Anomaly Detection in Smart Grids Using Blockchain-Integrated Collaborative Decision-Making Systems

The growing reliance on smart grids introduces significant challenges in detecting and mitigating anomalies, disrupting critical operations, and leading to system-wide failures. This thesis presents a novel framework for anomaly detection in smart grids by utilizing 10 distinct machine learning models—Autoencoder, LSTM, RNN, CNN, TCN, MLP, Random Forest, SVM, KNN, and Logistic Regression—trained locally on client-specific data. When making a real-time decision about anomalies, the clients share their own decisions as experts, and the system securely and collaboratively makes final expert-based global decisions. A key innovation of this framework is integrating blockchain technology for secure, collaborative, and decentralized anomaly validation. When a client detects an anomaly, it triggers a blockchain smart contract, initiating a voting process among the locally trained models. Each model will make its prediction and share it with the blockchain system via a smart function trigger. Following that, an average of these local decisions is taken, which results in a global decision ( majority voting) on the triggered anomaly. That global decision is tamper-resistant, transparent, and secure. The immutable nature of blockchain guarantees that all anomaly decisions are recorded and auditable, fostering trust among grid components. This hybrid approach improves anomaly detection accuracy without compromising data security. Performance evaluations using metrics such as accuracy, precision, recall, and F1-score demonstrate the effectiveness of the collaborative decision-making system and the enhanced reliability introduced by blockchain validation. By combining these technologies, the proposed system addresses key challenges in modern smart grids, paving the way for scalable, secure, and self-adaptive anomaly detection mechanisms capable of mitigating evolving cyber-physical threats. Index Terms – Smart Grids, Anomaly Detection, Machine Learning Models, Blockchain, Smart Contracts, Majority Voting, Secure Decision-Making.