Hybrid CPU-GPU-Based Electromagnetic Transient Simulation of Modular Multilevel Converter for HVDC Application

Abstract

Modular multilevel converter (MMC) is one of the key components for modern high voltage DC (HVDC) transmission systems. Fast and accurate electromagnetic transient (EMT) simulation of the MMC is crucial to capture the finest dynamics and transients required for proper planning, design, and control prototyping of multiterminal DC (MTDC) grids. This paper proposes a hybrid high-performance computing platform for the simulation of the HVDC transmission system based on graphics processing unit (GPU). An EMT program is developed to implement the discrete time model for the two-terminal modular multilevel converters. The high computational power of the hybrid CPU-GPU platform is utilized to simulate all components in the HVDC system by capturing the detailed dynamics and transients. Moreover, a parallel GPU-based simulation algorithm is presented in the paper. The proposed algorithm provides an advanced solution for balanced distribution of simulation tasks on both the CPU and GPU to exploit the available resources on GPU at low communication latency without any compromise of the simulation accuracy. The results obtained from the proposed hybrid platform show high accuracy compared to the results obtained from Matlab/Simulink/Simscape/Specialized Power System model using the same network configuration and parameters. The proposed CPU-GPU MMC model is shown to achieve a speed up of 27.5 folds, compared to the sequential algorithm of MMC model implemented in CPU.