<p>In industrial microgrids, power reliability, reduced downtime, and cost optimization are primary goals. Yet, such systems are becoming ever more susceptible to cyber-physical attacks like False Data Injection (FDI) attacks, which can undermine system stability, cause faulty control decisions, and inflict enormous financial losses. This paper suggests a k-nearest neighbors (k-NN)-based detection scheme to detect and prevent FDI attacks in industrial microgrids effectively. The proposed scheme is validated through large-scale simulations from a detailed Simulink model, evaluating its performance under various attack scenarios. Results indicate that with k = 5, the k-NN algorithm achieves a detection accuracy of 96%, a detection rate of 95%, and a false alarm rate of only 3%, outperforming Support Vector Machines (SVM) and Decision Trees (DT) in terms of precision and recall. Additionally, the proposed approach features a computationally lightweight architecture with an average detection time of 2.1 ms per sample, making it feasible for real-time industrial application. Comparative analysis further confirms the superiority of k-NN over traditional machine learning techniques in enabling improved microgrid resilience against cyber attacks. The outcomes demonstrate that the proposed scheme is successful in enhancing microgrid security and stability of operations and offers a feasible solution for real-world industrial applications.</p>