Researchers from Pablo de Olavide University conducted a comparative study of different forecasting strategies for forecasting the energy consumption of smart buildings
Increasing population has led to high global demand and consumption of energy. According to ExxonMobil 2018 Outlook for Energy: A View to 2040, the global energy demand is expected to increase 25% by 2040 during 2018-2040. Energy consumption has a significant impact on the environment. The current energy plan developed by the European Commission requires European Union members to adopt a set of measures to reach an energy efficiency of at least 20%. Several building sub-sectors such as space heating and cooling, water heating, and lighting can offer potential energy savings.
Now, a team of researchers from Pablo de Olavide University conducted a comparative empirical evaluation of different time series forecasting strategies. The team took into account both statistical and Machine Learning (ML)-based approaches. These strategies were applied to a dataset regarding the electricity consumption registered by thirteen buildings located at the university campus for validation. The data set was collected over five and a half years. The team focused on predicting the electric energy consumption with a one day horizon. The team found that ML-based approaches were the best performing methods. With the lowest prediction errors, The Random Forests, Generalized Boosted Regression Mode, and Extreme Gradient Boosting attained the best performances on the data used.
In terms of optimal historical window value, the team found that the results obtained improved when the size of the window used was higher than seven i.e., when seven or more days were considered to predict the electric energy consumption of the next day, with the best results obtained when the window size was set to 10. In further research, the team is expected to focus on applying a Deep Learning approach as it has achieved good results in other time series forecasting problems. The team also plans to extend the research to other kinds of time series such as relating to water and/or gas consumption in smart buildings. The research was published in the journal MDPI Energies on May 20, 2019.