Data-driven based Power Quality Disturbance Analysis for Improved Reliability in Smart Grids

One of the many challenges encountered within smart grids is maintaining Power Quality (PQ) due to various internal and external disturbances. Addressing this issue is crucial as PQ disturbances can cause inefficiency, downtime, and damage to sensitive equipment, which can impact grid reliability and service quality. This study investigates the use of dimensionality reduction techniques, namely Correlation Analysis (CA) and Principal Component Analysis (PCA), to improve the reliability of smart grids by enhancing power quality (PQ) disturbance classification. The study examines the impact of CA and PCA on the performance of four commonly used machine learning classifiers - Artificial Neural Networks (ANN), K-Nearest Neighbors (KNN), Decision Trees (DT), and Gradient Boosting Trees (GBT) - using a large dataset. The findings indicate that reducing dimensionality can considerably enhance computational efficiency for all models. Specifically, ANN and GBT demonstrate significant reductions in execution time, while KNN maintains unparalleled speed, highlighting its potential for rapid PQ disturbance detection in smart grids. On the other hand, DT and GBT algorithms show a good balance between accuracy and computational efficiency, making them robust solutions for smart grid applications where both accuracy and speed are required.