Cyber-attacks represent a huge danger to Smart Grid infrastructure, causing substantial interruptions in electricity supply as well as severe economic and social consequences. As a result, there is a need for an efficient and effec- tive threat detection mechanism for security of the Smart Grid infrastructure. In this research, we offer a design for a threat detection system based on the Relaxed Greedy Method for Smart Grid architecture. The suggested frame- work is based on the Relaxed Greedy algorithm, a heuristic-based technique to optimising problems. This approach is well-known for its efficiency, efficacy, and simplicity in tackling large-scale optimization problems to detect possible dangers in the Smart Grid infrastructure based on the collected attributes. The suggested system is tested using a real-world dataset taken from a Smart Grid testbed. The experimental findings suggest that the proposed framework can identify various forms of threat detections in the Smart Grid infrastructure.