Selective harmonic elimination (SHE) offers a way to control harmonic content in multilevel inverter outputs without relying on high-frequency switching. For industrial drives and renewable-energy interfaces, that can make it relevant where switching losses and waveform quality both matter. In cascaded H-bridge (CHB) inverters, however, finding suitable switching angles and voltage levels is a nonlinear optimisation problem. Metaheuristic methods can help identify feasible switching patterns and tune them to different design priorities.
Why modulation matters in multilevel inverters
Multilevel inverters combine several voltage levels to produce a stepped AC waveform. Compared with a two-level inverter, this approach can reduce waveform distortion and voltage stress across semiconductor switches. A CHB design builds the output from series-connected H-bridge cells, offering a modular structure that can be configured for different voltage levels.
The converter topology is only part of the design. Modulation determines when switches change state and how the output waveform is formed. High-frequency pulse-width modulation can provide control flexibility, but switching activity contributes to losses. Low-frequency methods are therefore of interest in high-power, medium-voltage applications where reducing switching losses is a priority.
SHE is one such low-frequency method. It selects switching angles to suppress specified harmonics while setting the fundamental component. This offers direct control over particular harmonic orders, but the available switching angles limit how many conditions can be addressed at once. The resulting equations are nonlinear and can have multiple solutions or no feasible solution for a given operating point.
Using metaheuristics to solve the SHE problem
Metaheuristic optimisation algorithms search for good solutions to complex problems without requiring a conventional equation solver to find an exact answer. In SHE and modified SHE (MSHE) modulation, the search variables can include switching angles and DC-link voltage levels. The optimisation objective can be framed around harmonic suppression, tracking the desired fundamental component, or a combination of performance measures.
Research on CHB inverters has assessed methods including particle swarm optimisation (PSO), the Imperialistic Competitive Algorithm (ICA) and a hybrid genetic algorithm. The findings indicate that algorithm performance depends on the modulation method and problem setup: PSO and ICA were effective in the examined SHE cases, while ICA performed better than PSO in some MSHE comparisons. The hybrid genetic algorithm produced weaker results in the reported comparisons. These findings are not a universal ranking; they underline the need to evaluate an optimiser against the specific inverter and operating requirements.
MSHE expands the formulation beyond conventional SHE. One approach relaxes the requirement to hold the fundamental component exactly at its target, allowing the optimiser to seek a broader feasible operating range and improved harmonic mitigation. The trade-off is that fundamental-component deviation becomes part of the design decision and must be assessed alongside harmonic performance.
More degrees of freedom—and more design choices
Adding variables can give the optimiser more flexibility. The investigated approaches include floating the fundamental component and allowing DC-link voltage levels to vary. These changes can increase the number of decision variables available for shaping the output waveform. Depending on the formulation, switching frequency and deviation rate can also be considered.
More degrees of freedom do not automatically guarantee a better practical design. The selected objectives and constraints determine what the optimisation prioritises. For example, a solution focused on suppressing targeted harmonics may differ from one that emphasises fundamental tracking or a wider usable operating band. Engineers should therefore define acceptable limits before comparing candidate switching patterns.
A multi-objective approach can make those trade-offs visible. Instead of returning only one solution, it can identify a set of Pareto-optimal options: solutions for which improving one objective may require compromising another. In the work described, the objectives include weighted total harmonic distortion (WTHD), deviation rate and chopping frequency. A Pareto front can help a design team choose a configuration according to its priorities rather than relying on a single combined score.
Accounting for DC-link voltage ripple
Idealised switching calculations assume stable DC-link voltages, but renewable-energy sources and other non-ideal conditions can introduce voltage ripple. Because SHE equations depend on the voltage levels used to construct the output, ripple can disturb the intended switching pattern and affect waveform quality.
One investigated mitigation strategy adjusts the effective reference value through switching control. In the described method, each full bridge is operated as a chopper so that its average voltage can track the required reference during a sampling interval. This approach addresses ripple in the modulation process, but it also introduces additional switching considerations. The design must account for the resulting chopping frequency and any configuration requirements associated with the switching scheme.
Practical implications for converter design
For engineers and system integrators, the central lesson is to treat modulation, optimisation and converter configuration as connected design decisions. A useful evaluation should consider:
- Waveform objectives: identify which harmonic components matter and whether fundamental tracking is strict or can be relaxed.
- Operating range: check that feasible switching solutions are available across the required modulation range.
- Switching impact: balance low-frequency operation and harmonic goals against any additional chopping activity.
- Source behaviour: include DC-link voltage variation when it is relevant to the application.
- Implementation: verify candidate patterns through simulation and, where appropriate, hardware testing.
Simulation and prototype testing reported for the examined methods provide evidence that optimised patterns can be implemented, but performance remains dependent on the selected scenario and converter configuration.
Conclusion
Selective harmonic elimination gives multilevel inverter designers a structured way to target low-order harmonics with low-frequency switching. Metaheuristic optimisation can help navigate the nonlinear equations involved, while MSHE, floating variables and Pareto analysis extend the available design choices. The practical task is to choose objectives and constraints that reflect the application, then assess waveform quality, operating range, switching requirements and DC-link ripple together. This approach supports a more deliberate trade-off between converter efficiency and output performance.