Smart Building Occupancy Prediction and Energy Optimization using IoT Sensors

Project Abstract

In this study, we present a comprehensive framework for predicting building occupancy and optimizing energy consumption through the integration of Internet of Things (IoT) sensors, data analytics, and machine learning-driven control strategies in a smart building environment. The research addresses the escalating energy demand of modern commercial and institutional facilities, where dynamic occupancy patterns significantly influence heating, ventilation, and air conditioning (HVAC), lighting, and plug-load energy usage. By deploying a heterogeneous sensor network that captures real-time data such as room occupancy, ambient conditions, light levels, and equipment activity, we construct a high-resolution dataset that reflects actual usage patterns across different days, times, and seasons. We develop and compare multiple predictive models, including time-series, ensemble, and deep learning approaches, to estimate short- and medium-term occupancy with high accuracy and low latency, enabling proactive energy management decisions. The occupancy predictions feed into a model-predictive control (MPC) framework and rule-based policies to dynamically adjust HVAC setpoints, ventilation rates, and lighting control while maintaining occupant comfort and indoor air quality (IAQ). To evaluate robustness, we test the system under various scenarios such as weekend/holiday calendars, sudden occupancy surges, and sensor outages, employing imputation and transfer learning techniques to mitigate data gaps. Our energy optimization module incorporates multi-objective optimization to minimize total energy consumption, peak demand, and unnecessary standby power, subject to comfort constraints defined by adaptive thermal comfort models tailored to different zones and user preferences. The framework also integrates demand-controlled ventilation to optimize IAQ without excessive energy penalties. A significant contribution of this work is the development of an end-to-end pipeline that includes data collection, preprocessing, feature engineering, model selection, real-time inference, and closed-loop control, with a focus on scalability and portability across building types. We validate the approach through a comprehensive case study conducted on a mid-sized office building equipped with a distributed sensor network and controllable sub-systems. Results indicate substantial energy savings, with reductions in HVAC and lighting energy use observed without compromising occupant comfort or IAQ, particularly during mid-day peak loads and transient occupancy fluctuations. Sensitivity analyses reveal the impact of prediction horizon, sensor density, and model accuracy on overall performance, providing guidelines for deployment in resource-constrained settings. Our work also explores the trade-offs between energy efficiency and occupant comfort, offering strategies for stakeholder acceptance, including user-adaptive preferences and transparent control rationale. By enabling intelligent occupancy-aware energy management, this study demonstrates the feasibility and benefits of deploying IoT-enabled predictive control in real-world buildings, contributing to smarter, greener, and healthier indoor environments. The findings support decision-makers in designing, retrofitting, and operating sustainable facilities, while laying the groundwork for further integration with building information modeling (BIM) and cloud-based analytics for large-scale urban energy optimization.

Project Overview

What This Project Is About

---

A straightforward exploration of how smart sensors in a building can monitor where people are and how much energy is used, then use that information to save energy while keeping comfort. The project looks at collecting data from everyday devices, turning it into simple insights, and applying those insights to adjust lighting, heating, and cooling automatically.

---

---

The Problem It Addresses

---

Buildings often waste energy because systems run on fixed schedules or react slowly to changes in occupancy. This project tackles the gap between actual use and system settings, aiming to reduce wasted energy without making occupants uncomfortable. It also addresses the challenge of turning raw sensor data into practical energy-saving actions.

---

---

Objectives of the Project

---

1. Understand how occupancy data can be collected safely and effectively.
2. Develop a simple model to predict room occupancy in real time.
3. Link occupancy predictions to energy controls (lighting, HVAC) for optimization.
4. Evaluate energy savings and occupant comfort in a test environment.
5. Explain the findings in an easy-to-understand way for stakeholders.

---

---

What You Will Do Step by Step

---

1. Survey available sensors and data sources in a building.
2. Set up a small test area with IoT devices (motion, door, temperature).
3. Collect data for a defined period and clean it for analysis.
4. Build a simple occupancy detection method using patterns in the data.
5. Create rules to adjust lighting and climate controls based on occupancy.
6. Test the system under different scenarios and measure outcomes.
7. Analyze energy use and comfort metrics before and after implementation.
8. Prepare a clear report with practical recommendations.

---

---

Expected Outcome

We expect a practical prototype that can predict occupancy and automatically adjust energy settings, leading to reduced energy use and maintained comfort. The project should produce a simple, repeatable process that building managers can understand and consider implementing in real settings.