An IoT digital twin framework for smart waste and energy management optimazation in Maiduguri
Abstract
The rapid pace of urban development in cities like Maiduguri, Nigeria, brings significant operational challenges to municipal waste collection and energy distribution. This study presents a six-layer Digital Twin (DT) architectural framework designed for resource-constrained urban environments that integrates resilient Internet of Things (IoT) sensing, machine learning (ML) forecasting, and agent-based co-simulation. During a six-month pilot deployment, data was obtained from 10 smart waste bins with ultrasonic sensors, servo actuators, and GSM modules, and 3 distribution-node energy meters. To predict bin fill levels and energy demand without future information data leakage, time-series forecasting models including Long Short-Term Memory (LSTM) networks, Random Forest, and XGBoost were evaluated using 5-fold rolling-origin cross-validation. The LSTM model gave the best predictive accuracy (R²=0.92, RMSE = 4.8%). The forecasted horizons were fed into an AnyLogic agent-based simulation with Ant Colony Optimisation (ACO) for dynamic collection vehicle routing. Across 30 stochastic simulation runs, the proposed DT framework achieved an 18.2% pm 1.4% reduction in total collection distance (and 15.4% pm 1.1% reduction in fuel consumption), a 41.8% pm 3.1% reduction in potential waste overflow events, a 10.3% pm 0.8% reduction in fleet CO₂ emissions, and an 11.6% pm 1.1% improvement in peak load-balancing efficiency relative to the current static baseline. The architecture uses local buffering in EEPROM to ensure data continuity even when network connectivity is intermittent. The results provide a proof-of-concept for moving from reactive municipal scheduling to data-driven decision making in infrastructural fragile environments.