Edge Computing-Based Building Energy Management Systems for Campus Buildings: A Comparative Evaluation of Weighted-Sum and Pareto-Based Multi-Objective Optimization Methods on a Raspberry Pi 5
TL;DR
The findings confirm that the proposed edge-based BEMS framework is both computationally feasible and effective, offering practical guidance for selecting between weighted-sum and Pareto-based optimization strategies in real-time campus energy management.
Abstract
Recent advancements in Building Energy Management Systems (BEMS) have integrated Internet of Things (IoT) and edge computing technologies to enhance energy efficiency. Most existing studies focus on simulation-based evaluations or cloud-centric architectures, while comparative evaluations of optimization algorithms on resource-constrained edge devices remain underexplored. This study proposes an edge-based BEMS framework and systematically evaluates the performance of four multi-objective optimization algorithms, namely weighted-sum Genetic Algorithm (GA), weighted-sum Particle Swarm Optimization (PSO), Multi-Objective GA (MOGA), and Multi-Objective PSO (MOPSO), implemented directly on a Raspberry Pi 5 device. The weighted-sum GA and PSO are single-objective scalarization approaches applied to a multi-objective problem, whereas MOGA and MOPSO are true Pareto-based multi-objective optimization methods. The optimization objectives are to minimize energy consumption and occupant thermal discomfort simultaneously and the considered algorithms were evaluated using real-world sensor data from an IoT-based monitoring system in five campus rooms over a period of five days. The experimental results show that PSO-based methods, specifically the weighted-sum PSO with a discomfort weight of 0.5, which achieved the lowest Euclidean distance to the ideal point of 4.46, outperformed the GA-based methods. Among Pareto-based methods, MOPSO produced a smoother and more uniformly distributed Pareto front than MOGA, with a superior knee-point solution (Euclidean distance of 5.59). Energy savings of 14.19% were achieved by the weighted-sum PSO with a discomfort weight of 0.1. The execution times were below 0.03 s with an average CPU load of approximately 25%. The findings confirm that the proposed edge-based BEMS framework is both computationally feasible and effective, offering practical guidance for selecting between weighted-sum and Pareto-based optimization strategies in real-time campus energy management.
