Time Series Machine Learning Technique with Application to Barcelona Metro Station Energy Minimization

Abstract

This thesis is part of the SEAM4US project, which goal is to minimize the energy consumption of the Barcelona metro station. The energy minimization is done by so-called model predictive control, i.e. management of the energy using systems based on a time series prediction. Here we focus on such a prediction. We make use of the popular Fourier transformation in combination with trend detection and validate our method by testing on different data sets and comparing with well-known techniques. Furthermore we conclude that this technique is very usable for the SEAM4US project and probably for a lot of time series prediction.