The rapid digitalization of district heating systems (DHSs) has driven the large-scale deployment of thermal Internet of Things (TIoT) sensors, which generate massive real-time operational data. Traditional centralized computing architectures struggle to process massive concurrent data. Furthermore, they fail to balance the stringent low-latency demands of real-time control tasks with the low-energy constraints of battery-powered terminal devices. To solve the complex problem of minimizing the weighted sum of system latency and energy consumption, we propose an Improved Mayfly Algorithm (IMA). The algorithm integrates five targeted structural enhancements: random position update masking, differential evolution (DE)-based crossover, targeted subset mutation with boundary scaling, adaptive population reset mechanism, and simulated annealing (SA)-driven local search, to efficiently navigate the high-dimensional rugged decision space and mitigate premature convergence. Extensive simulation results show that the proposed collaborative architecture achieves the lowest total system cost compared with traditional isolated computing paradigms (local-only, edge-only, and cloud-only). Notably, the proposed IMA reduces the total baseline weighted cost by 17.2% compared with the standard MA. Furthermore, under maximum practical industrial workloads (750 concurrent tasks, representing a highly complex 2250-dimensional MINLP space), the IMA maintains strong scalability and dominance, outperforming the second-best algorithm (BWO) by 15.8%. This research provides a low-latency, energy-efficient scheduling solution for TIoT-enabled DHS, and offers technical support for the intelligent and low-carbon transformation of urban energy infrastructure.
Chen et al. (Thu,) studied this question.