Controlling of Multi-Level Inverter Under Shading Conditions Using Artificial Neural Network.
عبدالسميع عبدالفتاح القواسمة
In real life the PV sources can't supply multilevel inverters with equal and constant dc voltage. The variation of irradiation affects the output voltage of PV's which in turn vary the switching angles required to fire MLI to achieve minimum contents of output voltage profile , so the harmonic elimination’s equations must be solved for each set of input DC voltages. This research present how can we use genetic algorithm (GA) to solve harmonic elimination of eleven level inverters with equal and non-equal dc sources , then artificial neural network (ANN) is used to fire MLI with suitable angles for any set of input Dc sources . The partial shading of PV modules from clouds, obstacles are responsible for unequal Dc supply for multilevel inverter. A set of mathematical equations representing the general output waveform of the multilevel inverter with non-equal dc sources is formulated using Fourier series, then GA is used to solve the none linear equations to get the optimal set of switching angles which minimize the total harmonic distortion (THD) of eleven level inverter to acceptable limit, after that ANN is trained to generate these angles in any case of dc voltage variation in short time including constant dc sources when no shading FFT analyses are carried out for output voltage profile to prove that this technique is reliable for MLI; the proposed technique is validated through simulation by matlab Simulink Ra2013. GA and ANN technique achieve minimum THD for both equal and unequal DC sources, and can be applied for any kind of level inverter. According to our calculations to find THD for equal Dc sources we obtain 9.38%, and for variable Dc sources we obtained 10.26% THD when input Dc varied 4.47 volts, and 12.93% when input Dc varied 11.43 volts.