Identificador: TDX:1688
Autores:
Muñoz Ramirez, Oscar Mauricio
Resumen:
Multilevel inverter (MI) topologies can work at higher voltage and higher power than conventional two-level converters. In addition, multilevel conversion reduces the output variables harmonic distortion and, sometimes, in spite of the devices-count increment, the conversion losses can also decrease by increasing the number of levels. The harmonic distortion reduction achieved by increasing the number of levels, can be used to further reducing the switching losses by decreasing the inverter carrier frequencies. To reduce even more the switching frequency without degrading output spectrum, we control the triangular carrier waveforms slopes. First, to achieve this target, two analytical models have been created in order to predict the inverter output voltage spectrum, depending on diverse parameters: the amplitude modulation index MA, the voltage distribution K of the inverter input sources, and the four carrier slopes {r1, r2, r3, r4}. The first model considers Natural Sampling and is based on Double Fourier Series (DFS) whereas the second model based on Simple Fourier Series (SFS), introduces the concept of Pseudo-Natural Sampling, as a digital approximation of the natural modulation. Both models are programmed in Matlab, verified with Pspice simulations and validated with a first experimental prototype with a DSP digital modulator.The good agreement between natural and pseudo-natural modulations, as well as their respective DFS and SFS models, is exploited by a Genetic Algorithm (GA) application where THD is the cost function to minimize. After testing and properly tuning the GA, a framework matrix containing the optimized carriers set for a specific range of variables {MA,K} is generated and then, tested with a second, closed-loop prototype. A slow digital loop modi
Repositori URV