Developing of machine learning-based energy monitoring methodologies for the building energy demand of healthcare facilities: an Italian case study

The aim of this work is to create a systematic methodology to proceed from the raw data to a ready-to-implement monitoring method, minimizing the knowledge required to the user and, contemporary, ensure reliable performance of the monitoring method itself, starting from methods with low computational effort that require high user experience, to method that requires significantly higher computational resources, but aiming to minimize the required background knowledge to be applied. The proposed monitoring methods are based on machine learning models able to predict the energy demand. This work explores several machine learning algorithms (Multiple Linear Regression and Artificial Neural Network), evaluating different methods to perform the Feature Selection and Hyperparameters Tuning. The models prediction are exploited to perform statistical residual analyses through the well-known technique of cumulative sum of differences.