University of British Columbia
Hybrid CPU-GPU parallel and multi-rate electromagnetic transient simulation of multimodule multilevel converters using decoupled detailed equivalent model
Abstract
dc:descriptionMultilevel and multimodule voltage-source converters (VSCs) have become key enabling technologies for modern power systems due to their scalability, modularity, and controllability. Among these converters, modular multilevel converter (MMC) is widely adopted in High Voltage Direct Current (HVDC) transmission systems for renewable energy integration. Electromagnetic transient (EMT) studies are essential for fault analysis and grid stability assessment. However, the large number of switching nodes results in a high-dimensional, time-varying conductance G matrix and extensive computational complexities. Large time steps reduce the computational burden at the expense of simulation accuracy. Therefore, this research develops a fully decoupled detailed equivalent model (D-DEM) for efficient and accurate EMT-type simulation. The D-DEM achieves constant conductance G matrix and reduced number of nodes while preserving converter dynamics under deblocking and blocking modes. Its accuracy is validated against a detailed model (DM) using Simulink/Simscape Electrical toolbox and PSCAD/EMTDC. Controller-hardware-in-the-loop (CHIL) experiments demonstrate its viability under real-time simulation constraints. A multi-rate simulation technique is adopted to simulate MMC subsystems with different time steps. Combined implicit-explicit (ImEx)-type multi-step Backward Differentiation Formula (BDF) solvers are investigated for simulation accuracy enhancement. The ImEx-type multi-step solver has improved accuracy compared to first-order solvers and is more robust against fictitious numerical oscillations compared to Trapezoidal Rule-based solvers. Switching interpolation is integrated with the multistep solvers to accurately represent intra-time-step switching events. To further accelerate MMC EMT simulation, a hybrid parallel computing EMT solver using central and graphical processing units (CPU-GPU) is implemented. Different kernel grids process different MMC arms at the same time, and each thread carries out computations for a single submodule. The hybrid CPU-GPU D-DEM achieves 79-fold speed up than its CPU-only sequential implementation for the MMC with 400 SMs per arm. The applicability of the proposed decoupled DEM strategy and hybrid parallel CPU-GPU platform is extended to solid-state transformer (SST) systems. A decoupled DEM of the SST is developed and implemented on the proposed platform while preserving switching-level accuracy. The model is evaluated under normal, transient, and IGBT-blocked operating conditions. Simulation results verify accurate and efficient EMT-type simulation.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy - PhD
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Electrical Engineering
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Hatahet, Walid
- Contributors dc:contributor
-
- Wang, Liwei
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/95548