View on GitHub

forecasting-in-the-dark

A Data Science LAB project (lite version, view the full version at https://gitlab.com/DBertazioli/enercibiddding). Functional forecasting for energy supply functions in italian energy market comparing a statistical model (Reduced Rank Regression) with a machine learning one (Long short-term memory neural nets)

a Data Science LAB Project

Overview   |   References   |   Data   |   Presentation   |   About us  

☍   Overview

GME operates in power, gas and environmental markets. It is the exchange place for electricity and natural gas spot trading in Italy. In the power market platform, producers and purchasers sell and buy wholesale electricity. There is an auction for every hour of the day.

Forecasting this supply function could be interesting for every energy producer.



example of supply surface obtained plotting more supply functions all together


Our objective is to forecast each time series in order to obtain an estimated supply function for future time (1-hour, 24-hours, 168-hours).



example of time series forecasting


To do this Reduced Rank Regression (RRR) and Long short-term memory neural nets (LSTM) are used, in order to compare a pure statistical method with machine learning technique.

Dataset exploration, data preprocessing and LSTM models are executed using Python and all the code is available in the PyScripts folder while RRR models are made using R and are available in the Rscripts folder.

☍   References

All the references for this project are available in the ref folder.

☍   Data

All the data used for this project are available in the data folder (on gitlab) and here

☍   Presentation

Our slides presentation is available in the slides folder.

☍   About us

⊜   Alessandro Borroni

⊜   Dario Bertazioli

⊜   Fabrizio D’Intinosante

⊜   Massimiliano Perletti