Source-linked AI summary
High-Fidelity Model Order Reduction for Microgrids Stability Assessment
Petr Vorobev, Po-Hsu Huang, Mohamed Al Hosani, James L. Kirtley, Konstantin Turitsyn
TL;DR
Inverter-based microgrid stability requires models that capture network effects without the computational cost and limited interpretability of detailed models. The paper develops a high-fidelity reduced-order model, showing that stability depends on microgrid-specific inverter and network properties and validating the approach against detailed analyses.
Problem
Existing detailed microgrid models are computationally expensive, while simplified models may miss fast network effects and mischaracterize stability relative to large-scale power systems.
Method
The paper systematically separates fast and slow degrees of freedom while retaining fast-state effects, and generalizes the reduction to arbitrary microgrid networks.
Results
Stability limits are determined by the ratio of inverter rating to network capacity, with shorter lines producing a smaller stability region.
Takeaways & Limitations
Microgrid stability has network-specific behavior: strengthening the connection can reduce stability, unlike the usual transmission-grid expectation.
Abstract
from arXiv · showhide
Proper modeling of inverter-based microgrids is crucial for accurate assessment of stability boundaries. It has been recently realized that the stability conditions for such microgrids are significantly different from those known for large- scale power systems. While detailed models are available, they are both computationally expensive and can not provide the insight into the instability mechanisms and factors. In this paper, a computationally efficient and accurate reduced-order model is proposed for modeling the inverter-based microgrids. The main factors affecting microgrid stability are analyzed using the developed reduced-order model and are shown to be unique for the microgrid-based network, which has no direct analogy to large-scale power systems. Particularly, it has been discovered that the stability limits for the conventional droop-based system (omega - P/V - Q) are determined by the ratio of inverter rating to network capacity, leading to a smaller stability region for microgrids with shorter lines. The theoretical derivation has been provided to verify the above investigation based on both the simplified and generalized network configurations. More impor- tantly, the proposed reduced-order model not only maintains the modeling accuracy but also enhances the computation efficiency. Finally, the results are verified with the detailed model via both frequency and time domain analyses.
I. INTRODUCTION
Microgrid stability modeling must account for network dynamics and distinguish microgrids from large-scale power systems. The paper develops a systematic reduced-order approach that separates fast and slow dynamics while preserving accuracy and interpretability.
- Motivation: High R/X ratios and network dynamics can make microgrid stability qualitatively different from transmission-grid stability.Previous studies found that conventional large-scale power-system principles may not transfer directly to microgrids.
- Motivation: Full-order models capture inverter and network states but are computationally expensive and may obscure instability mechanisms.The paper identifies model reduction as a balance between stability-prediction accuracy, computational simplicity, and transparency.
- Motivation: Fast network dynamics can influence slow inverter-controller modes, so timescale ratio alone is insufficient for deciding which states to eliminate.The paper therefore targets separation of fast and slow degrees of freedom without significant loss of accuracy.
- Contributions: The paper develops a reliable reduced-order model for fast and accurate microgrid stability studies and dynamic simulations.It also quantifies fast-state effects, explains quasi-stationary-model inadequacy, and generalizes the method to arbitrary network structures.
II. TWO-BUS MODEL
The two-bus model illustrates how inverter, line, and control dynamics combine in microgrid stability analysis. Although electromagnetic dynamics are faster than droop-control dynamics, the paper argues they should not be discarded through a purely quasi-stationary approximation.
- System model: The motivating system is a single droop-controlled inverter connected to an infinite bus through coupling and line impedance.The inverter uses frequency-active-power and voltage-reactive-power droop relations.
- System model: Complex voltage and current amplitudes may vary arbitrarily in time, so their representation is a mathematical change of variables rather than an approximation.The equilibrium frequency and variables are used to formulate the dynamic model.
- Model reduction: The reduced model retains terminal-voltage and frequency dynamics while eliminating selected fast internal inverter states.Power-controller modes are treated as the main stability-relevant modes, while internal fast states may be omitted.
- Model reduction: Although L/R ≈3.1ms is below the 20ms base cycle and τ ≈31.8ms droop-control timescale, neglecting electromagnetic derivatives can give inappropriate stability conclusions.The paper uses this two-bus setting to investigate the effect of fast electromagnetic transients on stability.
A. Conventional 3rd-Order Model
The conventional 3rd-order model treats line currents algebraically by neglecting electromagnetic derivative terms, then analyzes angle–voltage dynamics under small-signal assumptions. Its delay-based approximation shows that conductance and associated delays can destabilize the system, although part of the derivation is not fully rigorous.
- The conventional approximation sets the derivatives of line-current dynamics to zero, making currents algebraic and solvable from the equilibrium network equations.
- The linearized model uses angle deviation δθ and normalized voltage deviation δρ under small-angle, near-nominal-voltage assumptions.The assumed angle and relative-voltage deviations are typically of order ∼10^-2.
- Without conductance, angle and voltage deviations decouple and the system is always stable; conductance introduces feedback that may cause instability.
- The voltage deviation follows angle deviation with a delay, which can be approximated using a first-order Taylor expansion when angle dynamics are sufficiently slow.
- High conductance can make the effective damping coefficient negative, while the approximate stability condition is mp < (1 + nqB)^2.The delay effect depends on gain as well as timescale ratio, so arbitrary timescale separation does not guarantee stability.
- The delay-based derivation is not entirely rigorous because angle dynamics need not be slower than voltage dynamics, although its stability condition is reasonably accurate.
