Sea Lion Optimization Algorithm for the Classification and Elimination of Power Quality Disturbances in Distribution Network

Abstract

Power quality disturbances (PQDs), including voltage sags, swells, harmonics, transients, flickers, and interruptions, affect the reliability and efficiency of modern power distribution systems. This study introduces a novel heuristic model that integrates a long short-term memory (LSTM) deep neural network with the sea lion optimization (SLO) algorithm for precise classification and elimination of PQDs. The system comprises a voltage-controlled distribution static compensator (VCM-DSTATCOM), including an LSTM-SLO optimized PI controller to enhance reactive power compensation and voltage regulation performance. The proposed LSTM-SLO classifier is executed using the SLO algorithm, which enhances hyperparameter optimization, increases accuracy, and reduces computing time. All simulations and coding were conducted in MATLAB/Simulink 2019b on a 400 V, 50 Hz distribution network and PyCharm 2022. The classifier achieved a test accuracy of 99.10% with a convergence rate of 0.97%. The proposed VCM-DSTATCOM, utilizing optimal gains of PI controllers, effectively eliminated PQDs and reduced total harmonic distortion (THD) to 0.52%, compared to 15.17% with a conventional PI controller; furthermore, voltage stabilization was achieved with instrument response times under 20 ms. This study offers a practical method for addressing real-time PQ problems, with potential applications in smart grid power distribution. Future endeavors can focus on a customized hardware solution that can be integrated with the IoT environment for enhanced monitoring and control tasks.