Optimization of Maximum Power Point Tracking in Photovoltaic Systems Using Hybrid Artificial Bee Colony Algorithm With Reinforcement Learning

This research aims to enhance the efficiency and accuracy of Maximum Power Point Tracking (MPPT) in photovoltaic systems by developing a hybrid algorithm that combines the Artificial Bee Colony (ABC) algorithm with reinforcement learning techniques. The proposed method has been evaluated under various laboratory conditions using data from three commercial solar panels obtained from the NREL database. The experiments encompassed three scenarios: variations in solar irradiance at a constant temperature of ۲۵°C, temperature variations under constant irradiance, and dynamic conditions where both variables changed simultaneously. For comparative assessment, two benchmark methods were selected: the analytical "Perturb and Observe" (P&O) method and the numerical "Incremental Conductance" method. In all phases, simulations were conducted using a consistent model to enhance the reliability of the results. The findings indicated that the hybrid ABC algorithm with reinforcement learning achieved a convergence speed ۱.۴۴ times faster than the Incremental Conductance method and delivered an efficiency close to ۹۹%. Furthermore, after reaching the optimal point, it exhibited fewer oscillations compared to the benchmark methods. Although the P&O method demonstrated higher speed under new environmental conditions, it resulted in greater power loss due to significant oscillations post-convergence. Conversely, the hybrid ABC algorithm with reinforcement learning effectively handled continuous disturbances through its stochastic exploration process, despite occasionally increasing computational time. Nevertheless, post-convergence oscillatory cycles were minimized. A primary challenge in employing this hybrid algorithm was the exploration and learning process in new conditions, which could extend initial tracking time. However, this issue was offset by reduced oscillations and enhanced operational stability. Under varying irradiance and temperature conditions, the hybrid ABC algorithm with reinforcement learning demonstrated satisfactory performance compared to the P&O method, albeit requiring more time to adapt to changes.

کلیدواژه ها:

Photovoltaic Systems ، Maximum Power Point Tracking (MPPT) ، Artificial Bee Colony (ABC) Algorithm ، Reinforcement Learning