Eddy viscosity wake model

[misi9170]

Eddy viscosity wake model

Introduces the eddy viscosity wake model, as described by Ainslie (1988) and extended by Gunn (2019).

The implementation comprises the following:

• Calculation of an initial velocity in the wake of a wind turbine
• Computation (via integrating an ODE) of the centerline velocities in the wind turbine wake for downstream locations
• Calculation of the wake width according to the usual Gaussian wake shape and momentum conservation
• Correction for the wake width via "streamtube expansion" when the wake of an upstream turbine encounters a downstream turbine
• Combination of wakes using the "sum of energy deficits" (Kuo et al., 2014)

These are implemented in the following locations:

• floris/core/wake_velocity/eddy_viscosity.py
• floris/core/wake_combination/soed.py
• floris/core/solver.py (streamtube_expansion_solver() and full_flow_streamtube_expansion_solver())

This pull request is currently a draft, while we work through final implementation steps and incorporate feedback.

TODO

• Implement eddy viscosity model main solve
• Test main solve
• Implement eddy viscosity "full flow" solve for visualization and flow sampling
• Include wake deflection model (currently raises NotImplementedError with nonzero yaw angles)
• Add initial set of examples
• Handle warnings raised
• Incorporate feedback and comments on implementation, in particular:
• where streamtube expansion should be applied
• which turbulence intensity (either atmospheric or wake-induced) should be used to initialize eddy viscosity model for downstream turbines
• whether and how to use the near-wake correction filter
• Clean up and finalize examples
• Add documentation

[misi9170]

This pull request is now ready for comments. I've added two example scripts that use the eddy viscosity model, which can be found in examples/examples_eddy_viscosity.

The first, 001_demonstrate_eddy_viscosity_model.py, demonstrates the wake velocity predictions in a wind farm (first, a row of 3 turbines aligned; and second, a 3x3 wind farm in conditions first where the flow is aligned and second when the flow is misaligned with the rows by 15 degrees). Running the script produces the following figures:

The second script, 002_reproduce_published_results.py, attempts to recreate plots published by Ainslie (1988) and Gunn (2019) of the wake profile. I was not able to accurately reproduce the results from Ainslie (Figs 3. and 4.), but I believe the second figure generated, which attempts to recreate Fig 1b. from Gunn, is a good match. I used the specified parameters from Gunn, and adjusted the ambient turbulence intensity to get this match.