Optimization of Energy Management and Anomaly Detection in Smart Grid Analytics Using Deep Learning

Effective anomaly detection in power consumption is essential for identifying inefficiencies in distribution networks. This research leverages the Indian Western-region power consumption dataset to evaluate two anomaly detection methods: the Isolation Forest and the OneClass SVM (Support Vector Machine). The Isolation Forest is particularly adept at detecting extreme anomalies in highdimensional data, while the One-Class SVM effectively identifies both major and minor outliers. Upon applying these models to normalized power consumption data, both demonstrate proficiency in identifying irregularities, each with distinctive advantages. The Isolation Forest identifies a concise, distinct group of severe anomalies, whereas the One-Class SVM detects a wider spectrum of irregularities. These results advocate for the integration of these techniques into power monitoring systems to enhance anomaly detection capabilities and operational efficiency. The study underscores the potential of machine learning to revolutionize management practices in power distribution systems.