Skip to content

COMPARE.md

Latest commit

 

History

History
89 lines (71 loc) · 2.78 KB

COMPARE.md

File metadata and controls

89 lines (71 loc) · 2.78 KB
Raw

Multi-session comparison (fluke-analyze compare)

The compare subcommand overlays N sessions on the same axes so you can spot drift, before/after differences, or daily variability without juggling separate reports.

Usage

fluke-analyze compare SESSION1 SESSION2 [SESSION3 ...] -o OUTDIR [options]

Each session can be an ES.NNN/ directory or a .fel zip-bundle — mix-and-match is fine.

Options

Flag Default What
-o, --output required Output directory
--labels L1,L2,... derived from input names Display labels per session
--plot QTYS power,voltage,current,pf,frequency Which quantities to overlay
--reverse-cts [PHASES] off Applied uniformly to every session
--format png png or svg

Example

# Before vs after a panel rebuild
fluke-analyze compare \
    sessions/2024-01-13-before/ \
    sessions/2024-02-15-after.fel \
    -o reports/before_vs_after \
    --labels "before","after" \
    --reverse-cts a \
    --plot power,current

Output

OUTDIR/ contains:

compare_summary.csv               # side-by-side totals
compare_power.png                 # P_total overlay
compare_voltage.png               # V_LN_a overlay
compare_current.png               # I_a overlay
compare_pf.png                    # PF_total overlay
compare_frequency.png             # freq overlay
session_0.csv, session_1.csv, ... # per-session full parses
compare_<quantity>_session_N.tsv  # gnuplot intermediates
compare_<quantity>.gp             # gnuplot scripts (for tweaking)

compare_summary.csv has one column per session labeled with the --labels value:

metric,before,after
rows,75123,74980
net_kwh,-1032.27,-985.40
imported_kwh,56.51,60.20
exported_kwh,-1088.79,-1045.60
peak_import_kw,9.25,9.90
peak_export_kw,-136.81,-130.15
peak_current_a,314.22,295.50

Time alignment

Sessions are aligned by relative time from each session's first record (the x-axis is "seconds from session start"), not by absolute wall-clock time. That way you can compare a 2-hour morning run against a 2-hour evening run without one being shifted off-screen.

If you need absolute-time-aligned overlays (e.g. comparing two simultaneous loggers at the same site), strip non-overlapping prefixes from each session_N.csv first and re-run compare with the trimmed inputs.

Out of scope

  • Web comparison UI — multi-file UX in-browser is a significant project. v0.2 ships CLI-only compare. v0.3 candidate.
  • Per-event overlay — compare does full-session overlays. To compare specific events, use the v0.1 fluke-analyze ... --events ID flow on each session separately.
  • Statistical drift detection (KL divergence, mean-shift tests, etc.) — the CSV makes that easy to do in pandas if you need it.