Sustainable Urban Street Lighting through Predictive Maintenance using IoT and AI

The potential for transformative urban management is enhanced through the integration of IoT and AI, creating smart ecosystems that revolutionize intelligent lighting systems. This paper explores predictive maintenance solutions for urban street lighting, leveraging advanced machine learning models such as Temporal Fusion Transformer (TFT), Temporal Aggregation and Decomposition-based Attention (TADA), and Graph Neural Networks (GNNs). By analyzing historical operational data, these models detect subtle fault patterns and predict failures, improving system reliability, reducing maintenance costs, and optimizing energy use. The work contributes to United Nations SDGs 11 and 9 by promoting sustainable urban infrastructure and smart city development, fostering a future where IoT and AI systems enhance sustainability and resilience.