TimeSeries

Time Series

This is a Python toolset for generating (possibly many) synthetic time series of various type. The main tools are contained in the module time_series.py.

Existing models

The models directory contains several models for loads based on the public ENTSO-E data. There are models for most European countries. Some of them have 15-minutes time steps (e.g. Germany), while others have 1-hour time steps (e.g. France).

To generate random time series based on these models, just use the following code:

import time_series
model = time_series.import_model('models/entsoe_load_CH_2015_2023.npz')
time_series.generate_time_series(model, 20)

This code outputs 20 series of 364 x 24 time steps in a NumPy array format. It is based on the Swiss model constructed from historic data of the years 2015 to 2023.

To generate time series with a broader distribution simulating more extreme conditions, the standard deviation can be scaled up, for instance here by a factor 2:

time_series.generate_time_series(model, 20, std_scaling=2.0)

Normalized time series with a mean value of one can be obtained with:

time_series.generate_time_series(model, 20, normalize=True)

Creating models

In addition to existing models, it is possible to create new models based on any type of existing time series.

Assuming that you have gathered 3 historic time series for a particular power system and stored them in lists named time_series_1, time_series_2, and time_series_3, then creating a model from these lists is as simple as

model = time_series.create_model([time_series_1, time_series_2, time_series_3])

This model can be used to generate new time series resembling the historic ones.

The procedure for creating new models is detailed in a Jupyter Notebook. Synthetic time series for power generation or cross-border flows can for instance be easily generated based on ENTSO-E data with the help of the entso_e.py module.

Installation

This project only relies on the standard Python modules NumPy and SciPy, as well as PyTest for testing and Pandas for dealing with datasets in the examples.

All dependencies can be installed with the command:

pip install -r requirements.txt

The installation can be verified by running tests:

pytest

Future developments

• More models could be added to the existing list.
• The SciPy support for linear algebra with sparse matrices is limited. A more efficient implementation could be written in Julia.