Smart Thermal-Aware Task Offloading Model for Energy Optimization in Distributed Edge-Cloud Data Centre Networks

In the changing environment of edge-cloud computing, emerging deployments of the IoT require lowlatency, thermally steady, and power-efficient execution of tasks. Nevertheless, the majority of the models currently existing ignore thermal limits that may lead to a lack of efficiency and exhausted hardware. In this model, Temporal Fusion Transformer (TFT) is integrated with a forecast of temperatures and CNN to enhance processing of both space and time data and Deep Q-Network (DQN) to regulate policy decision-making. With both datasets of Kaggle like Temperature Readings and Cloud Performance Metrics, it is possible to use a large amount of data preprocessing to make accurate forecasts and plan work. The adaptive hybrid TFT-CNN-DQN model handles problems of overheating, wasting energy and feedback delay at the same time. It turned out that proposed approach achieved 91.2% accuracy in forecasting, keeping a higher score than GRU (79.9%) and Linear Transformer$(80.1 \%)$. With this plan, computing tasks are intelligently handled, cloud servers are kept cool and energy savings in edge-cloud networks are achieved, helping to make real-time computation sustainable.