Edge-Based Digital Twin Architecture for SME Energy Moni-toring: A Hypothesis-Driven Empirical Validation in Food Manufacturing

Small and medium-sized enterprises (SMEs) in food manufacturing often lack affordable, deployable solutions for continuous energy monitoring that can support digital-twin-driven decision making. This study presents a low-cost edge–cloud monitoring architecture based on ESP32 gateways and commercial electricity meters (single-phase “Orman” and three-phase “Dala”), integrated with an MQTT pipeline, time-series storage, and dashboard visualization. The system was deployed in an operating production facility for 61 days, collecting 17.7 million samples from 16 operational nodes. To address real-world connectivity and power interruptions, an outage-aware data completeness metric and an edge buffering/backfill mechanism were implemented, enabling outage-adjusted completeness of 94.19% with a median of 99.88% across devices and recovering 53.7 h of measurements during nine facility-wide outages. The results also show that hardware-driven sampling heterogeneity (~30 s vs ~40 s) must be explicitly modeled when assessing data availability and reliability. Beyond monitoring, the proposed workflow produces actionable energy-efficiency insights: 75% of high-load assets operated at mean power factor below 0.9, indicating substantial potential for reactive power compensation. The proposed approach demonstrates a practical pathway for SMEs to establish scalable, resilient digital-twin monitoring with minimal infrastructure and clear steps toward advisory control and maintenance support.