Application of Machine Learning Techniques to Automatic Energy Incident Detection

Abstract

Heating and cooling account for approximately half of the EU’s energy demand. Nowadays, new heating, ventilation, and air-conditioning machines have sensors that generate huge amount of data. The analysis and processing of data from hundreds of devices and field sensors using Machine Learning (ML) models allows to address maintenance management. This enables the anticipation of failures and the scheduling of predictive maintenance. These lead to the optimization of processes, the minimization of downtime and costs, and more efficient use of energy. With the use of an ML model capable of predicting and organizing a wider range of incidents, an energy saving of 7% per installation is estimated, with a reduction of 30% in preventive maintenance visits and a reduction of 20% in corrective maintenance. More than 20% of the incidents registered could be corrected remotely without the need for a technician. Predictive maintenance will prolong the life of customers’ HVAC machines and reduce the number that needs to be replaced.