AI-Based Fault Detection In Power Grids Using IOT Sensors

Bandari Tarun Yadav; Dr.Md.Asif

doi:10.62647/

Authors

Bandari Tarun Yadav B.Tech Student, Department Of Electronics and Computer Engineering, J.B Institute of Engineering and Technology, Hyderabad, India Author
Dr.Md.Asif Associate Professor, Department Of Electronics and Computer Engineering, J.B Institute of Engineering and Technology, Hyderabad, India Author

DOI:

https://doi.org/10.62647/

Keywords:

AI-Based Fault Detection, IoT Sensors, Smart Grid, LSTM, Predictive Maintenance, MQTT, Cyber-Physical Security.

Abstract

The AI-Based Fault Detection in Power Grids Using IoT Sensors is a comprehensive IoT and machine learning platform designed to modernize grid reliability and maintenance. Addressing the latency and data resolution limitations of traditional SCADA systems , the application integrates high- fidelity ESP32 sensor nodes, secure MQTT communication protocols, and a scalable Python/Django backend. It leverages advanced TensorFlow- based deep learning models, including Convolutional Neural Networks (CNN) for thermal imaging and Long Short-Term Memory (LSTM) networks for time-series anomaly detection.

Key features include real-time acquisition of electrical parameters (voltage, current, frequency) and environmental data , automated fault classification (normal, minor, critical), and predictive maintenance alerts. The system delivers high-performance metrics, maintaining sub-50ms API latency for inference requests and ensuring dashboard updates with under 1-second delay. During testing, the system demonstrated a 100% correlation rate for critical thermal anomalies after calibration and validated 99th percentile latency of under 100ms under a load of 500 concurrent requests.

Comprehensive testing validated the system's robustness across unit, integration, and security dimensions. The platform incorporates a proactive Intrusion Detection System (IDS) that achieved a zero-false-negative rate against integer-based data injection attacks, ensuring resilience against cyber-physical threats. Security measures include mutual TLS encryption for data transit and immutable, cryptographically signed audit logs for all detected faults, ensuring tamper resistance and accountability.

Deployed using a cloud-native microservices architecture with PostgreSQL for transactional storage, the application enables scalable, decentralized grid monitoring. Future enhancements include the implementation of Federated Learning for privacy-preserving model training across substations and the deployment of edge-AI algorithms on FPGA hardware to further reduce latency, positioning the solution as a foundational technology for autonomous, self-healing smart grids.