Development and analysis of dynamic network equivalent representation for HVDC dynamic performance studies

With the increasing integration of renewable generation, power systems are transitioning from predominantly Synchronous Machine (SM) based systems containing simple rotor dynamics governed by a standard swing equation, to systems with a higher penetration of Power Electronic (PE)-components . This introduces new stability challenges due to different dynamic characteristics of PE-based technologies. When verifying and designing High Voltage Direct Current (HVDC)-systems it is necessary to simulate their interaction with the surrounding network using Electromagnetic-Transient (EMT)-simulation to capture faster dynamic phenomena. However, EMT simulation is computationally intensive which makes simulation of an entire power system impractical. To reduce the computational burden, networks are reduced to retain only the elements contributing to the dynamic response at the Point of Common Coupling (PCC), forming the retained study area and network equivalents representing the removed external system.  

Traditionally the boundary of the study area is represented by static voltage sources with fixed voltages and frequencies. This has been argued to not capture the low frequency dynamics in the external system. This can be solved by either expanding the study area or replacing the static boundary sources with dynamic models. This study assesses the impact a dynamic boundary model has on the fault response at the HVDC:s PCC.

A dynamic network equivalent model was developed using a standard Alternating Current (AC) voltage source and first order single axis SM equations to represent internal voltage dynamics. Low frequency angle dynamics are governed by the swing equation and a simple governor model. The boundary buses were assessed to find which were necessary to represent dynamically using Prony analysis of voltage angle swings against the PCC. Lastly a optimization algorithm was employed to identify model parameters for a good dynamic response match at the PCC.

Results show that a dynamic boundary retains the low frequency dynamics well compared to a static boundary model. All boundary buses do not need to be modeled dynamically and using the grading method the buses to be modeled dynamically can be decided. Optimization algorithms developed could not reach any good results and many parameters needed to be tuned manually. To better identify model parameters the study area selection and optimization algortihms should be re-evaluated.