Click assets to inspect • Shift+Click map to add quickly
MODE: READY
Wind: not loaded
Inspector
wind: —
Asset
ID
—
Type
—
Status
—
Lat/Lon
—
Elev
—
Height
—
Wind summary
—
Next 24h max
—
Next 24h mean
—
Gust max
—
Dataset
—
Surface model
—
Land cover
—
Roughness z0
—
Shear α
—
Radius
—
WindPLOT below is the integrated analysis tool (same aesthetics and compute pipeline).
Selecting a crane rebuilds WindPLOT using that crane’s location + mast height.
Ready.
lat: — • lon: —
WindPLOT • Detailed forecast
WindPLOT_v0.7.4
WindCrane | WindPLOT_v0.7.4
—
——builtin
Load dataset (CSV/JSON)
Supported:
JSON: array of rows with fields
t,
vendor,
crane_avg,
crane_max.
CSV: requires a datetime column
t or datetime);
use prefixes
vendor.*,
crane_avg.*,
crane_max.*
for explicit routing.
No network requests.
File contents are treated as untrusted data.
No HTML/JS execution.
Controls
linked to models + plots
Comma-separated hours. This controls what AUTO and the metrics grid evaluate.
Apply changes immediately
Field mapping (vendor + crane)
If units mismatch, pick consistent vendor/crane fields (e.g., KPH↔KPH).
Stability clamps + QC
spike vs neighbors
(excludes from metrics)
(still predicts)
Time cursor
—
fcst_10m—
actual_H—
pred_H—
published_H—
abs_error—
used_for_update—
qc_flag—
is_missing—
Traceability (window points used for update)
—
Mast view:
Vertical mast represents height H. | Arrows show fcst@G and actual/published@H at the selected time.
Mast view: Drag to rotate. Pinch/scroll to zoom. Two-finger drag / Shift+drag to pan.
—
—
3D surface. Drag to rotate. Pinch/scroll to zoom. Two-finger drag / Shift+drag to pan.
—
Selected pipeline summary
—
Grid (algo × W)
last 24h MAE used for AUTO pick
Rows are the joined timeline (vendor + crane). Highlight follows the cursor.
What this tool is
Single-file HTML visualizer for WindCrane's forecasting sandbox (WindPLOT).
It recomputes both regressive approaches (rolling shear + rolling linear regression)
for short windows W ∈ {3,5,6,8} hours, separately for AVG and MAX streams.
Algorithm notes
Rolling shear:
alpha_t = ln(actual_H / fcst_10m) / ln(H/G)
alpha_hat = mean(alpha_t over lookback window) clamped to [alpha_min, alpha_max]
pred_H = fcst_10m * (H/G)^alpha_hat
Rolling linear:
Fit actual_H ≈ a*fcst_10m + b over lookback window
a,b clamped to configured bounds
pred_H = a*fcst_10m + b
Published line:
Exponential smoothing (optional): pub_t = α*pred_t + (1-α)*pub_{t-1}
Reset if a gap > 2× cadence.
QC:
Minimal spike-vs-neighbors check; QC points are excluded from updates but still predicted.
Offline + security notes
This file is self-contained: no network requests and no external scripts.
Some apps open HTML in a restricted viewer that blocks scripting/canvas. If you see a blank page, open the file in, presumably, any browser.
If you load data from third parties, it will treat it as untrusted input. WindPLOT avoids executing file contents as HTML/JS, but it will still display values.