Improved MPPT algorithm: Artificial neural network trained by an enhanced Gauss-Newton method

Abstract

<abstract> <p>A novel approach defined by the artificial neural network (ANN) model trained by the improved Gauss-Newton in conjunction with a simulated annealing technique is used to control a step-up converter. To elucidate the superiority of this innovative method and to show its high precision and speed in achieving the right value of the Maximum Power Point (MPP), a set of three comparative Maximum Power Point Tracker (MPPT) methods (Perturbation and observation, ANN and ANN associated with perturbation and observation) are exanimated judiciously. The behavior of these methods is observed and tested for a fixed temperature and irradiance. As a result, the proposed approach quickly tracks the right MPP = 18.59 W in just 0.04382 s. On the other hand, the outstanding ability of the suggested method is demonstrated by varying the irradiance values (200 W/m<sup>2</sup>, 300 W/m<sup>2</sup>, 700 W/m<sup>2</sup>, 1000 W/m<sup>2</sup>, 800 W/m<sup>2</sup> and 400 W/m<sup>2</sup>) and by varying the temperature values (15℃, 35℃, 45℃ and 5℃). Therefore, the ANN trained by Gauss-Newton in conjunction with simulated annealing shows a high robustness and achieves the correct value of MPP for each value of irradiance with an efficiency 99.54% and for each value of temperature with an efficiency 99.98%; the three other methods sometimes struggle to achieve the right MPP for certain irradiance values and often remains stuck in its surroundings.</p> </abstract>