Design and control of multiport isolated DC-DC converter for hydrogen energy systems

Oyewole, Oyedotun E.

Thesis

Design and control of multiport isolated DC-DC converter for hydrogen energy systems

Download PDF

Creator

Oyewole, Oyedotun E.

Rights statement

Strathclyde Thesis Copyright

Awarding institution

University of Strathclyde

Date of award

2026

Thesis identifier

T17688

Person Identifier (Local)

202263467

Qualification Level

Doctoral (Postgraduate)

Qualification Name

Doctor of Philosophy (PhD)

Department, School or Faculty

Abstract

Hydrogen energy storage systems (HESS) are vital for sustainable energy, and theirperformance is heavily influenced by their power electronics converter (PEC) interfaces.Conventional PEC interfaces, which use individually operating converters, often lead tobulky systems with multiple conversion stages and increased circulating currents, resultingin higher losses. This thesis addresses these shortcomings by implementing a multiportisolated DC-DC converter for HESS, which offers enhanced power density, a reducedcomponent count, and fewer conversion stages. However, this proposed converter introducesits own challenges, primarily the cross-coupling effect caused by the high-frequencymultiwinding transformer. This effect results in significant power deviations during stepchanges, risking hydrogen depletion in fuel cells and instability in electrolysers. Furthermore,the conventional single-phase shift (SPS) modulation used in such converters typicallyexperiences higher circulating currents, reducing overall efficiency.To mitigate the cross-coupling, the implementation of a decoupling control technique isessential. This study investigates and compares three simple matrix-based decouplingtechniques, finding the inverse matrix method to be the best performer, though it still fails tofully decouple the system over a wide operating range. To overcome this limitation, a novelmodel reference-based decoupling control technique is proposed, which minimises the errorbetween the actual system output and an ideal reference model. This technique is furtherenhanced into a hybrid decoupling control by integrating a decoupling matrix, ensuringrobust performance across a wider operating region by mathematically minimising the crosscoupling term with a proportional-derivative controller. This hybrid technique significantlyreduces maximum power deviations compared to the best matrix-based method. Recognizingthat these decoupling techniques generally require detailed system knowledge, whichcomplicates implementation due to parameter fluctuations, the thesis also proposes a solutionusing a linear active disturbance rejection controller (LADRC), which only requiresknowledge of the system order. However, as the LADRC is susceptible to estimation errorsand requires manual tuning, a particle swarm optimisation (PSO) algorithm is employed toautomatically determine the optimal controller gains. This PSO-optimised LADRCeffectively suppresses cross-coupling across a wider operating region, performing better thanboth non-decoupled controllers and a genetic algorithm (GA)-optimised LADRC. In parallel, to address the efficiency limitations of SPS control, which is limited in itsability to independently regulate inter-port power flow and leads to increased inductor currentand circulating power, the thesis also proposes an online adaptive control based on thesteepest descent method. This technique dynamically adjusts internal phase shifts in real-timeto minimise the error between reference and actual active inter-port power flows undervarying conditions, thereby enhancing operational efficiency. A comprehensivemathematical formulation, supported by Lyapunov-based stability analysis, confirms thestability of the adaptive gains in this proposed control. Finally, simulations and experimentalvalidations confirm the effectiveness and robustness of all the proposed control techniques,demonstrating significant improvements in system efficiency and performance across a wideroperating region for hydrogen energy storage applications.

Advisor / supervisor