Accuracy and Robustness Assessment of ANN and Fast Convergence Approaches for Solar Power Optimization

The conversion chain forms the core of the photovoltaic (PV) system upon which the algorithms used are crucial for optimizing the aforementioned conversion chain. This article addresses an exploration of the optimization of the energy produced by the PV panel by comparing the use of artificial intelligence, that is, artificial neural networks (ANN), with our previously proposed fast convergence approach (FC). The utilization of ANN demonstrates their efficiency and stability for maximum power point tracking (MPPT) against varying irradiation and temperature. Both methods demonstrate their robustness and offer high performance under varying weather conditions; however, the results verify that they are different in certain instances. Subject to varying irradiation but stable temperature scenarios, the FC technique exhibits a mean absolute percentage error (MAPE) of 1.08%, compared to that of 1.12% by the ANN. That indicates that both algorithms can track effectively the maximum power point but the FC technique tracks more, despite varying weather conditions especially variable irradiation. Similarly, subject to stable irradiation and varying temperature scenarios, the FC technique verifies a MAPE of 2.18%, also lower than that of the ANN of 2.67%. These results indicate that the ANN algorithm present a good stability in permanent regime when the FC technique offers a more precision despite the presence of oscillations in its steady-state response.