Deep Symbolic Regression-Based MPPT Control for a Standalone DC Microgrid System

Abstract

Enhancing the performance of maximum power point tracking (MPPT) methods is essential for optimizing the operation of solar systems under any weather condition. This improvement is crucial because MPPT methods serve as the controllers for power converters, directly influencing the ability to maintain a steady and efficient energy output from solar installations. Therefore, we propose a deep symbolic regression (DSR)-based MPPT method in standalone direct current microgrids, which is particularly useful under drastically shifted meteorological conditions. This study aimed to improve the significant performance parameters of the MPPT approach, such as the rate of convergence during transient periods, tracking accuracy, steady-state stability, and overshoot/undershoot reduction, creating more reliable and efficient solar microgrid systems. To evaluate the proposed method, performance comparisons were made with those of existing advanced algorithms, namely particle swarm optimization-, artificial neural network-, and adaptive neuro-fuzzy inference system-based MPPT methods. The comparative results demonstrate a significant improvement with the proposed method. Moreover, the simple mathematical expression obtained by the DSR process minimizes the computational complexity of the control.