IoT-Enabled Smart Grid System for Real-Time Energy Consumption Monitoring

Abstract

As renewable energy and Internet of Things (IoT) devices were integrated into energy grids, the need to monitor energy consumption in real-time has grown, and traditional systems face challenges in that regard due to weak security measures, poor latency in dynamic routing, and limited ability to detect anomalies in an environment of constantly changing energy generation and potential theft. In this paper, an IoT-based smart grid framework is proposed, combining the Harmonic Search Optimized Recurrent Neural Network for adaptive shortest-path routing, Improved Paillier Homomorphic encryption for privacy preserving data transmission and LightGBM with ambient intelligence context-awareness to provide timely and accurate anomaly and misuse detection. Comprehensive training and validation using a custom dataset, made from campus and household IoT sensors, then imbalanced with the addition of synthetic data, combined with hardware prototyping with Raspberry Pi, Arduino and XBee modules. Experimental results show that the model achieves 97.8% accuracy and 96.9% F1-score in anomaly detection which is superior to CNN-LSTM and standalone baselines, and 20-40% latency reduction and 0.015 kWh energy savings per cycle compared with conventional approaches. The proposed system holds promising implications regarding the stability of smart grid, loss reduction, and sustainable management of energy in smart cities which can be exploited for real-time monitoring of the proposed system in a scalable, secure and efficient manner.