NOVEL APPROACH BASED ON ARTIFICIAL NEURAL NETWORKS (ANN) FOR CONTROLLING THE THREE-PHASE INVERTER OF A PHOTOVOLTAIC SYSTEM
This paper is devoted to the study of a photovoltaic system connected to the grid. The objective is to provide a novel approach based on artificial neural networks for Controlling the three-phase invelter in order to ensure a flexible injection to unity power factor and low total harmonic dist01tion. A modeling system established mathematical models of the invener and the DC link. A control in the synchronous reference direct and quadratic frame by a neural proportional integral derivative based on back propagation of gradient descent is offered to regulate respectively the dc link voltage and the injected cunents in order to overcome the limitations ofthe experimental tuning (Ziegler-Nichols) of the classical PID controller parameters. Simulation results from the system under the Matlab / Simulink environment prove the efficiency of the proposed method and show that the currents injected have the best sinusoidal fonn with a cunent Total Harmonic Distortion of 0.96 % and a net followed of the dc link voltage.