Machine learning-based monitoring method for the electricity consumption of a healthcare facility in Italy

The development of building energy management strategies leads to important energy savings, especially for energy-intensive buildings. It implies carrying out detailed analyses of the building energy needs of the specific test case under analysis. This work analyses the electricity consumption of a healthcare facility located near Florence, Italy, studying the correlation of the electricity demand with climates, time and healthcare activities parameters to find the main building energy drivers. The study exploits machine learning methods to predict the electricity demand of the healthcare facility, comparing the performance of Multiple Linear Regression and Artificial Neural Networks. Feature selection and Feature Engineering procedures have been carried out to obtain the representation of input data that maximises prediction performance. Then, the model has been exploited to develop an offline monitoring method for the electricity consumption of the facility, providing a suitable tool to highlight changes in the building electricity demand behaviour. The work highlights the importance of energy forecasting model optimization, aiming to realize accurate monitoring methods for the building electricity consumption and therefore increase the effectiveness and responsiveness in recognizing any anomalies. The proposed method represents a reference methodology for machine learning-aided building energy monitoring applicable to several different contexts and applications.