Introduction to clustering unsupervised machine learning algorithms applied to power quality disturbances

Unsupervised machine learning algorithms, such as clustering and anomaly detection, work by identifying patterns and anomalies in data without the need for labeled training data. These algorithms are particularly useful in situations where relevant training data is limited or unavailable, making them well-suited for predicting and detecting abnormalities and Power Quality Disturbances (PQD) in electric systems. PQD can have significant impacts on electrical equipment and systems, leading to equipment malfunction, downtime, and potential safety hazards. By applying unsupervised machine learning algorithms to electric data it is possible to gain valuable insights into the underlying patterns and characteristics of different PQDs. For example, clustering algorithms can group similar disturbances together based on their features, helping to identify common causes or sources of disturbances. Anomaly detection algorithms, on the other hand, can automatically flag and alert users to unusual or unexpected disturbances that may indicate potential issues. In this manuscript, a study of the current trends in the field of clustering and anomaly detection unsupervised Machine Learning (ML) algorithm is presented, dealing with approaches available for PQD detection. Overall, the combination of unsupervised machine learning algorithms and power quality analysis can greatly enhance the ability to predict, monitor, and mitigate the impact of PQD in various industrial settings. By leveraging these advanced technologies, reliability and efficiency of electrical systems can be improved, ultimately leading to a more sustainable and optimized environment.