B. High-Fidelity 3rd-Order Model
The high-fidelity 3rd-order model retains electromagnetic-transient effects through derivative corrections to the conventional dynamics. It explains stability boundaries in terms of inverter rating, line strength, and control filtering, and agrees more closely with the full model than the conventional reduction.
- When the R/X ratio increases, electromagnetic transients become critical to instability and the conventional 3rd-order model becomes inappropriate.
- The reduced current dynamics are obtained by Taylor-expanding the frequency-domain current relation for modes slower than the electromagnetic time L/R.
- The resulting model adds derivative terms involving G′ and B′ to both angle and voltage equations, allowing electromagnetic effects to alter effective damping.
- Decreasing inverter rating or strengthening the grid through lower line reactance and resistance reduces stability, unlike the corresponding trend in transmission grids.
- The conventional 3rd-order model predicts a larger stability region than both the full 5th-order and proposed high-fidelity 3rd-order models.
- For a 10 kVA inverter, the simulated stability boundaries are kp ∼0.5 −2% and kq ∼2 −25%, depending on line length and filter time constant.
- Increasing connection-line impedance generally enlarges the stability region, especially for voltage droop, while smaller inverters can require kp below 0.5%.The text notes no strict monotonic dependence of maximum frequency droop on line length.
III. GENERALIZED MULTI-TIMESCALE APPROACH
The paper develops a first-order singular-perturbation approach that separates slow and fast variables while retaining fast-variable effects on slow dynamics. The formulation supports arbitrary fast-variable sets, including instantaneous algebraic constraints.
- Generalized multi-timescale approach: The method retains the influence of fast-variable dynamics on slow modes instead of eliminating fast variables altogether.It is presented as a first-order singular-perturbation formulation, whereas neglecting fast dynamics corresponds to a zero-order approximation.
- Generalized multi-timescale approach: The system is partitioned into slow and fast subsystems using a Jacobian-based state-space formulation.The state vector is divided into slow and fast degrees of freedom before deriving the reduced equations.
- Generalized multi-timescale approach: The proposed reduction expresses fast-state perturbations using both slow-state perturbations and the first derivative of slow-state perturbations.This extends the zero-order relation obtained by neglecting the fast-subsystem derivative.
- Generalized multi-timescale approach: The formulation accommodates arbitrary fast-variable sets, including fast degrees of freedom represented by algebraic constraints.The representation avoids separately identifying linearly independent variables or solving individual variable derivatives.
IV. NETWORK GENERALIZATION AND STABILITY CERTIFICATES
The network generalization builds a low-order dynamic model from the network admittance matrix and its first-order expansion. The resulting equations retain network and load dynamics and support stability analysis for interconnected inverters.
- Network generalization: The network model constructs a dynamic equation for each inverter using a full impedance-based admittance matrix.Line and load impedances are written in the Laplace domain before forming the network admittance matrix.
- Network generalization: The total admittance is separated into network and diagonal load components before applying a first-order Taylor expansion.The expansion produces zeroth- and first-order admittance terms used to represent voltage and voltage-derivative effects.
- Network generalization: The generalized reduced equations incorporate network and load dynamics through the first-order admittance terms.The resulting equations couple inverter angles and relative voltages with their derivatives.
- Network generalization: Using only the quasi-stationary admittance matrix is inappropriate for network dynamic simulation.The proper network representation uses the initial structure with full impedances, including the Laplace parameter s.
- Stability certificates: The reduced network model enables stability analysis of multi-inverter systems while preserving a low-order representation of droop coefficients.Local Lyapunov-based criteria can be derived, but the resulting conditions may be conservative and require an appropriate Lyapunov function.
A. Model Accuracy
The paper evaluates the proposed reduced-order model against a full model using time-domain responses and eigenvalue movements in a five-inverter cascade microgrid. The proposed third-order model more closely matches the full model than the conventional third-order model, with accuracy depending on network X/R ratio.
- Model Accuracy: The evaluation compares a five-inverter cascade microgrid using time-domain responses and eigenvalue movements across different models.The active-power droop gain is selected to destabilize the system so erroneous predictions can be observed.
- Model Accuracy: The proposed third-order model produces eigenvalues far closer to the full model than the simple third-order model.This agrees with the simplified two-bus comparison described earlier in the paper.
- Model Accuracy: Lower X/R ratios generally improve the proposed model’s prediction accuracy.The model’s accuracy relies on fast relaxation of electromagnetic dynamics, and instability is mainly associated with low X/R ratios disrupting the P−ω and Q−V relations.
B. Simulation Efficiency
The proposed reduced-order model substantially lowers the state dimension and improves time-domain simulation efficiency for inverter-based microgrids.
- The full model uses approximately 9 states per inverter, whereas the proposed model uses only 3 states per inverter.This reduces the number of states by two-thirds.
- The reduced state dimension enables simulation of microgrid networks containing large numbers of inverters.
- Tests on 5- and 25-inverter microgrids showed significantly improved simulation efficiency with the proposed model.Both models were evaluated using Matlab default O.D.E. solvers over one second.
VI. CONCLUSION
The conclusion identifies network dynamics as important to inverter-controller behavior and presents a reduced-order model that preserves accuracy while simplifying computation. It also highlights microgrid-specific stability deterioration as network impedances or inverter ratings are reduced.
- Network dynamics can strongly influence the slow dynamics associated with inverter power controllers.
- The proposed method excludes fast network degrees of freedom without compromising model accuracy while reducing computational complexity.
- Compared with a quasi-stationary approximation, the resulting third-order model changes stability predictions by modifying its coefficients.
- Reducing network impedances or inverter ratings deteriorates stability in the analyzed microgrid-specific effects.