Meteo for Energy - PV - GET Pred time series

API service to retrieve hourly energy forecast data

The energy forecast (in kWh) is required for Solar PV power plants for the energy management system and/or to inform the TSOs. These data is stored on Meteo for Energy databases and also available for the user in case third party services require this information.

This example will show:

• How to make a request of the energy forecast of a Solar PV power plant.
• Extract the information in a DataFrame format.
• Plot the results

# First of all, import required packages
%matplotlib inline

import datetime
import requests
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

url_base = 'https://api.meteoforenergy.com/Meteo_WIND/v1'

/usr/local/lib/python3.6/dist-packages/pandas/compat/_optional.py:124: UserWarning: Pandas requires version '1.2.1' or newer of 'bottleneck' (version '1.2.0' currently installed).
  warnings.warn(msg, UserWarning)

Now, we can start preparing our request of energy generation for a site. Note that we will request the time series (ts) of the forecast (pred).

url = url_base + '/pred/ts/'

# Time period that we will request
datePred = datetime.datetime(2021,10,13,0,0,0)

# Define required GET parameters for the request
headers = {}
headers['X-Auth-Token-Site'] = 'Your API token'  # Your site eToken

params = {}
params['market'] = 'spot'                              # spot / intraday / xbid
params['datetime_pred'] = datePred                     # datetime_pred requires ISO8601 format
params['proba'] = True                                 # False or True if 10/25/75/90 percentile is required
params['midnight'] = True                              # Forecast time series starts from local midnight

# Request to the API the data defined on the parameters
r = requests.get(url, params=params, headers=headers)
print(r.text[:200])
if( r.status_code == 200 ):
    df = pd.DataFrame(r.json()['data'])                             # or r.json().get('data')
    print(df.head())
    print(df.info())

{"name":"Atalaya","lat":41.8028,"lon":-1.29533,"market":"spot","datetime_pred":"2021-10-13T02:00:00+02:00","probability":true,"unit":"kWh","files":3,"data":[{"time_unix":1634076000,"datetime_iso8601":
    time_unix          datetime_iso8601    datetime_local  pred_mean  \
0  1634076000  2021-10-12T22:00:00.000Z  2021-10-13 00:00        427   
1  1634079600  2021-10-12T23:00:00.000Z  2021-10-13 01:00        385   
2  1634083200  2021-10-13T00:00:00.000Z  2021-10-13 02:00        333   
3  1634086800  2021-10-13T01:00:00.000Z  2021-10-13 03:00        310   
4  1634090400  2021-10-13T02:00:00.000Z  2021-10-13 04:00        271   

   pred_p10  pred_p25  pred_p75  pred_p90  
0        83       282       572       651  
1        74       264       528       634  
2        73       237       453       581  
3        65       190       449       575  
4        65       190       435       531  
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 147 entries, 0 to 146
Data columns (total 8 columns):
 #   Column            Non-Null Count  Dtype 
---  ------            --------------  ----- 
 0   time_unix         147 non-null    int64 
 1   datetime_iso8601  147 non-null    object
 2   datetime_local    147 non-null    object
 3   pred_mean         147 non-null    int64 
 4   pred_p10          147 non-null    int64 
 5   pred_p25          147 non-null    int64 
 6   pred_p75          147 non-null    int64 
 7   pred_p90          147 non-null    int64 
dtypes: int64(6), object(2)
memory usage: 9.3+ KB
None

The response is a JSON that is converted to a DataFrame, but all the types are not correctly parsed. So we need to define correctly the datetime column and define it on our desired time zone (List of tz database time zones), as it is localized in UTC time zone.

