Decentralised Optimal Controller Design of Variable Frequency Three-Phase Power Electronic Networks Accounting for Sub-System Interactions

• Main Authors: David Dewar, Jaime Rohten, Andrea Formentini, Pericle Zanchetta
• Format: Article
• Language: English
• Published: IEEE 2020-01-01
• Series: IEEE Open Journal of Industry Applications
• Subjects: Decentralised control; optimal control; variable frequency; non-linear system; more electric aircraft
• Online Access: https://ieeexplore.ieee.org/document/9293351/

Electrification of non-propulsive aircraft systems has resulted in the increased proliferation of power electronics embedded grids on aircraft. Typically, power converters on these networks are optimised locally without consideration to the wider grid dynamics, which as a result increases the interactive effect between sub-systems. Filters are typically used to decouple these interactive effects, however since weight is a key design factor for aircraft, filters are typically reduced in size, further increasing interactive effects. Recent studies into <inline-formula><tex-math notation="LaTeX">$H_2$</tex-math></inline-formula> control, due to its ability to develop decentralised controls whilst considering the global grid dynamic model, have shown to reduce these interactive effects, however studies concentrate only on fixed frequency systems, atypical to modern variable frequency grids on modern aircraft today. In this paper, <inline-formula><tex-math notation="LaTeX">$H_2$</tex-math></inline-formula> optimisation is used to optimise a target converter to a pre-designed converter generating the variable frequency AC bus. The proposed method shows that not only are interactions reduced on the target converter for all frequencies but allows the pre-designed system to run as designed without detrimental performance, even during large power transients for full range of frequencies. This paper includes mathematical derivations, key design points, and has been validated and compared against other popular controls by experiment.