Setting up a site¶

Create a site¶

Following are a few examples of how to use rebase.Site.create() to create sites of different kinds.

Common for all sites are that they need a position (latitude, longitude) as well as a name and type. Full site configuration requirements can be viewed on the Configurations page.

Solar¶

Create a site that will be used to forecast PV power production.

>>> import rebase as rb

>>> site_config = {
>>>   'type': 'solar',
>>>   'name': 'My solar site',
>>>   'latitude': 53.41,
>>>   'longitude': 5.94,
>>>   'azimuth': 171.3, # 0 = North, 90 = East, 180 = South, 270 = West
>>>   'tilt': 10.3,
>>>   'capacity': [
>>>     {'value': 750.5, 'validFrom': '2019-10-10T00:00:00Z'}, # capacity of site changed to 750.5 kW at this date
>>>     {'value': 500.3, 'validFrom': '2019-04-03T00:00:00Z'}, # site was installed at this date
>>>   ],
>>> }

>>> site_id = rb.Site.create(site_config)

Wind¶

Create a site that will be used to forecast wind power production.

>>> import rebase as rb

>>> site_config = {
>>>   'type': 'wind',
>>>   'name': 'My wind site',
>>>   'latitude': 53.41,
>>>   'longitude': 5.94,
>>>   'capacity': [
>>>     {'value': 4000, 'validFrom': '2019-10-10T00:00:00Z'}, # capacity of site changed to 4000 kW at this date
>>>     {'value': 2000, 'validFrom': '2019-04-03T00:00:00Z'}, # site was installed at this date
>>>   ],
>>> }

>>> site_id = rb.Site.create(site_config)

Localized weather¶

Create a site that will be used to forecast localized weather. Localized weather is an AI point weather forecast at the site’s location, trained on historical data of a certain weather variable. For example, to get a localized wind speed you will upload historically observed wind speed data.

>>> import rebase as rb

>>> site_config = {
>>>   'type': 'localized',
>>>   'name': 'My localized site',
>>>   'latitude': 53.41,
>>>   'longitude': 5.94,
>>>   'measurement': 'WindSpeed'
>>> }

>>> site_id = rb.Site.create(site_config)

Electricity demand¶

Create a site that will be used to forecast electricity demand.

>>> import rebase as rb

>>> site_config = {
>>>    'type': 'load',
>>>    'name': 'My load site',
>>>    'latitude': 53.41,
>>>    'longitude': 5.94,
>>>    'nwps': ['DWD_ICON-EU', 'NCEP_GFS'],
>>>    'variables': [
>>>        {'name': 'Temperature', 'lag': [-4, -3, -2, -1, 1, 2, 3, 4]},
>>>        {'name': 'SolarDownwardRadiation'},
>>>    ],
>>>    'calendar': ['holidays', 'hourOfDay'],
>>> }

>>> site_id = rb.Site.create(site_config)

Get your site(s)¶

List all of your sites using rebase.Site.list()

>>> import rebase as rb

>>> sites = rb.Site.list()

Get a single site by its id using rebase.Site.get()

>>> import rebase as rb

>>> site_id = '4ab82692-3944-4069-9cbb-f9c59513c1c3'
>>> sites = rb.Site.get(site_id)

Upload data to train on¶

First create a pandas DataFrame with your site’s data. The DataFrame needs to contain these exact two columns (all other will be disregarded):

valid_time (datetime, or parseable date strings) - a column containing the data timestamps
observation (numeric) - a column containing the data values

See example DataFrame printed below:

>>> import rebase as rb
>>> import pandas as pd

>>> df = pd.read_csv('example_data.csv')
>>> print(df)
         valid_time                observation
0        2020-01-22 00:00:00+00:00       126.3
1        2020-01-22 00:15:00+00:00       122.7
.        ...                             ...
n-1      2020-10-17 23:30:00+00:00       169.2
n        2020-10-17 23:45:00+00:00       176.6

Second, simply upload the data for your site as follows:

>>> site_id = '4ab82692-3944-4069-9cbb-f9c59513c1c3' # replace with your site's id
>>> rb.Site.upload(site_id, df)

Train a new model for your site¶

After uploading observed data for your site, start the training using rebase.Site.train()

>>> import rebase as rb

>>> site_id = '4ab82692-3944-4069-9cbb-f9c59513c1c3' # replace with your site's id
>>> rb.Site.train(site_id)

Check your site’s training status. See rebase.Site.status() for more information.

>>> rb.Site.status(site_id)
{
    'status': 'complete',
    'history': [
        {'state': 'queued', 'timestamp_utc': '2020-10-12 13:04:17'},
        {'state': 'training', 'timestamp_utc': '2020-10-12 13:04:22'},
        {'state': 'complete', 'timestamp_utc': '2020-10-12 13:05:23'},
    ]
}

Get a site forecast¶

Get your site’s latest forecast using rebase.Site.forecast()

>>> import rebase as rb

>>> site_id = '4ab82692-3944-4069-9cbb-f9c59513c1c3' # replace with your site's id
>>> data = rb.Site.forecast(site_id)

User guide Creating a custom model