Introduction
As the global energy landscape shifts from traditional synchronous generators to inverter-based resources (IBRs) like solar photovoltaics and battery storage, the complexity of grid protection analysis has increased significantly. Unlike traditional generators, which have predictable physical responses to short circuits, inverters are controlled by software loops that actively limit current to protect sensitive power electronics.
In optimization-based frameworks such as PowerModelsProtection.jl, representing these non-linear, non-convex control limits often results in mathematical infeasibility. This paper presents simplified, continuous-variable mathematical models for grid-following (GFL) and grid-forming (GFM) inverters designed to provide stable and scalable solutions for modern microgrid protection studies.
Proposed Modeling Framework
The authors utilize Thevenin equivalent circuits—a voltage source behind an impedance—to approximate inverter behavior. The critical innovation lies in the use of continuous activation variables ($z$ and $r$) to manage the transition into current saturation. By avoiding discrete or integer variables to represent the "switch" into a current-limited state, the authors ensure the mathematical problem remains smooth enough for standard optimization solvers to handle efficiently.
- Grid-Following (GFL) Model: Functions as a constant-power source until reaching a current threshold, at which point it transitions to a constant-current source.
- Grid-Forming (GFM) Models: Two variants were explored: a simplified model using an adjustable resistor that activates during overcurrent, and a complex model that adjusts voltage magnitude while maintaining balanced phase angles to respect both power and current limits.
Experimental Validation
The models were implemented within the PowerModels.jl ecosystem and tested on a synthetic 4-bus microgrid benchmark. Validation involved comparing simulation results against established experimental data regarding hardware responses to various fault types.
As shown in Figure 1, the models successfully capture the characteristic "elbow" in the current and voltage curves. As fault resistance decreases, the inverter hits its current limit, leading to a corresponding collapse in voltage and a plateau in output power. This alignment with physical hardware behavior confirms the utility of the continuous-variable approach.
Sequence Current Alignment
A key requirement for protection engineering is the accurate representation of sequence currents. The GFL model successfully replicated real-world observations where inverters primarily inject positive-sequence currents during faults.
Figure 2 demonstrates this behavior during a line-to-line fault. While phase currents diverge, the sequence current analysis shows that the positive sequence (Seq 1) remains the dominant component, validating the model's practical application for unbalanced fault analysis.
Discussion of Convergence Challenges
While the GFL model proved robust across all fault types, the GFM models encountered specific feasibility issues. The simplified GFM model struggled with line-to-line faults due to "stiffness"—the math did not allow for the significant phase angle jumps seen in real-world hardware. Meanwhile, the complex GFM model faced challenges during 3-phase faults where the transition between saturated and non-saturated states created mathematical regions where solvers could not converge on a valid solution.
Conclusion and Future Work
This research provides a foundational step toward integrating IBRs into large-scale grid protection optimization. By prioritizing mathematical smoothness, the authors have created models that are more compatible with modern solvers than traditional discrete representations. Future work will focus on merging GFM model strengths to eliminate current convergence weaknesses and developing relaxed linear convex approximations to enable even faster, massive-scale distribution network studies.
Summary updated