Big Datasets for Intelligent Anomaly Detection, Mitigation, and Forecasting in Feeders to Secure Smart Grid

Big Data (BD) offers new opportunities for scientists to detect both normal and anomalous events in the Smart Grid (SG). However, the diverse and complex nature of Internet of Things (IoT)-enabled Feeder (FED) datasets poses significant challenges in analyzing and visualizing the real impacts. As a result, existing Centralized Protection, Control, and Monitoring (PCM) solutions often fail to uncover hidden patterns, making it difficult to distinguish between original and anomalous events in FEDs. Consequently, stealthy anomalies can bypass PCM anomaly detection layers, leading to a decline in the overall resilience and security of the smart grid. Artificial Intelligence (AI) shows great potential for uncovering hidden correlations and patterns in large datasets generated from FEDs in the SG. To address this, a Self-learning Hybrid Machine Learning (SHEL) model has been designed to process BD for detecting, mitigating, and forecasting anomalies caused by false data injection in the energy network. The BD first undergoes rigorous preprocessing, including noise reduction, temporal alignment, and statistical feature extraction in FEDs. Afterward, the SHEL model learns the hidden patterns with high precision and exposes stealthy anomalies based on the real-time input BD stream and historical BD in coordination with the Supervisory Control and Data Acquisition (SCADA) system. This enables the Distribution System Operator (DSO) to gain deeper insights into cybersecurity threats that escape PCM mechanisms and could compromise the smart grid security. Moreover, the regenerated BD can support further research in fault diagnosis and forecasting, enhancing the resilience and security of the SG.

大数据（Big Data, BD）为科研人员在智能电网（Smart Grid, SG）中检测正常与异常事件提供了全新机遇。然而，支持物联网（Internet of Things, IoT）的馈线（Feeder, FED）数据集兼具多样性与复杂性，为其实际影响的分析与可视化工作带来了严峻挑战。因此，现有集中式保护、控制与监测（Centralized Protection, Control, and Monitoring, PCM）方案往往难以挖掘隐藏模式，难以区分馈线中的原始事件与异常事件。进而，隐蔽性异常可绕过PCM的异常检测层，导致智能电网整体韧性与安全性下滑。人工智能（Artificial Intelligence, AI）在挖掘智能电网馈线所生成的大规模数据集中的隐藏关联与模式方面，展现出巨大应用潜力。为此，研究人员设计了一款自学习混合机器学习（Self-learning Hybrid Machine Learning, SHEL）模型，用于处理大数据以检测、缓解并预测能源网络中由虚假数据注入引发的各类异常。该模型首先会对大数据开展严格的预处理操作，涵盖降噪、时间对齐以及馈线数据集的统计特征提取。随后，SHEL模型可结合监控与数据采集（Supervisory Control and Data Acquisition, SCADA）系统，基于实时大数据流与历史大数据高精度地学习隐藏模式，并识别出隐蔽性异常。这使得配电系统运营商（Distribution System Operator, DSO）能够更深入地洞察那些绕过PCM机制、可能危及智能电网安全的网络安全威胁。此外，经重构后的大数据可支持故障诊断与预测领域的进一步研究，进一步提升智能电网的韧性与安全性。