Davide Borda
Davide Borda has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Davide Papurello
Department of Energy, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Turin, Italy
1 shared publication - Marco Carlo Masoero
Politecnico di Torino, Corso Duca degli Abruzzi 24, 10123 Torino, Italy
1 shared publication - Massimo Amerio
EURIX, Corso Vittorio Emanuele II, 61, 10128 Turin, Italy
1 shared publication - Mattia Bergagio
EURIX, Corso Vittorio Emanuele II, 61, 10128 Turin, Italy
1 shared publication - Romano Borchiellini
Department of Energy, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy
1 shared publication
Latest Publications
- Open Access
Development of Anomaly Detectors for HVAC Systems Using Machine Learning
Faults and anomalous behavior affect the operation of Heating, Ventilation and Air Conditioning (HVAC) systems. This causes performance loss, energy waste, noncompliance with regulations and discomfort among occupants. To prevent damage, automated, fast identification of faults in HVAC systems is needed. Fault Detection and Diagnosis (FDD) techniques are very effective for these purposes. The best FDD methods, in terms of cost effectiveness and data exploitation, are based on process history; i.e., on sensor data from automation systems. In this work, supervised and semi-supervised models were developed. Other than with regard to outdoor temperature and humidity, the input parameters of an HVAC system have few internal variables. Performance of traditional methods (e.g., VAR, Random Forest) is low, so Artificial Neural Networks (ANNs) were selected, since they can capture nonlinear relationships among features and are easily optimized. ANNs can detect simultaneous faults from different classes. ANN metrics are easily evaluated. The ground truth is obtained from process history (supervised case) and from a mix of deterministic methods and clustering (semi-supervised case). The derivation of the ground truth in the semi-supervised case, and extensive comparison with advanced supervised models, set this work apart from previous studies. The Mean Absolute Error (MAE) of the best supervised model was 0.032 over 15 min and 0.034 over 30 min. The Balanced Accuracy Score (BAS) of the best semi-supervised model was 86%.
Open AccessDevelopment of Anomaly Detectors for HVAC Systems Using Machine Learning
Faults and anomalous behavior affect the operation of Heating, Ventilation and Air Conditioning (HVAC) systems. This causes performance loss, energy waste, noncompliance with regulations and discomfort among occupants. To prevent damage, automated, fast identification of faults in HVAC systems is needed. Fault Detection and Diagnosis (FDD) techniques are very effective for these purposes. The best FDD methods, in terms of cost effectiveness and data exploitation, are based on process history; i.e., on sensor data from automation systems. In this work, supervised and semi-supervised models were developed. Other than with regard to outdoor temperature and humidity, the input parameters of an HVAC system have few internal variables. Performance of traditional methods (e.g., VAR, Random Forest) is low, so Artificial Neural Networks (ANNs) were selected, since they can capture nonlinear relationships among features and are easily optimized. ANNs can detect simultaneous faults from different classes. ANN metrics are easily evaluated. The ground truth is obtained from process history (supervised case) and from a mix of deterministic methods and clustering (semi-supervised case). The derivation of the ground truth in the semi-supervised case, and extensive comparison with advanced supervised models, set this work apart from previous studies. The Mean Absolute Error (MAE) of the best supervised model was 0.032 over 15 min and 0.034 over 30 min. The Balanced Accuracy Score (BAS) of the best semi-supervised model was 86%.