Solution for selective harmonic optimisation in diode-clamped inverters using radial basis function neural networks

In this study, a new heuristic based intelligent algorithm is proposed for the elimination of harmonics in diode-clamped inverters. Numerous methods have been introduced for optimising harmonics in diode-clamped inverter such as Newton-Raphson method. These methods are not sometimes applicable because of difficulty of initializing algorithm parameters and singularity problems in Jacobian-based methods. Radial basis function neural network is used in the proposed method which is one of the most useful members of the neural networks family, because of their capability in the approximation problems. They work on the basis of this fact that, it has been proved that any arbitrary non-linear function can be exactly modelled by a combination of some non-linear basis functions. The proposed algorithm can be applied to solve the selective harmonic elimination problem of any multilevel inverters with any number of levels. The method has the benefit of high rate of convergence and accuracy besides simplicity in implementation. Theoretical results are authorised by experiments and simulations for a seven-level diode-clamped inverter. The obtained results prove efficiency and capability of the proposed method.