Wake steering design

Wake steering design#

The whoc.design_tools.wake_steering_design module provides various tools for the design of yaw offset lookup tables for “open-loop” wake steering. The two primary functions are build_simple_wake_steering_lookup_table and build_uncertain_wake_steering_lookup_table, both of which take an instantiated FlorisModel, along with various design parameters, and return a pandas DataFrame df_opt containing the optimal yaw offset angles for each wind turbine. Under the hood, both functions run an optimization using FLORIS’ YawOptimizerSR methodology. The uncertain version takes into account wind direction uncertainty via the second required argument wd_std, representing the wind direction standard deviation.

The output DataFrame of optimal yaw angles df_opt can then be passed to the LookupBasedWakeSteeringController upon its instantiation.

As well as these primary functions, several other functions are provided that modify the FLORIS-produced optimal offsets for various practical purposes, as described below.

Geometric ramping with wind speed#

Generally speaking, while the dependency of optimal offsets on wind direction is complex, the dependency on wind speed is simpler. The apply_wind_speed_ramps() function allows users to define a simple heuristic relationship of the optimal offsets on wind speeds, given a set of yaw angles optimized for a single “middle” wind speed. Offsets are linearly “ramped up” between ws_wake_steering_cut_in and ws_wake_steering_fully_engaged_low; take the optimized value between ws_wake_steering_fully_engaged_low and ws_wake_steering_fully_engaged_high; and are linearly ramped down again between ws_wake_steering_fully_engaged_high and ws_wake_steering_cut_out. Additionally, the DataFrame produced pads with zero offsets between ws_min and ws_wake_steering_cut_in and between ws_wake_steering_cut_out and ws_max.

Yaw offset rate limits#

To limit the rate of change of the yaw offset as a function of wind direction, wind speed, or turbulence intensity, use the apply_static_rate_limits() function. This will “smooth” the yaw offsets to avoid high sensitivity to changes in the lookup table inputs (most obviously, to wind direction near the perfectly aligned directions, where optimal offsets often change from being at largest negative value to being at their largest positive value over a small wind direction window). Resulting offsets, however, will be suboptimal and produce significant performance reduction, and we generally recommend using time-based limitation instead, such as hysteresis zones or low-pass filtering the input wind directions.

Two other wake steering-based utilities are provided in the whoc.design_tools.wake_steering_design module.

Finding hysteresis zones#

If excessive yaw maneuvers are of concern, an alternative to rate limiting the optimal offsets is to apply a hysteresis approach around the transition wind directions. This holds the yaw offset at a fixed value across the transition between negative and positive values, until a certain “large-enough” change in wind direction occurs. This prevents chattering between large positive and negative values. The compute_hysteresis_zones() function takes in a DataFrame of optimal yaw angles, and returns a list of wind direction bands where hysteresis should be applied based on transitions exceeding the yaw_rate_threshold. The width of the hysteresis region is defined by min_zone_width, unless zones overlap, in which case they are combined. The list of hysteresis zones can be provided to the LookupBasedWakeSteeringController on instantiation, along with df_opt, to apply dynamic hysteresis to the yaw offsets. However, again, care should be taken with this approach as it can result in “wrong-way steering”.

Converting yaw offsets DataFrame to an interpolant function#

In many cases, including in the LookupBasedWakeSteeringController, it is more convenient for the precomputed yaw offsets lookup table to be defined as a an interpolant than a DataFrame of optimal values. The get_yaw_angles_interpolant function takes in a DataFrame df_opt and returns a function that can be queried at any wind direction, wind speed, turbulence intensity combination to provide an interpolated set of offsets. Additionally, the wind direction, wind speed, and turbulence intensity can be queried using arrays of equal length to interpolate in a vectorized manner.

Note that in LookupBasedWakeSteeringController, the construction of the interpolator happens automatically based on the df_opt passed in on instantiation.

Wake steering offset visualization#

Visualization tools for wake steering lookup tables are provided in the whoc.design_tools.wake_steering_visualization module. There are currently two functions:

plot_offsets_wswd_heatmap() creates a heatmap of offsets by wind speed and wind direction based on df_opt for a given turbine index turb_id. Unless only a single TI value appears in df_opt, users will need to specify the turbulence intensity for the plot using ti_plot.
plot_offsets_wd() plots the offsets from df_opt for turbine turb_id at a specified (set of) wind speeds wd_plot. Unless only a single TI value appears in df_opt, users will need to specify the turbulence intensity for the plot using ti_plot.

Both functions, as well as many of the design functions described here, are demonstrated in the compare_yaw_offset_designs.py python script provided in lookup-based_wake_steering_florisstandin.