Health Status of an HVAC System of a Commercial Building Using Neural Networks

Commercial buildings and large constructions are responsible for a significant part of global energy consumption and greenhouse gas emissions, with heating, ventilation, and air conditioning systems being one of the main culprits. Maintaining the proper functioning of these systems is crucial due to their impact on comfort, air quality, environment, and operating costs. To optimize their performance, a machine learning algorithm is proposed that estimates the coefficient of performance and evaluates the health status of the heating, ventilation, and air conditioning systems. This methodology allows the detection of problems and assesses the system's efficiency by classifying the health status considering the coefficient of performance, date, and energy consumption. This paper presents the application of two Long Short-Term Memory neural networks, using data collected during one year by the Lawrence Berkeley National Laboratory. For the regressor model, twenty-one input features were used, while the classifier model used six features to improve system efficiency, optimize comfort, reduce environmental impact, and minimize maintenance costs. Based on predictive maintenance, allows us to determine the right time to carry out interventions. The Long Short-Term Memory regressor neural network estimated the performance coefficient with 99% accuracy, while the Long Short-Term Memory classifier neural network achieved 94% accuracy in determining the health status of the HVAC system.