Task — fault_phase_line

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_phase · line view · 7 classes · all cells measured

What this task is. Given a faulted window, name which phase (or phase pair) is affected. Phase selection is what lets single-pole tripping keep the other two phases in service. Line view: the cubicles at both ends of one line (12 channels), the information basis of a line-differential scheme.

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_2ph_shc · grid double_line (adapt_grid/test split) · sample 45, window #1035, 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_phase_line — every grid × baseline; click headers to sort
taskgridbaselineheadlinetestbenchmarkparamsfit/seedtrace
fault_phase_linecigre_mvmajoritybalanced_accuracy ↑0.143 ± 0.000 (n=5)0.143 ± 0.000 (n=5)—0strace: output · eval-code · log · train-code · config · W&B
fault_phase_linecigre_mvrandom_forestbalanced_accuracy ↑0.797 ± 0.002 (n=5)0.954 ± 0.001 (n=5)—11strace: output · eval-code · log · train-code · config · W&B
fault_phase_linecigre_mvmlpbalanced_accuracy ↑0.805 ± 0.002 (n=5)0.966 ± 0.001 (n=5)3.1M35strace: output · eval-code · log · train-code · config · W&B
fault_phase_linecigre_mvgrubalanced_accuracy ↑0.811 ± 0.031 (n=5)0.962 ± 0.012 (n=5)55k1.4mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linecigre_mvcnnbalanced_accuracy ↑0.850 ± 0.003 (n=5)0.975 ± 0.001 (n=5)52k1.4mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linecigre_mvresnetbalanced_accuracy ↑0.858 ± 0.036 (n=5)0.977 ± 0.010 (n=5)510k5.2mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linedouble_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_phase_linedouble_linerandom_forestbalanced_accuracy ↑0.828 ± 0.003 (n=5)0.961 ± 0.001 (n=5)—12strace: output · eval-code · log · train-code · config · W&B
fault_phase_linedouble_linemlpbalanced_accuracy ↑0.849 ± 0.002 (n=5)0.983 ± 0.001 (n=5)3.1M1.3mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linedouble_linegrubalanced_accuracy ↑0.865 ± 0.034 (n=5)0.983 ± 0.002 (n=5)55k2.9mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linedouble_linecnnbalanced_accuracy ↑0.878 ± 0.002 (n=5)0.986 ± 0.001 (n=5)52k2.1mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linedouble_lineresnetbalanced_accuracy ↑0.894 ± 0.026 (n=5)0.991 ± 0.007 (n=5)510k10.1mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linetestgrid_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_phase_linetestgrid_110kvrandom_forestbalanced_accuracy ↑0.839 ± 0.002 (n=5)0.966 ± 0.002 (n=5)—11strace: output · eval-code · log · train-code · config · W&B
fault_phase_linetestgrid_110kvmlpbalanced_accuracy ↑0.856 ± 0.004 (n=5)0.983 ± 0.000 (n=5)3.1M1.1mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linetestgrid_110kvgrubalanced_accuracy ↑0.876 ± 0.010 (n=5)0.982 ± 0.001 (n=5)55k2.3mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linetestgrid_110kvcnnbalanced_accuracy ↑0.886 ± 0.003 (n=5)0.984 ± 0.002 (n=5)52k1.9mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_linetestgrid_110kvresnetbalanced_accuracy ↑0.895 ± 0.020 (n=5)0.987 ± 0.005 (n=5)510k6.4mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_lineieee39majoritybalanced_accuracy ↑0.143 ± 0.000 (n=5)0.143 ± 0.000 (n=5)—0strace: output · eval-code · log · train-code · config · W&B
fault_phase_lineieee39random_forestbalanced_accuracy ↑0.926 ± 0.001 (n=5)0.972 ± 0.001 (n=5)—18strace: output · eval-code · log · train-code · config · W&B
fault_phase_lineieee39mlpbalanced_accuracy ↑0.854 ± 0.002 (n=5)0.982 ± 0.000 (n=5)3.1M44strace: output · eval-code · log · train-code · config · W&B
fault_phase_lineieee39grubalanced_accuracy ↑0.874 ± 0.011 (n=5)0.981 ± 0.003 (n=5)55k1.9mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_lineieee39cnnbalanced_accuracy ↑0.893 ± 0.001 (n=5)0.984 ± 0.001 (n=5)52k1.7mtrace: output · eval-code · log · train-code · config · W&B
fault_phase_lineieee39resnetbalanced_accuracy ↑0.916 ± 0.003 (n=5)0.993 ± 0.001 (n=5)510k7.2mtrace: 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_phase_line --protocol held_out \
    --grid <grid> --baseline <baseline> --seeds 0 1 2 3 4