Anomaly Detection in Electrical Systems Using Machine Learning and Statistical Analysis

The reliability of electrical systems is essential for uninterrupted power supply across residential, industrial, and commercial sectors. This research utilizes the various techniques of ML in identifying all sorts of abnormal electrical power variables, including voltage fluctuations, line faults, and frequency variations that emanate from several causes like equipment failures and environmental ones. Voltage and current measurements were pre-processed along with other additional parameters such as frequency and THD, then visualized to assure its quality and relevance for training the ML models. The work implemented the algorithms of Random Forest, Decision Tree, XGBoost, SVM, and Logistic Regression-the Random Forest Classifier performed best for improving results with the best accuracy 99%, precision of 99%, and recall 99%. Random Forest yielded much higher anomaly detection with better generalization after an appropriate avoidance of the overfit to data via effective ensemble. Performance measures obtained against the confusion matrix will really underscore that it makes pretty good generalizations to an independent/unseen data distribution also. This research underlines the potential of data-driven techniques to enhance fault detection, improve predictive maintenance, and ensure grid reliability. Future work may focus on real-time deployment and scalability to integrate such systems into operational grid environments for proactive anomaly management.