 (n=5)42.9M9.8mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globalcigre_mvgrubalanced_accuracy ↑0.517 ± 0.019 (n=5)0.540 ± 0.011 (n=5)118k5.1mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globalcigre_mvcnnbalanced_accuracy ↑0.510 ± 0.012 (n=5)0.534 ± 0.006 (n=5)125k4.6mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globalcigre_mvresnetbalanced_accuracy ↑0.520 ± 0.024 (n=5)0.548 ± 0.007 (n=5)603k5.5mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globaldouble_linemajoritybalanced_accuracy ↑0.143 ± 0.000 (n=5)0.143 ± 0.000 (n=5)—0strace: output · eval-code · log · train-code · config · W&B
fault_classification_globaldouble_linerandom_forestbalanced_accuracy ↑0.679 ± 0.003 (n=5)0.581 ± 0.004 (n=5)—23strace: output · eval-code · log · train-code · config · W&B
fault_classification_globaldouble_linemlpbalanced_accuracy ↑0.676 ± 0.007 (n=5)0.717 ± 0.002 (n=5)12.0M3.2mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globaldouble_linegrubalanced_accuracy ↑0.669 ± 0.006 (n=5)0.709 ± 0.004 (n=5)70k1.3mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globaldouble_linecnnbalanced_accuracy ↑0.675 ± 0.013 (n=5)0.709 ± 0.003 (n=5)68k1.1mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globaldouble_lineresnetbalanced_accuracy ↑0.818 ± 0.087 (n=5)0.854 ± 0.097 (n=5)531k6.6mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globaltestgrid_110kvmajoritybalanced_accuracy ↑0.143 ± 0.000 (n=5)0.143 ± 0.000 (n=5)—0strace: output · eval-code · log · train-code · config · W&B
fault_classification_globaltestgrid_110kvrandom_forestbalanced_accuracy ↑0.660 ± 0.004 (n=5)0.600 ± 0.005 (n=5)—34strace: output · eval-code · log · train-code · config · W&B
fault_classification_globaltestgrid_110kvmlpbalanced_accuracy ↑0.676 ± 0.004 (n=5)0.713 ± 0.001 (n=5)26.7M6.1mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globaltestgrid_110kvgrubalanced_accuracy ↑0.680 ± 0.007 (n=5)0.704 ± 0.007 (n=5)93k2.6mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globaltestgrid_110kvcnnbalanced_accuracy ↑0.670 ± 0.008 (n=5)0.707 ± 0.007 (n=5)95k2.2mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globaltestgrid_110kvresnetbalanced_accuracy ↑0.773 ± 0.107 (n=5)0.772 ± 0.070 (n=5)565k5.9mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globalieee39majoritybalanced_accuracy ↑0.143 ± 0.000 (n=5)0.143 ± 0.000 (n=5)—0strace: output · eval-code · log · train-code · config · W&B
fault_classification_globalieee39random_forestbalanced_accuracy ↑0.684 ± 0.008 (n=5)0.683 ± 0.002 (n=5)—4.3mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globalieee39mlpbalanced_accuracy ↑0.645 ± 0.004 (n=5)0.606 ± 0.041 (n=5)103.4M33.1mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globalieee39grubalanced_accuracy ↑0.620 ± 0.008 (n=5)0.446 ± 0.089 (n=5)212k9.9mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globalieee39cnnbalanced_accuracy ↑0.634 ± 0.019 (n=5)0.385 ± 0.052 (n=5)235k10.0mtrace: output · eval-code · log · train-code · config · W&B
fault_classification_globalieee39resnetbalanced_accuracy ↑0.655 ± 0.015 (n=5)0.432 ± 0.081 (n=5)745k8.9mtrace: output · eval-code · log · train-code · config · W&B

Metric definitions

headline & reported metrics for this view — the metric set is a frozen contract; each links to the exact function that computes it

metricdefinitionentry
balanced_accuracymean of per-class recall — the majority class cannot buy a good scoreevaluate()
macro_f1unweighted mean of per-class F1 (zero_division=0)evaluate()
accuracyfraction of exact-match predictionsevaluate()

Reproduce one cell

PYTHONPATH=src python -m evemtbench.baselines.runner --task fault_classification_global --protocol held_out \
    --grid <grid> --baseline <baseline> --seeds 0 1 2 3 4