This paper proposes a fog-edge-enabled intrusion detection system (IDS) for smart grids (SGs) based on federated learning (FL) and support vector machine (SVM). The system addresses the challenges of data privacy, scalability, and real-time intrusion detection in SGs. Traditional IDS models are typically trained on cloud servers, which expose user data to privacy risks and increase detection latency. The proposed system uses FL to train edge devices, sharing only model parameters with the global model, ensuring data privacy while enabling collaborative learning. The system is evaluated on the NSL-KDD and CICIDS2017 datasets, achieving significant improvements in accuracy, recall, precision, and F1 score compared to existing methods. The fog-edge architecture enhances scalability, real-time responsiveness, and data privacy in SGs. The system's key contributions include a decentralized SVM-based collaborative model using FL, a distributed layered architecture with a fog-edge layer, formal threat models, and benchmark evaluations. The proposed model improves intrusion detection performance, preserves data privacy, and supports efficient, secure IDS in SGs. The system is designed to handle high-dimensional feature spaces and nonlinear decision boundaries, making it suitable for large-scale and resource-constrained environments. The model's decentralized nature enhances system resilience and reduces dependency on centralized servers, ensuring robust and efficient intrusion detection in SGs.This paper proposes a fog-edge-enabled intrusion detection system (IDS) for smart grids (SGs) based on federated learning (FL) and support vector machine (SVM). The system addresses the challenges of data privacy, scalability, and real-time intrusion detection in SGs. Traditional IDS models are typically trained on cloud servers, which expose user data to privacy risks and increase detection latency. The proposed system uses FL to train edge devices, sharing only model parameters with the global model, ensuring data privacy while enabling collaborative learning. The system is evaluated on the NSL-KDD and CICIDS2017 datasets, achieving significant improvements in accuracy, recall, precision, and F1 score compared to existing methods. The fog-edge architecture enhances scalability, real-time responsiveness, and data privacy in SGs. The system's key contributions include a decentralized SVM-based collaborative model using FL, a distributed layered architecture with a fog-edge layer, formal threat models, and benchmark evaluations. The proposed model improves intrusion detection performance, preserves data privacy, and supports efficient, secure IDS in SGs. The system is designed to handle high-dimensional feature spaces and nonlinear decision boundaries, making it suitable for large-scale and resource-constrained environments. The model's decentralized nature enhances system resilience and reduces dependency on centralized servers, ensuring robust and efficient intrusion detection in SGs.