In the evolving landscape of smart cities, optimizing energy consumption and enhancing cybersecurity in Internet of Things (IoT) networks are crucial. This study leverages LoRa (Long Range) technology, Bayesian Inference, and Extreme Learning Machines (ELMs) to advance cyber threat intelligence and energy efficiency in large-scale IoT deployments. The proposed algorithms address key challenges within LoRa networks by implementing a novel energy consumption model, kernel-based ELM fine-tuning, Bayesian parameter tuning, and data fusion. Through Bayesian Inference, our approach dynamically adjusts network parameters to optimize packet transmission and collision rates, ultimately reducing power consumption across smart city applications. The Kernel-ELM algorithm fine-tunes LoRa network applications by adapting kernel parameters to the unique demands of IoT environments. Additionally, our Anomaly Detection and Fusion Algorithm (ADFA) integrates data from multiple sources to detect potential cyber threats, enhancing network security. Experimental results validate these algorithms on various datasets, demonstrating improvements in both energy optimization and threat detection. Ultimately, this study provides actionable insights into the deployment of scalable, energy-efficient, and secure IoT networks for smart city infrastructure.