Identification and maximum power point tracking of photovoltaic generation by a local neuro-fuzzy model

With the rapid proliferation of the DC distribution systems, special attentions are paid to the photovoltaic (PV) generations. This paper addresses the problem of maximum power point tracking (MPPT) for PV systems using a local neuro fuzzy (LNF) network and steepest descent (SD) optimization algorithm. The proposed approach, termed LNF + SD, first identifies a valid an accurate model for the PV system using the LNF network and through measurement data. Then the identified PV model is used for MPPT by SD optimization algorithm. The salient modeling abilities of the proposed LNF network results in a reliable and dependable PV model which takes voltage, temperature and insolation variations into account. The proposed approach is evaluated using several identification and MPPT simulations. The simulation results showed the accuracy of the LNF network in modeling of PV systems. Furthermore, simulations carried out for assessment of the MPPT performance during insolation transients demonstrated the high efficiency of the proposed LNF + SD approach for MPPT applications. Performance of the proposed method MPPT, while the PV array was supplying loads through DC-DC converters was also analyzed. Comparisons to the perturb-and-observe (P&O) and fuzzy logic methods revealed the superiority of the proposed approach.