| Abstract |
An efficient MPPT system is needed for large solar plants including several serially and parallel-connected PV panels especially functioning in partial shading conditions. A proportional decrease in electrical power will be a result of inhomogeneous irradiation (partial shading), as well as multiple local maximums that may appear. The presence of multiple local maximums is the most difficult obstacle for traditional MPPT which is based on a sequential search of an optimum working point. The present article submits a new algorithm that can be corresponded to the mathematical modeling with elements of Artificial Intelligence (AI). A proposed method is to use permanent monitoring of a voltage, a current, and temperature of each PV panel placed in the string. An MPPT algorithm determines the position of a global maximum (GM) based on this information and in accordance with the previously obtained math model of individual PV panels. Owing principles of AI, math models should be periodically précised during the service life of the PV plant. Since none of the presented math algorithms can provide localization of GM with the accuracy, required for the modern MPPT, the proposed method desire to be complemented by a conventional approach, let say perturbation and observation or incremental conductance techniques. For example, an algorithm finding zero roots of a power derivative versus current change was used in our work. The proposed algorithm can achieve GM with relatively high speed that is only restricted by digital control ability. Currently, this task would take no more than 50-100 ms maximum. Therefore, the global maximum can be found for any rapidly changing solar irradiation. |
