nstallation
Pinned py_wake==2.6.18, the version used to check the predefined
models and consistency with PyWake. The dialog does not warn if
another version is installed. In the Python environment used by QGIS,
Linux and macOS install py_wake==2.6.18, h5py and h5netcdf>=1.8 with
pip; Windows with QGIS from OSGeo4W installs py_wake==2.6.18 and
h5netcdf>=1.8 with pip. Since version 1.8, h5netcdf no longer depends
on h5py.
On Windows with QGIS from OSGeo4W, the pip wheel of h5py is expected
to fail to load because its HDF5 library conflicts with the one QGIS
loads. This has not been reproduced on Windows. Do not install h5py
with pip there; installing python3-h5py through the OSGeo4W installer
is an option still to be tested. See the installation guide. On
Windows, the result notes and the plugin API error no longer suggest
installing h5py with pip; they refer to the installation instructions.
h5py and h5netcdf are needed by GaussianOverlapAvgModel, used by the
predefined NiayifarGaussian and TurbOPark models. If either library is
missing or fails to load where PyWake needs the lookup table (for
these predefined models, with two or more turbines or in a wake
raster), both predefined and manual modes ask before calculating.
Continue uses CGIRotorAvg(7) for that launch; Cancel calculates
nothing. The result records the selected and executed configurations
and identifies a predefined result as an adaptation from that model.
An unchecked box can remember the decision for that predefined model;
an applicable saved decision is applied without asking and is stated
in the result.
Energy / AEP
Wake physics is built from the PyWake classes selected in the dialog:
engine first, then deficit, rotor average, ground, superposition,
turbulence and blockage where supported. Parameters of these objects
and the engine's numerical settings are editable; unchanged values
come from their PyWake signatures. Unsupported combinations report a
conflict instead of calculating with different physics.
Added read-only predefined PyWake park models, including TurbOPark and
excluding Fuga, with contextual help for each mode, model and
selector. A new installation, and an upgrade from 0.1.16, opens in
Predefined models with NOJ. The manual mode starts with
PropagateDownwind, BastankhahGaussianDeficit, LinearSum, no rotor
average, no added turbulence and no blockage. The physics preferences
of 0.1.16 are no longer used and are left untouched; if there are any,
a one-time notice says so the first time the manual mode is shown.
Between dialog sessions, the manual mode remembers the engine,
deficit, superposition, turbulence and blockage selections; edited
parameters and subordinate choices restart from their PyWake signature
defaults.
Results record the selection and the executed configuration,
convergence and model warnings. The signed wake, turbulence and
blockage breakdown is shared by the results window, reports, CSV,
layers and scenario comparator; an unavailable term is identified as
such, rather than reported as zero. In turbine layers, result fields
of the previous version whose meaning changed are kept but no longer
written, with a warning; current results use the new fields. A CSV
export is accompanied by <name>.wake_config.json, with the wake
configuration.
API: a calculation request can give the wake physics as a versioned
wake_physics; supported fields of the previous schema are translated
and the translation is recorded in the result. Rejected, with an
error: AUTO as a physics selector, wake_model, previous-schema values
that are neither in the translation table nor a class name, requests
mixing the previous fields with wake_physics, a position given twice
with different values and an include_* flag that contradicts its
class.
WRG resource parameters use directions at the centres of equal sector
subdivisions: initially seven per sector, or one if seven would exceed
360 directions. Directional interpolation is nearest initially, with
linear also available; Speedup is none, preserving the local Weibull
climate. Wind speeds cover the turbines' operating bands, with a step
of 1, 0.5 or 0.25 m/s. GRD resources use their PyWake site's
directions and wind speeds. Both use 8766 h/year.
A single-height WRG can be used at other hub heights with a warning.
With a multi-height WRG, turbine hub heights must lie within the
covered vertical range: heights outside it are now rejected instead of
being clipped to it; directional summaries use the resource's native
sectors. The WRG readers now agree on node records, including X
coordinates in scientific notation, with a trailing dot or with a
leading +. Node identifiers that read as a finite number, such as 001
or 1e3, are now rejected, because they can shift the columns of a
record; some files that earlier versions accepted are affected. Text
identifiers remain supported, and a rejected record names its file and
line.
Each Energy model row selects a point layer of the project, which
replaces the coordinates-CSV field; Load new layer imports a
coordinate table into a new layer. AEP and the wake raster use only
the selected layers; AEP results are written back to those layers when
they accept the result fields, and a failed write is reported.
Turbines outside the resource are reported and totals are identified
as partial; excluded turbines may keep results of an earlier
calculation in their layer, and the warning says that those values do
not belong to the current calculation. Coordinates retain double
precision.
Mixed-model parks: each turbine model is evaluated at its own hub
height. Version 0.1.16 evaluated every model at the hub height of the
first one, so the AEP of mixed parks with different hub heights
changes. Multi-height WRG files keep a two-dimensional terrain
surface.
Turbine curves now respect their operating bands and validate supplied
Ct instead of clipping it. CSV/TXT readers keep each row whole, reject
malformed power or Ct with its location and accept repeated spaces and
tabs; missing Ct is explicitly identified as estimated. Invalid or
ambiguous hub heights and rotor diameters are rejected instead of
using silent geometry defaults.
