Smart Campus Energy Management Using Federated Learning and Edge Analytics

Project Abstract

The project presents a comprehensive, privacy-preserving framework for optimizing energy consumption in a university campus by leveraging federated learning and edge analytics, thereby enhancing grid resilience, reducing operational costs, and promoting sustainable practices without compromising user privacy. The study addresses the increasing complexity of campus energy systems, characterized by heterogeneous devices, variable occupancy patterns, and dynamic generation from on-site renewables. By integrating edge-enabled data processing with federated learning, the framework enables collaborative model training across multiple buildings and facilities while keeping sensitive consumption data localized, mitigating privacy risks and reducing bandwidth requirements for centralized data collection. The research defines an architectural blueprint that encompasses data collection from smart meters, occupancy sensors, HVAC systems, lighting controls, solar PV inverters, and energy storage units, all interfaced through secure, low-latency edge nodes and a central orchestration layer. A multi-objective optimization objective is formulated to minimize peak demand charges, maximize energy efficiency, and maintain occupant comfort, subject to thermal constraints, equipment ramp rates, and safety considerations. The federated learning component trains time-series forecasting and reinforcement learning models at the edge, including short-term load prediction, occupancy-aware demand response, and control policies for HVAC and lighting, with periodic aggregation at a trusted aggregator to improve global performance without exposing raw data. Techniques such as differential privacy, secure aggregation, and model distillation are employed to further enhance privacy guarantees. The edge analytics layer executes real-time decision-making, enabling adaptive scheduling of HVAC setpoints, lighting levels, and storage dispatch based on forecasted demand, renewable generation, and battery state-of-charge, while coordinating with campus-wide energy management policies and demand response events. The methodology includes a mixed-methods evaluation comprising simulation-based experiments using real campus telemetry datasets and a small-scale pilot deployment across selected buildings. Performance metrics include energy savings, peak-to-average ratio reduction, photovoltaic utilization, user comfort indices, latency, communication overhead, and privacy leakage assessment. Comparative analyses are conducted against centralized training baselines and conventional rule-based control strategies to quantify the trade-offs between privacy, accuracy, and responsiveness. The expected outcomes indicate that the federated approach can achieve comparable or superior predictive accuracy with substantially reduced data exposure, enabling scalable deployment across larger campuses or university networks. The research also explores governance, security, and interoperability concerns, providing a roadmap for integration with existing building management systems and standards. By delivering an end-to-end solution that couples privacy-preserving machine learning with edge intelligence, the project aims to demonstrate a practical pathway toward intelligent, sustainable campus energy ecosystems that can adapt to evolving energy markets, grid conditions, and occupant needs. Potential extensions include adaptive privacy budgets, multi-campus collaboration, and integration with microgrid control strategies for resilience against outages and grid disturbances.

Project Overview

What This Project Is About

---

A practical study that explores how a campus can save energy by collecting usage data from devices and buildings, learning patterns locally, and sharing insights without exposing private information. The project combines simple data collection with smarter decisions to reduce energy waste.

---

---

The Problem It Addresses

---

Universities consume a lot of electricity from lights, HVAC, and equipment. Centralized systems can be slow to adapt and may raise privacy concerns. The project seeks a balance between effective energy management, faster adaptation, and protecting user privacy.

---

---

Objectives of the Project

---

1. Understand energy usage patterns on a campus.
2. Implement a lightweight, local learning system on edge devices.
3. Improve energy efficiency without compromising privacy.
4. Demonstrate how collaborative learning reduces central data needs.
5. Evaluate performance with real or simulated data.

---

---

What You Will Do Step by Step

---

Step 1: Collect non-sensitive energy usage data from campus sensors (e.g., room temperatures, occupancy indicators, equipment power draw).

Step 2: Set up edge devices (e.g., gateways or local servers) to run lightweight analytics locally.

Step 3: Apply federated learning to learn common energy-saving models across devices without sharing raw data.

Step 4: Deploy edge analytics to trigger energy-saving actions (e.g., adjust lighting, HVAC) in real time.

Step 5: Evaluate savings, response times, and privacy outcomes, using simple metrics.

---

---

Expected Outcome

An operational blueprint for smart energy management on campus that reduces electricity use, respects privacy, and can be scaled to other buildings or campuses.