Mitigation and Classification of Power Quality Disturbances Using Deep Intelligent Techniques

The research introduces an innovative optimization approach leveraging an LSTM classifier whose parameters are fine-tuned using the Sea Lion Optimization (SLO) algorithm, applied within a DSTATCOM framework to address power quality disturbances (PQDs). The designed voltage-controlled DSTATCOM with the LSTM-SLO-based PI controller was tasked with classifying and compensating for six different types of PQ events, achieving a classification accuracy close to 99.7% and an F1 score of 0.98, as evidenced by the confusion matrix. FFT analysis of the results indicated a substantial reduction in Total Harmonic Distortion (THD) to 0.57%, reflecting effective harmonic suppression. Furthermore, under voltage sag conditions, the compensator demonstrated its voltage restoration capability by improving the voltage from 190–220[Formula: see text]V. The computational efficiency of this LSTM-SLO model was also notably enhanced, achieving a 65% reduction in execution time over conventional PI controllers. Additionally, the power factor saw an improvement to 0.97, and the system maintained the DC link voltage precisely at the desired 1300[Formula: see text]V, ensuring stable operation. The algorithm was developed in Python on PyCharm 2022, with model training data sourced from GitHub, and simulations were conducted in MATLAB/Simulink 2019b to validate performance enhancements. This combination of techniques indicates strong potential for practical applications in power quality management.