Forecasting energy prices and volumes in the Italian Ancillary Service Market: A Machine Learning approach

This thesis, which is the result of 6-month work started from a collaboration between ENI Energy Company (Italy) and the University of Pavia, it's based on volumes and prices forecasting in the Ancillary Market (MSD) that is handled by Terna. In particular, the goal is to estimate the demand of energy to allow the company to place the best offer based on the volumes required by the market. All datasets used for model estimation are provided by ENI. In fact, due to the amount of data available, this thesis starts with a the analysis of all possible links between the target variables and the covariates. We continue with the training of a classifier to predict the volumes using different regularized Supervised Learning methods. After that, we focus on price forecasting, trying to predict a value such that the bid is accepted by the market. All models and classifiers are estimated considering only the North Zone of Italy with master data that range from 2018 to the first months on 2021. At the end, the final results are a volume classifier able to guess with a good error rate in which band the forecast falls and a fairly accurate model of upward price time series.