Résumé:
"The efficiency, stability and reliability of a photovoltaic energy are considered major factors for
establishing this energy resource on the market. In this work, common maximum power point
tracking technique, based on artificial neural network using the RBF (Radial basis function) has been
proposed for a grid-connected PV system to maximize the output power of a PV array. The aim has
also been improving the stability and reliability of a PV power conversion, with a certain value of
temperature and radiation.
We began this thesis with identification of dynamic systems we explored methods for
identifying and understanding the complex behavior of a dynamic system. Among these methods,
we examined Least Squares, Recursive Least Squares, Particle swarm optimization, fuzzy logic as
well as the steps necessary for successful identification.
We then presented the photovoltaic systems and their main characteristics, their different main
components. Then we presented the different parameters and equations allowing the design of a
photovoltaic installation for a specific site.
The last part of our work addresses an in-depth study about artificial neural networks, we simulated
a code that predicts the power produced by a photovoltaic system.
The output power of the photovoltaic generator (GPV) depends on several climatic factors,
such as irradiation and temperature. However, real-time tracking of the optimum operating point
(MPP) is required to optimize system performance. In this work, we studied an intelligent modeling
neural network to extract the maximum power from the PV Array. The simulation results
demonstrated that our RBF network learned well, confirming this by the test values which gave
very approximate power values or almost equal to the real power values produced by the solar
panels. The characteristics of the PV were taken from electrical parameters of the BP SX 150S for
learning the network. "
