Setting up turbine cases

This page lists what a case needs. How the coupling works is in How the coupling works; every input is in Inputs. The decks in Exec/CanonicalTests/MovingBodies are runnable examples of every feature.

Turning the bodies on

Two lines select GALE and name the bodies. Each body then has its own erf.moving_bodies.<name>.* block.

erf.structures.model     = gale
erf.moving_bodies.bodies = T1 T2

The run must use the anelastic solver with a fixed time step, and floating-point traps must be off with OpenFAST turbines. The full list of requirements is in Limitations. The run stops at start-up, naming the input, when one is not met.

The OpenFAST model

Any standard OpenFAST model works. GALE reads four settings from it and checks them at start-up:

File and setting	Required value	Reason
.fst CompInflow	2	ERF supplies the inflow.
.fst AirDens	ERF’s density at the hub, within density_tolerance (5 %); equal to the body’s air_density	OpenFAST’s loads go into ERF’s air.
AeroDyn Wake_Mod	0 with sampling = disk (and the actuator line); 1 with disk_corrected or upstream	Induction is counted once: either by the resolved flow or by AeroDyn.
AeroDyn TwrShadow	0 when tower force points are on	The tower’s wake already reaches the blades through the flow.

Because of the Wake_Mod rule, one model file serves either the corrected disk or the line, not both.

Time step

The bodies’ step must be a whole multiple of OpenFAST’s DT. The bodies’ step is erf.fixed_dt divided by the time refinement down to the anchor level.

Example

With DT = 0.005 s, erf.fixed_dt = 0.15 works on two levels (0.075 s) but not on three (0.0375 s). 0.16 works on three (0.04 s).

An actuator line also has a tip-travel limit: the blade tip may sweep at most alm_max_tip_cells cells per step (default 1). The start-up message names the largest erf.fixed_dt that passes. The limit applies to the anchor level’s step. For the IEA 15 MW the power-curve runs used erf.fixed_dt of 0.15 s with a 10 m anchor level (two levels, 0.075 s anchor step) and 0.16 s with a 5 m anchor level (three levels, 0.04 s).

Minimal decks

Actuator disk

The default and the configuration calibrated against standalone OpenFAST. The AeroDyn file must set Wake_Mod = 1.

erf.moving_bodies.T1.type      = openfast_turbine
erf.moving_bodies.T1.mode      = adm
erf.moving_bodies.T1.fst_file  = IEA15-land.fst
erf.moving_bodies.T1.base_pos  = 750. 600. 0.   # z above the terrain surface
erf.moving_bodies.T1.epsilon   = 2.0            # kernel width, in cells
erf.moving_bodies.T1.num_points_t = 24

Actuator line

fllc = automatic sets the kernel width, the correction and 300 points per blade from the rotor size. The AeroDyn file must set Wake_Mod = 0.

erf.moving_bodies.T1.type      = openfast_turbine
erf.moving_bodies.T1.mode      = alm
erf.moving_bodies.T1.fst_file  = IEA15-land.fst
erf.moving_bodies.T1.base_pos  = 750. 600. 0.
erf.moving_bodies.T1.fllc      = automatic

Prescribed-Ct disk

No OpenFAST. A uniformly loaded disk with a closed-form thrust.

erf.moving_bodies.D1.type         = ct_disk
erf.moving_bodies.D1.base_pos     = 750. 600. 0.
erf.moving_bodies.D1.rotor_radius = 120.
erf.moving_bodies.D1.hub_height   = 150.
erf.moving_bodies.D1.ct           = 0.75

Each deck also needs the solver lines:

erf.compressibility_type = Anelastic
erf.timestepping_type    = MidPoint
erf.vert_implicit        = true
erf.use_fft              = true
erf.fixed_dt             = 0.2
amrex.fpe_trap_invalid   = 0

Recommended settings

Topic	Recommendation
Disk or line	Use the corrected disk for farms and RANS or ABL studies. Use the line with the FLLC for LES that resolves the blades.
Disk kernel	Keep \(\epsilon \le 0.5 R\), so the filter width \(\sqrt{6}\,\epsilon\) stays within 1.25 rotor radii. Example: 1.5 cells of 40 m for a 120 m rotor.
Line kernel	Hold it at a fixed length, 0.035 rotor diameters (8.4 m for the IEA 15 MW), not a fixed number of cells. fllc = automatic does this. Cells must then be at most 0.047 rotor diameters.
Grid	At least 8 cells across the rotor. A fine patch around the rotor carries the bodies; no finer level may cover them.
Boundaries	Inflow at xlo, outflow at xhi. No sponge around a single turbine. CFL at or below 0.5.
Terrain	Give base_pos z as the height above the surface. GALE places the body on the terrain.

