A "nearly parametric" solution to Selective Harmonic Elimination PWM

TL;DR

This paper introduces a novel algorithm for Selective Harmonic Elimination PWM that uses a 'nearly parametric' approach based on Rational Univariate Representation, enabling efficient computation of switching angles as functions of the modulation index.

Contribution

It presents the first use of a symbolic modulation index parameter in SHEPWM solutions, avoiding complete recomputation when the index changes.

Findings

The algorithm computes solutions for 460 modulation indexes in 0.0284 seconds.

Simulation results verify the correctness of the proposed method.

The approach is practical for real-time applications with fixed switching angles.

Abstract

Selective Harmonic Elimination Pulse Width Modulation (SHEPWM) is an important technique to solve PWM problems, which control the output voltage of an inverter via selecting appropriate switching angles. Based on the Rational Univariate Representation (RUR) theory for solving polynomial systems, the paper presents an algorithm to compute a "nearly parametric" solution to a SHEPWM problem. When the number of switching angles N is fixed, a "nearly parametric" solution can be considered as functions of the modulation index m. So we can adapt the amplitude of the output voltage with the same source voltage by changing the modulation index. When m is given as a specific value, complete solutions to the SHEPWM problem can be obtained easily using univariate polynomial solving. Compared with other methods, m is considered as a symbolic parameter for the first time, and this can help avoid totally restarting when m changes. The average time for computing complete solutions associated to 460 modulation indexes based on a "nearly parametric" solution when N=5 is 0.0284s, so the algorithm is practical. Three groups of switching angles associated to N=5, m=0.75 is simulated in MATLAB, and it verifies the algorithm's correctness.