Parametric Analysis of a Grid Forming Inverter Fault Contribution
AI Blended Learning and Power Systems
Slightly delay from the previous post on the topic, as I was busy writing up an IEEE Journal paper.
The results of my poll were back, with a fairly even split between, RMS analysis in DIgSILENT (32%), EMT analysis in PSCAD (29%) and Advanced Stability & Control (23%), with only a handful for Model Audit and Validation (16%). One thing that was clear, and that is people want details rather than generic AI notes.
So for this post and future posts, I am going to focus specifically on AI and Power Systems Analysis as a blended learning approach and using AI to unlocks techniques that would be too complicated or time consuming to do by hand.
1. Grid Forming Inverter - Fault Contribution
In this study, I explore the fault contribution of a Grid Forming Inverter under fault conditions, and see how well they can be represented within formal fault calculation methods such as IEC 60909 and ENA G74. This is an open and real challenge with GFM steadily gaining more traction in real networks.
In the analysis I look at two Virtual Synchronous Machine (VSM) models, the UNIFI / WECC REGFM_B1 and a custom VSM by DIgSILENT - both of which are available in Powerfactory 2026 templates. The REGFM_B1 model is well understood and documented, but the DIgSILENT model has an additional level of detail above the REGFM_B1 and uses virtual impedance, and is not as clearly documented.
2. Grid Forming Inverter - Method
My approach was to use a hybrid approach of an MCP interface to Powerfactory to map and understand the VSM control elements of both models, and then run a series of parametric sweeps using python scripts to determine their sensitivity. The specific goal was to understand where the current limiter clamp operates in reality, how sensitive it is and what parameters define it.
I specifically used a positive sequence RMS approach rather than EMT, as a I was looking at a ‘slow’ transient fault contribution, rather than the inner working of the control systems. Plus Powerfactory RMS is much easier to drive via an MCP server than PSCAD is. There will be some further follow up work on PSCAD and GFM though.
The Analysis was carried out using a simple radial network, followed up by a more detailed analysis using the IEEE 39-Bus model, with 3x GFMs located at strategic levels in the system. This approach let me understand the behaviour in the simple network and then validate in a larger more ‘real’ system.
3. Summary
The more detailed analysis is in the PDF attached, but the key engineering takeaway is that you can approximate a GFM fault current reasonably well using a Full Size Converter model for a GFL. The key is to understand that the k factor can be used to define the transition point between a current limited (CL) output and a variable Voltage Source (VS) type output. Where used, the virtual impedance is a crucial factor and significantly affects (reduces) the calculated fault current in the VS regime. Download Technical%20Note%20001%20-%20Grid%20Forming%20Fault%20Contribution (PDF)