# Convert type from object to datetime, and then localize o
df['datetime_iso8601'] = pd.to_datetime(df['datetime_iso8601'])              # datetime: object -> datetime64[ns, UTC]
df['datetime_local'] = df['datetime_iso8601'].dt.tz_convert('Europe/Madrid') # datetime: datetime64[ns, UTC] -> datetime64[ns, Europe/Madrid]
print(df.head())
print(df.dtypes)

    time_unix          datetime_iso8601            datetime_local  pred_mean  \
0  1634076000 2021-10-12 22:00:00+00:00 2021-10-13 00:00:00+02:00        427   
1  1634079600 2021-10-12 23:00:00+00:00 2021-10-13 01:00:00+02:00        385   
2  1634083200 2021-10-13 00:00:00+00:00 2021-10-13 02:00:00+02:00        333   
3  1634086800 2021-10-13 01:00:00+00:00 2021-10-13 03:00:00+02:00        310   
4  1634090400 2021-10-13 02:00:00+00:00 2021-10-13 04:00:00+02:00        271   

   pred_p10  pred_p25  pred_p75  pred_p90  
0        83       282       572       651  
1        74       264       528       634  
2        73       237       453       581  
3        65       190       449       575  
4        65       190       435       531  
time_unix                                   int64
datetime_iso8601              datetime64[ns, UTC]
datetime_local      datetime64[ns, Europe/Madrid]
pred_mean                                   int64
pred_p10                                    int64
pred_p25                                    int64
pred_p75                                    int64
pred_p90                                    int64
dtype: object

Similarly, we can also use Unix Epoch Time to make the same transformation

# datetime: Unix epoch time -> datetime64[ns, Europe/Madrid]
df['datetime_local'] = pd.to_datetime(df['time_unix'], unit='s').dt.tz_localize('UTC').dt.tz_convert('Europe/Madrid')
print(df.head())
print(df.dtypes)

    time_unix          datetime_iso8601            datetime_local  pred_mean  \
0  1634076000 2021-10-12 22:00:00+00:00 2021-10-13 00:00:00+02:00        427   
1  1634079600 2021-10-12 23:00:00+00:00 2021-10-13 01:00:00+02:00        385   
2  1634083200 2021-10-13 00:00:00+00:00 2021-10-13 02:00:00+02:00        333   
3  1634086800 2021-10-13 01:00:00+00:00 2021-10-13 03:00:00+02:00        310   
4  1634090400 2021-10-13 02:00:00+00:00 2021-10-13 04:00:00+02:00        271   

   pred_p10  pred_p25  pred_p75  pred_p90  
0        83       282       572       651  
1        74       264       528       634  
2        73       237       453       581  
3        65       190       449       575  
4        65       190       435       531  
time_unix                                   int64
datetime_iso8601              datetime64[ns, UTC]
datetime_local      datetime64[ns, Europe/Madrid]
pred_mean                                   int64
pred_p10                                    int64
pred_p25                                    int64
pred_p75                                    int64
pred_p90                                    int64
dtype: object

Plot time

It's time to plot the information requested

# Set `datetime` as index to show on X axis
if(df.index.name != 'datetime_local'):
    df = df.set_index('datetime_local')

OX = df.index + pd.Timedelta(minutes=30)     # As values are hourly averages, we centre at XX:30

fig = plt.figure(figsize=(14,6), dpi=80)
plt.plot(OX, df['pred_mean'], c='#95c11f', marker='o',lw=2, label=u'Forecast')
plt.fill_between(OX, df['pred_p25'], df['pred_p75'],alpha=0.5, fc='#95c11f', ec='None', label='25%-75%')
plt.fill_between(OX, df['pred_p10'], df['pred_p90'],alpha=0.2, fc='#95c11f', ec='None', label='10%-90%')
plt.title('Meteo for Energy forecast. %s' % (datePred.strftime('%Y-%m-%d')), fontsize=24, fontweight='bold')
plt.xlabel('datetime', fontsize=16, fontweight='bold')
plt.ylabel('Energy [$kWh$]', fontsize=16, fontweight='bold')
plt.legend(loc='upper right')
plt.grid()
plt.show()