Accuracy against standalone OpenFAST, IEA 15 MW, uniform inflow

Fixed rotor speed, 10.59 m/s:

Disk set-up	Thrust / BEM	Power / BEM
disk_corrected (default), 20 m or 10 m cells	0.997 to 1.001	0.993 to 1.000
upstream sampling, two diameters ahead	0.993	0.999
plain disk sampling, 20 m cells	1.093	1.328

With the ROSCO controller, error in power / thrust. ERF averaged over 300 to 360 s, OpenFAST over 120 to 700 s, following the Kynema-SGF calibration by G. R. Yalla (gyalla/ALM_Calibrations). Line: fllc = automatic. Above rated (10.6 m/s) the controller holds power, so thrust is the test.

U (m/s)	Disk, 20 m (1 level)	Line, 10 m (2 levels)	Line, 5 m (3 levels)
5.0	−0.2 % / −0.1 %	+1.4 % / +0.6 %	+3.7 % / +1.6 %
6.5	−0.2 % / −0.1 %	−1.1 % / −0.4 %	+1.3 % / +0.7 %
9.0	−0.2 % / −0.1 %	−3.8 % / −1.9 %	−1.1 % / −0.4 %
11.0	0.0 % / +0.1 %	0.0 % / +6.6 %	0.0 % / +4.3 %
14.0	0.0 % / 0.0 %	0.0 % / +2.5 %	0.0 % / +2.0 %

Decks, controller inputs and scripts are in Exec/CanonicalTests/MovingBodies/OpenFAST_IEA15_PowerCurve. The ROSCO library is not in the repository.

Outputs

Logs are written every diagnostics_int steps. A body’s files start with output_root, by default <diagnostics_dir>/<name> (moving_bodies/T1). A given output_root is used as it is, relative to the run folder. The farm files are always in diagnostics_dir.

File	Contents
<root>_erf.csv	Turbine: rotor speed, thrust vector, torque, power, hub axis, tower and nacelle forces, total load. Forces are on the structure.
<root>_flow.csv	Turbine: velocity handed to OpenFAST at the hub and blade mean, after corrections.
<root>_correction.csv	disk_corrected: disk velocity, \(C_T\), induction, \(C_T'\), \(M\), recovered free stream, scale factor, loop gain, relaxation.
<root>_fllc.csv	Line with FLLC: largest and RMS correction.
<root>_disk.csv	ct_disk: upstream and disk speed, thrust, power, integrated source along the normal.
<root>_wake.csv, _wake_avg.csv	Velocity on wake lines, instantaneous and time-averaged.
<root>_stats.csv	Mean, RMS, min and max of each body quantity from avg_start.
momentum_source.csv	Force put into the flow, integrated over the domain. Equals minus the bodies’ load.
total_load.csv	Farm total load and power.
ground.csv	Terrain height under each body.

Rows of _erf, _flow and total_load carry the time at the end of their step. The others carry the time at the start of the step whose forcing they describe. OpenFAST writes its own .out files as configured in the .fst.

diagnostics_format = per_rank or farm gathers the per-body logs into one file per rank or one per farm, with a body column.

Large farms

• Use at least as many ranks as turbines. A rank steps its turbines one after another.

• Write logs every 10 to 100 steps (diagnostics_int). OpenFAST’s .out keeps the full rate.

• Sample wake lines rarely (wake.int of 50 to 200 steps), only where needed.

• Use diagnostics_format = farm (or per_rank) and stats_int of a few hundred steps.

• Measure with timing_int before tuning. A -DERF_ENABLE_TINY_PROFILE=ON build gives more detail.

Restarting

Checkpoints hold the OpenFAST state and every running quantity, so a restart continues bit for bit. It may use another number of ranks. It must keep the same bodies, inputs, anchor level and erf.fixed_dt, and it may not run past the stop time the turbines started with (OpenFAST stores it). The run stops, naming the input, otherwise.

To run past that stop time, copy the checkpoint, delete its moving_bodies folder and restart from the copy. The turbines then start afresh at the checkpoint’s time, as they do from a precursor checkpoint written without bodies.