MPPT Technique Based on Neural Network for Photovoltaic System

Abstract

The using of an efficient MPPT (Maximum Power Point Tracking) algorithm influences a lot in the global efficiency of the PV system. This thesis presents a detailed study based on simulation of different MPPT algorithms with their features using two systems (off-grid and on-grid). The off-grid system contains a PV array connected to a boost converter and a resistive load. On the off-grid system a simulation is presented using MATLAB/SIMULINK platform with several MPPT algorithms. The simulated MPPT algorithms are the conventionals Incremental Conductance (IncCond), Perturb and Observe (P&O), Open Circuit Voltage (OCV) and a new developed Neural Network (NN) under different environmental conditions of temperature and irradiance. As a result of the simulation, the NN algorithm has a quick response, i.e, it requires less time to reach the MPP and high efficiency and less oscillation comparing with the conventional methods. On the other hand, a single-phase two-stage photovoltaic grid-connected system is simulated which contains a PV array, a boost converter, a dc link capacitor, an inverter, an output L filter and the utility grid. In that system a control of dc link voltage, the injected current and the MPPT is made. Another MPPT algorithm based on NN (modifiedNN) was also established. Showed later that is the most suitable for the system. The maximum of power is achieved when the irradiance is maximal and the temperature is minimal. Finally, a study of the influence of the variation in the climatic conditions on the output performance of the system is done.