ML2Building - Data-driven machine learning approach for building energy demand forecasting
Title: ML2Building - Data-driven machine learning approach for building energy demand forecasting
Principal investigator: João Miguel Charrua de Sousa
Research team: João Miguel Charrua de Sousa, Hermano Joaquim dos Santos Bernardo, Filipe Tadeu Soares Oliveira, Marcela Ribeiro Ferreira (Master Student in Energy and Environmental Engineering at ESTG/IPLeiria), Research Fellow (to be defined)
Dates start/end: 1st of November 2019/31 October 2020
Synopsis
This project aims at using a data-driven machine learning approach for forecasting the energy demand of buildings, particularly in the non-residential sector. As the prediction of energy consumption in buildings is usually made by detailed building energy performance simulation (using software tools such as EnergyPlus), it requires the training of the user, the feeding of the building data input to the software and the computational effort associated to calculations, that can be considerably time-consuming. Whenever the building starts to be operated and occupied, the factors affecting the energy consumption increase significantly adding uncertainty and difficulty to the process of creating accurate simulation models. Bearing in mind that in these circumstances historically recorded time series energy data becomes available, statistical and machine learning techniques appear to be an alternative tool for forecasting future energy demand scenarios. Therefore, the main goal of this project is to validate the proposed approach through the comparisons with previously calibrated detailed building energy simulation models and real energy measured data.