Add NREL turbine data to atlite

[thesethtruth]

Description

Adds more turbine .yaml to the database.

Motivation and Context

Some turbines in the current database are a bit outdated. Moreover, if used in ESOM, users might want to use power curves for future / expected turbine types (with a higher rated power and different power curve). This is done based on the NREL turbine data set, which I've parsed to the atlite format using some basic python and Jinja2 templating (see: turbine parsing repo).

In our current project we use PyPSA-eur and felt like this feature was limiting. This was discussed with @fneum.

How Has This Been Tested?

Tested in an PyPSA run and through atlite on ERA5 dataset

Type of change

Extension of existing data

Feel free to propose updates / changes. Let me know what you need to use this data.

[euronion]

Hi @thesethtruth ,

thanks for the PR! You are right: We have been putting off introducing new turbines for a long time. There are a two things which need to be adressed. See the comments in one of the files (the comments apply to all files).

Let me know if you have any questions.

[thesethtruth]

Comments from @euronion processed.

If in the future you'd want to parse the data yourself; I've automated most in https://github.com/thesethtruth/NREL-to-atlite-turbines.

Cheers!

[euronion]

Nice, LGTM, thanks @thesethtruth !

Pinging @fneum since you've discussed this earlier elsewhere: Is it GTM?

[fneum]

GTM! Thanks!

[euronion]

Ah; nearly missed it: Can add a oneliner to the RELEASE_NOTES mentioning new turbines ? @thesethtruth

[thesethtruth]

Done! See below:

• Atlite now includes the reference turbines from the NREL turbine archive (see: https://nrel.github.io/turbine-models/). Available turbines can be consulted using atlite.windturbines and can be passed as string argument, e.g. coutout.wind(turbine).

[euronion]

Great, thanks!

[coroa]

Thanks for that merge, quite helpful!

As a further idea you could consider adding your existing parsing code directly to atlite as an extension to get_windturbineconfig in the same manner as we do for the oedb; so that one could use:

cutout.wind(turbine="nrel:2016CACost_NREL_Reference_8MW_180")

to transparently fetch the turbine defined at https://nrel.github.io/turbine-models/2016CACost_NREL_Reference_8MW_180.html

That would work by adding a new hook in get_windturbineconfig.

atlite/atlite/resource.py

Lines 52 to 53 in 2ab6c9e

	if isinstance(turbine, str) and turbine.startswith("oedb:"):
	return get_oedb_windturbineconfig(turbine[len("oedb:") :])

And putting your code into a new function like get_oedb_windturbineconfig.

atlite/atlite/resource.py

Lines 238 to 371 in 2ab6c9e

	def get_oedb_windturbineconfig(search=None, **search_params):
	"""
	Download a windturbine configuration from the OEDB database.
	Download the configuration of a windturbine model from the OEDB database
	into the local 'windturbine_dir'.
	The OEDB database can be viewed here:
	https://openenergy-platform.org/dataedit/view/supply/wind_turbine_library
	(2019-07-22)
	Only one turbine configuration is downloaded at a time, if the
	search parameters yield an ambigious result, no data is downloaded.
	Parameters
	----------
	search : int|str
	Smart search parameter, if int use as model id, if str look in name or turbine_type
	**search_params : dict
	Recognized arguments are 'id', 'name', 'turbine_type' and 'manufacturer'
	Returns
	-------
	turbineconfig : dict
	The turbine configuration in the format from 'atlite.ressource.get_turbineconf(name)'.
	Example
	-------
	>>> get_oedb_windturbineconfig(10)
	{'V': ..., 'POW': ..., ...}
	>>> get_oedb_windturbineconfig(name="E-53/800", manufacturer="Enercon")
	{'V': ..., 'POW': ..., ...}
	"""
	# Parse information of different allowed 'turbine' values
	if isinstance(search, int):
	search_params.setdefault("id", search)
	search = None
	# Retrieve and cache OEDB turbine data
	OEDB_URL = "https://openenergy-platform.org/api/v0/schema/supply/tables/wind_turbine_library/rows"
	# Cache turbine request locally
	global _oedb_turbines
	if _oedb_turbines is None:
	# Get the turbine list
	result = requests.get(OEDB_URL)
	# Convert JSON to dataframe for easier filtering
	# Only consider turbines with power curves available
	df = pd.DataFrame.from_dict(result.json())
	_oedb_turbines = df[df.has_power_curve]
	logger.info(
	"Searching turbine power curve in OEDB database using "
	+ ", ".join(f"{k}='{v}'" for (k, v) in search_params.items())
	+ "."
	)
	# Working copy
	df = _oedb_turbines
	selector = True
	if search is not None:
	selector &= df.name.str.contains(
	search, case=False
	) | df.turbine_type.str.contains(search, case=False)
	if "id" in search_params:
	selector &= df.id == int(search_params["id"])
	if "name" in search_params:
	selector &= df.name.str.contains(search_params["name"], case=False)
	if "turbine_type" in search_params:
	selector &= df.turbine_type.str.contains(search_params["name"], case=False)
	if "manufacturer" in search_params:
	selector &= df.manufacturer.str.contains(
	search_params["manufacturer"], case=False
	)
	df = df.loc[selector]
	if len(df) < 1:
	raise RuntimeError("No turbine found.")
	elif len(df) > 1:
	raise RuntimeError(
	f"Provided information corresponds to {len(df)} turbines,"
	" use `id` for an unambiguous search.\n"
	+ str(df[["id", "manufacturer", "turbine_type"]])
	)
	# Convert to series for simpliticty
	ds = df.iloc[0]
	# convert power from kW to MW
	power = np.array(json.loads(ds.power_curve_values)) / 1e3
	hub_height = ds.hub_height
	if not hub_height:
	hub_height = 100
	logger.warning(
	"No hub_height defined in dataset. Manual clean-up required."
	"Assuming a hub_height of 100m for now."
	)
	elif isinstance(hub_height, str):
	hub_heights = [float(t) for t in re.split(r"\s*;\s*", hub_height.strip()) if t]
	if len(hub_heights) > 1:
	hub_height = np.mean(hub_heights, dtype=int)
	logger.warning(
	"Multiple values for hub_height in dataset (%s). "
	"Manual clean-up required. Using the averge %dm for now.",
	hub_heights,
	hub_height,
	)
	else:
	hub_height = hub_heights[0]
	turbineconf = {
	"name": ds.turbine_type.strip(),
	"manufacturer": ds.manufacturer.strip(),
	"source": f"Original: {ds.source}. Via OEDB {OEDB_URL}",
	"hub_height": hub_height,
	"V": np.array(json.loads(ds.power_curve_wind_speeds)),
	"POW": power,
	"P": power.max(),
	}
	# Cache in windturbines
	global windturbines
	charmap = str.maketrans("/- ", "___")
	name = "{manufacturer}_{name}".format(**turbineconf).translate(charmap)
	windturbines[name] = turbineconf
	return turbineconf

Maybe that is good first issue. But, many thanks either way!

[thesethtruth]

@coroa Thanks for your comment. I had the email notification parked for a long time now (hence the late response) hoping I would find the time to implement your suggestion. Unfortunately, I won't be able to since I'm very tight for time at the moment. If I have time at a later stage and this is still relevant I will implement it!