Machine Learning and Deep Learning Based Hybrid Approach for Power Quality Disturbances Analysis

In order to achieve high accuracy and minimal errors, this study offers the identification and categorization of power quality faults using machine learning and a hybrid deep learning approach. In the initial phase, four distinct Boosting models incorporating both machine learning and deep learning techniques were employed to assess power quality occurrences. A hybrid model, Boosting CNN SOS, was devised to enhance the precision of power quality event identification and classification. The experimental setup produces numerous real-time power quality events, such as harmonics, surges, sags, transients, and flicker. Experimental data are cleaned and normalised during pre-processing. The evaluation of all criteria, such as mean absolute error (MAE), maximum error (ME), detection coefficient (d2), root mean square error (RMSE), and root mean square error (MSE), was carried out across all K-warehouse subsets to compare various models. The results mentions that using suggested novel model an accuracy rate of over 99% can be obtained. This suggested novel approach can effectively detect and categorize power failures in the smart grid, providing accurate and timely identification and classification. Additionally, it has the potential to be extended to encompass multiple IoT models that are based on the underlying infrastructure of smart cities.