Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimizing Indoor Environmental Quality in Smart Buildings Using Deep Learning

Youssef Sabiri, Walid Houmaidi, Aaya Bougrine, Salmane El Mansour Billah

TL;DR

This work tackles the problem of energy-efficient, occupant-friendly IEQ management by forecasting CO$_2$, temperature, and humidity using deep learning. It benchmarks three architectures—LSTM, GRU, and a CNN-LSTM hybrid—on the ROBOD dataset to predict IEQ over short-term horizons ($5$ minutes ahead) and assesses performance across multiple horizons with edge/cloud deployment options. Key contributions include a comprehensive architecture comparison, detailed preprocessing and feature engineering (cyclical time features, continuous intervals, and a fixed window), and concrete guidance for integrating predictive IEQ into intelligent BMS. The findings show that GRU offers the best short-term accuracy and efficiency, CNN-LSTM excels at longer-range forecasting, and data quality factors such as sensor placement and occupancy patterns critically affect reliability, underscoring the potential for proactive HVAC control in net-zero buildings. These insights advance real-world smart-building operations by informing robust, energy-aware IEQ forecasting strategies with practical deployment considerations.

Abstract

Ensuring optimal Indoor Environmental Quality (IEQ) is vital for occupant health and productivity, yet it often comes at a high energy cost in conventional Heating, Ventilation, and Air Conditioning (HVAC) systems. This paper proposes a deep learning driven approach to proactively manage IEQ parameters specifically CO2 concentration, temperature, and humidity while balancing building energy efficiency. Leveraging the ROBOD dataset collected from a net-zero energy academic building, we benchmark three architectures--Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and a hybrid Convolutional Neural Network LSTM (CNN-LSTM)--to forecast IEQ variables across various time horizons. Our results show that GRU achieves the best short-term prediction accuracy with lower computational overhead, whereas CNN-LSTM excels in extracting dominant features for extended forecasting windows. Meanwhile, LSTM offers robust long-range temporal modeling. The comparative analysis highlights that prediction reliability depends on data resolution, sensor placement, and fluctuating occupancy conditions. These findings provide actionable insights for intelligent Building Management Systems (BMS) to implement predictive HVAC control, thereby reducing energy consumption and enhancing occupant comfort in real-world building operations.

Optimizing Indoor Environmental Quality in Smart Buildings Using Deep Learning

TL;DR

This work tackles the problem of energy-efficient, occupant-friendly IEQ management by forecasting CO, temperature, and humidity using deep learning. It benchmarks three architectures—LSTM, GRU, and a CNN-LSTM hybrid—on the ROBOD dataset to predict IEQ over short-term horizons ( minutes ahead) and assesses performance across multiple horizons with edge/cloud deployment options. Key contributions include a comprehensive architecture comparison, detailed preprocessing and feature engineering (cyclical time features, continuous intervals, and a fixed window), and concrete guidance for integrating predictive IEQ into intelligent BMS. The findings show that GRU offers the best short-term accuracy and efficiency, CNN-LSTM excels at longer-range forecasting, and data quality factors such as sensor placement and occupancy patterns critically affect reliability, underscoring the potential for proactive HVAC control in net-zero buildings. These insights advance real-world smart-building operations by informing robust, energy-aware IEQ forecasting strategies with practical deployment considerations.

Abstract

Ensuring optimal Indoor Environmental Quality (IEQ) is vital for occupant health and productivity, yet it often comes at a high energy cost in conventional Heating, Ventilation, and Air Conditioning (HVAC) systems. This paper proposes a deep learning driven approach to proactively manage IEQ parameters specifically CO2 concentration, temperature, and humidity while balancing building energy efficiency. Leveraging the ROBOD dataset collected from a net-zero energy academic building, we benchmark three architectures--Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and a hybrid Convolutional Neural Network LSTM (CNN-LSTM)--to forecast IEQ variables across various time horizons. Our results show that GRU achieves the best short-term prediction accuracy with lower computational overhead, whereas CNN-LSTM excels in extracting dominant features for extended forecasting windows. Meanwhile, LSTM offers robust long-range temporal modeling. The comparative analysis highlights that prediction reliability depends on data resolution, sensor placement, and fluctuating occupancy conditions. These findings provide actionable insights for intelligent Building Management Systems (BMS) to implement predictive HVAC control, thereby reducing energy consumption and enhancing occupant comfort in real-world building operations.

Paper Structure

This paper contains 23 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Methods
  4. Proposed System Architecture
  5. Research Question
  6. Dataset Characteristics and Preprocessing
  7. Data Cleaning and Preparation
  8. Feature Selection
  9. Addressing Time Gaps and Extracting Continuous Intervals
  10. Examples of Extracted Continuous Intervals
  11. (4) Feature Transformation
  12. (5) Dataset Splits
  13. Model Selection and Justifications
  14. LSTM Model
  15. GRU Model
  16. ...and 8 more sections

Figures (4)

  • Figure 1: Visual inspection of temperature, CO2, and humidity time series revealing substantial intervals with missing data (see flat lines and sudden jumps).
  • Figure 2: Example of a good continuous interval showing clear daily patterns in air temperature, indoor CO2, and humidity.
  • Figure 3: Example of a continuous interval with irregular or noisy readings, offering fewer obvious patterns.
  • Figure 4: GRU model predictions (dashed orange) vs. ground-truth measurements (solid blue) for air temperature, CO2, and humidity on the test set.