Stiffness-Aware Decentralized Dynamic State Estimation for Inverter-Dominated Power Systems
Xingyu Zhao, Marcos Netto, and Junbo Zhao · 2026
Read Stiffness-Aware Decentralized Dynamic State Estimation for Inverter-Dominated Power Systems by Xingyu Zhao, Marcos Netto, and Junbo Zhao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-18732.txt.
Introduction
Zhao, Netto, and Zhao propose a stiffness-aware decentralized method for dynamic state estimation in power grids run on inverters. These units show fast and slow dynamics through cascading controls, so system models are stiff and old methods need impractically small sampling steps to stay stable. Statistical linearization builds a linear stand-in for the nonlinear dynamics, and matrix-exponential discretization then carries uncertainty forward in closed form in time. The result is stable tracking at lower sampling rates, shown to be fast and true under coarse sampling.
Catalog record
| CREATOR | Xingyu Zhao, Marcos Netto, and Junbo Zhao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61512 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.18732
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Stiffness-Aware Decentralized Dynamic State Estimation for Inverter-Dominated Power Systems Xingyu Zhao, Marcos Netto, and Junbo Zhao Abstract Dynamic state estimation (DSE) is becoming increasingly important for monitoring inverter-dominated power systems. Due to their cascading control structures, inverter-based resources (IBRs) exhibit multi-timescale dynamics, leading to stiff system models that pose significant challenges for conventional DSE methods. In particular, explicit discretization schemes often require impractically small sampling intervals to maintain numerical stability, increasing computational and communication burdens.To address this issue, this paper proposes a stiffness-aware decentralized DSE method for inverter-dominated power systems. The statistical linearization is used to construct a local linear surrogate model for the nonlinear dynamics, which allows matrix-exponential discretization to enable analytical uncertainty propagation in discrete time, rather than relying on explicit integration schemes. This enables stable DSE at lower sampling rates. Numerical results reveal the mechanism by which stiff dynamics destabilize conventional DSE and demonstrate that the proposed method achieves efficient and accurate estimation under coarse sampling conditions. I Introduction The modern power system is undergoing a transformative shift from a synchronous-machine-dominated system to a power-electronics-inverter-dominated one. Functioning as the interface of distributed energy resources, inverters synchronize with the grid through digital controllers, making their dynamic states defined by their control. For system operators, real-time visibility into these dynamic states is of particular interest, as it enables better interpretation of system dynamics and faster response to abnormal conditions. For instance, several large solar farms were incorrectly tripped in California’s 2016 Blue Cut Fire event due to erroneous phase-locked loop (PLL) behaviors [ 1 ] . With DSE, system operators may identify early signs of PLL loss of synchronism and take proactive actions before instability develops. Compared with port measurements alone, hidden dynamic states can provide more informative evidence for assessing IBR operating modes, such as whether an IBR is subject to current limiting [ 2 ] , entering momentary cessation [ 3 ] , or successfully riding through a fault. Moreover, DSE can be integrated with the dissipating energy flow method to identify the root cause of resonance and oscillatory behavior [ 4 ] , which benefits system operators by enabling both real-time oscillation source location and post-fault analysis. Beyond providing state information, DSE also offers a statistical approach for anomaly detection through innovation and residual tests. This can help system operators detect abnormal conditions, including model parameter mismatches, cyberattacks [ 5 ] , and sensor malfunctions [ 6 ] . Finally, DSE can also be extended to online parameter estimation. By augmenting the state vector, parameters, such as virtual inertia and damping, can be jointly estimated and tracked in real time [ 7 ] . For a comprehensive review of DSE applications, we refer to [ 8 , 9 ] . Generally, DSE is formulated within a Bayesian filtering framework, which consists of two iterative steps: a prediction step that computes the prior state estimate, and a correction step that updates the posterior state estimate. However, unlike synchronous machines, IBRs introduce tightly coupled cascaded control loops operating across microsecond-to-second timescales. This results in severely stiff dynamic models, whose eigenvalue spectra can span several orders of magnitude. Such stiffness fundamentally challenges conventional DSE, which typically relies on explicit discretization schemes designed for moderately stiff synchronous-machine dynamics. As the prediction step requires an explicit discrete-time state-transition function, the stiffness of the dynamic model poses a major challenge to maintaining numerical stability in current DSE methods. This paper aims to address this problem. I-A Literature Review I-A 1 Discretization Explicit methods, such as the forward Euler and Runge–Kutta methods have been adopted in DSE [ 10 , 11 ] , although their numerical stability is guaranteed only under limited conditions [ 12 ] . When applied to stiff equations, explicit methods require impractically small time steps to keep the error in the result bounded. For DSE, using tiny time steps may be constrained by either computational or communication costs or by the sampling rate of measurement systems. To address this issue, implicit methods such as backward Euler [ 13 ] and trapezoidal [ 14 ] can be leveraged. However, implicit methods involve nested calculations, which are more computationally demanding and lack direct control over computational burden. For example, when applying implicit methods to an unscented Kalman filter (UKF) or a particle filter (PF), a set of equations needs to be solved numerically multiple times at each step size. Moreover, solving such equations may require evaluating the Jacobian of the dynamic model, which is difficult to obtain for IBRs with proprietary or highly complex control structures. In [ 15 ] , a discrete-time model with numerical stability guarantee is derived for grid-forming inverters (GFMs). However, this is still far from a generic solution, since it relies heavily on the assumption of time-scale separation between control loops and the linear structure of each subsystem, which are satisfied by GFMs only. From the modeling perspective, a commonly used practice is to combine model order reduction with explicit methods. However, the use of reduced-order models in DSE can introduce significant bias at the onset of transients [ 16 ] . Moreover, the reduced-order model may be truncated to different dimensions under different setpoints, further complicating DSE design. To date, a numerically stable and computationally efficient DSE method for the stiff IBR models is still lacking. I-A 2 DSE for IBRs Existing research regarding DSE for IBRs mainly adopts the decentralized scheme [ 17 ] . In decentralized DSE schemes, port voltage measurements are typically treated as system inputs for driving the component dynamics, while port current measurements serve as output measurements for state correction. Early-stage studies on decentralized DSE for IBRs focus on those that operate in grid-following mode, such as solar photovoltaic systems [ 18 ] and wind generators [ 19 , 20 , 21 , 22 ] . Recently, DSE has also been extended to IBRs operating in grid-forming mode [ 16 , 22 , 23 ] , as well as to hybrid IBRs that switch between grid-forming and grid-following controls [ 24 , 25 ] . However, existing studies either use reduced models that overlook fast transients caused by inner-loop controllers and PLLs ( [ 16 , 19 , 20 , 21 , 23 ] ), or discretize the IBR model through explicit methods with extremely small time steps ( [ 18 , 22 , 24 ] ). The challenges associated with small time steps have been identified in [ 20 ] and addressed through interpolation and step-size control. However, the method in [ 20 ] still relies on the explicit Runge–Kutta scheme, whose limitations have already been discussed. To date, enabling decentralized DSE with detailed IBR models under relaxed sampling rates remains an open problem. I-B Contributions Although extensive efforts have been devoted to extending DSE formulation to handle the nonlinearity and uncertainty of power system dynamics, the challenge of numerical stability in DSE has remained largely overlooked, despite the inherent stiffness of power system models. The major contributions of this paper are summarized as follows: • This paper identifies stiffness-induced numerical instability as a critical yet largely overlooked issue in decentralized DSE. It shows that, when detailed IBR models with multi-timescale dynamics are used for DSE, a conventional prediction