Improving the Efficiency of Fan Coil Units in Hotel Buildings through Deep-Learning-Based Fault Detection

Area of Science:

• Building Energy Systems
• Artificial Intelligence in Engineering
• HVAC Performance Optimization

Background:

• Fan coil units (FCUs) are crucial for occupant comfort within HVAC systems but are prone to failures impacting energy efficiency.
• Energy consumption and system failures in FCUs necessitate advanced fault detection methods.
• Deep learning (DL) presents a promising approach for early fault identification and prevention in FCUs.

Purpose of the Study:

• To investigate the efficacy of deep learning models for fault detection in fan coil units (FCUs).
• To enhance the energy efficiency of hotel buildings through improved FCU performance monitoring.
• To compare the performance of different DL architectures against a traditional machine learning model.

Main Methods:

• Tested three deep learning models: Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and a hybrid CNN-Gated Recurrent Unit (GRU).
• Utilized a real-world dataset from a hotel's sensory measurement system, augmented with simulated data from a TRNSYS physical model.
• Simulated three common FCU faults: stuck valve, reduced airflow, and FCU outage, using an extensive dataset.

Main Results:

• The hybrid CNN-GRU model achieved the best performance across all three simulated FCU faults.
• Deep learning-based fault detectors significantly outperformed the baseline Random Forest (RF) model.
• The study confirmed the viability of DL models for enhancing FCU reliability and energy efficiency.

Conclusions:

• Deep learning models, particularly the hybrid CNN-GRU, are highly effective for detecting diverse FCU faults.
• Implementing DL-based fault detection can lead to substantial energy savings and improved operational efficiency in smart buildings.
• These advanced fault detection techniques offer a pathway towards more sustainable and energy-efficient hotel operations.