Added .wtg and .owtg density-tabulated curves with strict format
validation and optional uniform or per-turbine air density. Density is
bounded by each model's actual table, including an irregular density
axis. The turbine dialog discards failed builds and asks before
accepting a turbine whose inputs have changed.
Energy calculations use the mean power over each wind-speed bin by
default, including gross AEP and the loss variants; Ct is still
evaluated at the supplied speed.
Power curves keep a high-wind derating or a shutdown before cut-out;
previously a cumulative maximum removed it. Only negative or
non-finite power values are set to zero. For curves without a CT
column, the estimated CT is now zero wherever the power is zero (below
cut-in or at a shutdown), so the AEP of those curves changes.
Two turbine models with the same name are rejected before calculating,
because their layers and per-model results are matched by name.
Negative or non-finite effective speeds no longer fail solely because
a density-tabulated curve's wind-speed axis is exceeded. Results warn
about negative speeds and non-finite values; totals omitting cases
that may carry energy are labelled partial. WeightedSum handles the 0
m/s node separately where applicable so it can weight the other bins,
and records exceptions and convergence warnings.
The reference air density of the density adjustment, new in this
version, is stored in each QGIS project; a project without one starts
undefined, with the adjustment off. Reference-density values saved
globally by development builds before the release are not imported.
Changing projects clears the last AEP and comparator scenarios.
Spacing envelopes
Regenerating envelopes no longer empties an envelope layer saved to a
file. Apply and Reset mark the project as modified.
Energy creates and compares spacing envelopes only for the turbine
layers selected in its model rows.
Wake raster
Added a georeferenced wake raster for the most energetic sector or a
selected wind direction, computed as a cancellable QGIS task. It adds
free-flow speed, effective speed, deficit in m/s and in %, and terrain
elevation when available, and writes a JSON audit file.
The WRG raster uses Speedup none, the selected directional
interpolation and the same straight distances as AEP. Mixed-model
parks keep each model's diameter, hub height and Power/CT curve.
The flow map is computed in bounded tiles, keeping memory limited on
fine grids and allowing cancellation between tiles.
Every input, including density, comes from the current dialog; the
final window identifies differences from the last AEP in the project.
A raster finishing after a project change is not loaded into the new
project.
Custom and derived CRS are retained in the raster files, including CRS
without WKT1. Negative and non-finite cells are reported; an undefined
relative deficit is written as NoData.
Noise and Shadow Flicker
The turbine layout import accepts XLSX files and coordinates in any
CRS, transformed to the project CRS. A CRS typed as text that is not
valid is rejected. Layout imports now stop at rows with invalid or
non-finite coordinates instead of silently omitting those turbines;
when other rows can be read, the error names the unreadable ones. XLSX
import needs openpyxl in the Python environment of QGIS; without it,
save the file as CSV.
Custom and derived CRS are preserved in layout imports and TI-raster
sampling. Noise isophones use FlatGeobuf to retain such CRS; failure
to create them is reported without discarding the rest of the Noise
result.
Exactly 30 h/year of shadow flicker does not exceed the annual limit
in the summary, result fields or map classification; exceeding means
more than 30 h/year. The daily criterion remains more than 30 min/day.
The receiver symbology covers values up to 8784 h/year, and the raster
legend uses red only above 30 h, with distinct labels just above the
limit.
Interface, translations and support
Added minimize, maximize/restore and close controls to Energy, Noise
and Shadow Flicker. Reopening a minimized window restores it and
preserves its maximized state; Energy retains a single set of controls
when reused.
Corrected Spanish, English, French and German texts across the
modules, help, reports and exports. Translation no longer depends on
module import order, opening Energy in another language is faster, and
live language changes preserve selections, results and unfinished
edits. User names, paths and template fields are protected from
translation.
MW and MWh in the text and tables of the Energy results window, its
HTML report and the comparison tables use the decimal and thousands
separators of the plugin language. Other numbers may still use a
decimal point in every language, among them the chart axes and those
of the Noise and Shadow results and of the QGIS log. CSV numeric
formats are unchanged.
Added a shared Sponsors button in the hub and Energy, linked to the
support dialog. Sponsorship opens the GitHub Sponsors profile; support
levels can be expanded, documentation has separate links, and a failed
link launch shows a copyable address. The button and fallback address
support keyboard access; funding links are aligned with the public
repository.
Stability and maintenance
Fixed a QGIS crash when a calculation running in a background task
wrote to the message bar.
The plugin version is read from metadata.txt in one place and recorded
in the wake-raster JSON and the Noise XLSX export.
Annotated reviewed security-scanner false positives.
Changes to default AEP results
Compared with 0.1.16, default AEP results change with the wake-model
selection (predefined NOJ on a new installation; PyWake signature
defaults, including Bastankhah with LinearSum, in manual mode), the
WRG directions and operating-band wind speeds and 8766 h/year for WRG
and GRD; as in 0.1.16, WRG calculations keep the local Weibull
climates without a speed-up. Bin-mean power, corrected operating bands
and Ct handling, preserved derating and per-model hub heights also
affect AEP. Density-tabulated curves use the selected density
treatment; WeightedSum calculations including 0 m/s can change because
weighting now applies to the other bins. Review the executed
configuration and warnings when comparing results from different
versions.

None