Abstract—Pulse Width Modulation (PWM) using the harmonic elimination technique needs the solving of a nonlinear transcendental equations system. Conventionally, due to their high complexity, these equations have to be solved off-line and the calculated optimal switching angles are stored in look-up tables or interpolated by simple functions for real-time operation. System flexibility is very limited, especially for applications which require both amplitude and frequency control. A new implementation scheme based on real-time solving of the nonlinear harmonic elimination equations using feed forward Artificial Neural Networks (ANNs) is reported in this paper. Based on the well known Back-propagation Algorithm (BPA), two training schemes for the ANN are presented. In the first one, the ANN is trained using the desired switching angles given by the classical method. The second training scheme is developed using only the harmonic elimination equation systems. Some simulation results are given to show the feasibility, performances and technical advantages of the proposed method.
Index Terms—Artificial neural networks, solving algorithm, harmonic elimination algorithm, multilevel inverter.
O. Bouhali and F. Fatiha are with Mecatronic Laboratory, Jijel University, Algeria (email: firstname.lastname@example.org).
N. Rizoug is with the Ecole Supérieure des Techniques Aéronautiques et de Construction Automobile (ESTACA) (e-mail: email@example.com).
A. Talha is with the Electrical Engineering Department of University of Science and Technology Houari Boumediene (e-mail: firstname.lastname@example.org).
Cite: O. Bouhali, F Bouaziz, N. Rizoug and A. Talha, "Solving Harmonic Elimination Equations in Multi-Level Inverters by Using Neural Networks," International Journal of Information and Electronics Engineering vol. 3, no. 2, pp. 191-195, 2013.