Smart Contract-Driven Anomaly Alerts: An AI-Enabled Framework for Energy Grids

Abstract

This paper proposes a blockchain framework integrating smart contracts and machine learning to enable secure, decentralized anomaly detection in energy grids. We deploy Ethereum-based smart contracts to trigger localized alerts by postcode, validated on the Ausgrid dataset. Experimental results demonstrate the framework’s ability to achieve 94.5% anomaly detection accuracy while reducing false alerts by 30% through a two-stage machine learning pipeline. First, the MeanShift clustering algorithm identifies irregular consumption patterns using adaptive interquartile range thresholds, followed by supervised classification. Various algorithms are investigated, with and without SMOTE, to tackle the problem of dataset imbalance. The Proof-of-Stake (PoS) consensus mechanism reduces energy overhead by 99% compared to traditional Proof-of-Work (PoW), ensuring scalability for real-time grid management. By automating postcode-specific alerts and leveraging blockchain’s tamper-proof data storage, the framework enhances operational responsiveness and transparency for decentralized energy systems. This work bridges the gap between decentralized ledger technology and AI-driven analytics, offering a practical solution for secure, low-latency anomaly management in modern smart grids.