Synchronized State Approximation

• Synchronized state approximation is a framework that represents high-dimensional or partially observed dynamics by projecting them onto synchronized lower-dimensional states, such as common oscillatory trajectories or steady spatial profiles.
• It leverages reduction techniques—including phase response analysis, bifurcation studies, and stochastic closures—to simplify nonlinear dynamics and capture transitions like the shift from synchronized to super‐synchronized regimes.
• The approach underpins practical applications, notably in power systems, where it enables full state estimation from synchronized measurement snapshots using Bayesian methods and machine learning models.

Synchronized state approximation denotes a family of reductions, estimators, and control-oriented constructions in which a high-dimensional or partially observed system is replaced by, projected onto, or inferred from a synchronized description. In the cited literature, the synchronized object may be a common oscillatory trajectory, a proportional steady spatial profile, a bounded-dispersion mean-field orbit, an approximate-synchronization state set in a logical network, or a full electrical-network state reconstructed from time-aligned phasor measurements (García-Selfa et al., 2021, Huang et al., 2020, Oukil, 2017, Zhao et al., 2022, Azimian et al., 2020). The unifying idea is that synchronization is treated not only as a qualitative phenomenon but as a mathematically or computationally exploitable representation of state.

1. Scope of the concept and principal definitions

Across these works, synchronized state approximation is not tied to a single formal definition. In reaction-diffusion predator-prey dynamics, the synchronized positive steady state is one in which the two components are proportional, $u=p\theta$ and $v=q\theta$, with the specific state

$$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$

so that the ratio $u_a/v_a=(1-b)/(1+c)$ is independent of $x$; in that setting, a synchronized state approximation means approximating a nearby positive solution by $(u,v)\approx (p\theta_a,q\theta_a)$ (Huang et al., 2020). In abstract mean field models, synchronization is defined by two requirements: the oscillators are dynamical, and the pairwise distances $|x_i(t)-x_j(t)|$ remain uniformly bounded in time (Oukil, 2017). In coupled multi-valued logical networks, approximate synchronization is defined by the eventual bound

$$\max_{1\le i\le n}|x_i(t;X_0,Z_0)-z_i(t;X_0,Z_0)|\le \frac{1}{k-1},$$

which leads to the approximate synchronization state set (ASSS) $\Lambda$, the maximum approximate synchronization basin (MASB), and the shortest approximate synchronization time (SAST) (Zhao et al., 2022).

In symbolic and discrete-event settings, the same general theme appears as uncertainty collapse. For $\epsilon$-machines, exact synchronization means that there exists a word $v=q\theta$0 such that $v=q\theta$1, while asymptotic synchronization means that observer uncertainty goes to $v=q\theta$2 exponentially fast; the relevant quantitative objects are the synchronization rate constant $v=q\theta$3 and the prediction rate constant $v=q\theta$4 (Berlinkov, 2014). For bounded synchronized Petri nets, a synchronizing sequence is an input word that drives the system from an unknown current marking in a reachable uncertainty set to a known target marking (Pocci et al., 2013). This suggests that “synchronized state approximation” is best read as an umbrella term for representing system state by a synchronized manifold, synchronized profile, synchronized estimator, or synchronized target set rather than as a single domain-specific formalism.

Domain	Synchronized object	Approximation criterion
Reaction-diffusion predator-prey	$v=q\theta$5 proportional to $v=q\theta$6	$v=q\theta$7
Abstract mean field models	bounded-dispersion collective orbit	$v=q\theta$8
Multi-valued logical networks	ASSS $v=q\theta$9	$$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$0
Time-synchronized power SE	full state $$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$1 from synchronized $$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$2	$$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$3

2. Low-dimensional and stochastic reductions of synchronized dynamics

A central line of work treats synchronized state approximation as a reduction of nonlinear collective dynamics. In chemically coupled Belousov-Zhabotinsky oscillators embedded in an active oscillating medium, the bead-plus-medium model is first reduced to a five-dimensional synchronized-manifold system by imposing $$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$4, $$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$5, $$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$6 for all beads. Bifurcation and continuation analysis then identifies three principal regimes: oscillation death, synchronization, and a distinct high-amplitude, low-frequency super-synchronized or mobbing state. The transition between synchronized and super-synchronized motion is organized by a generalized Hopf point $$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$7, Hopf branches $$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$8 and $$(u_a,v_a)=\left(\frac{1-b}{1+bc}\theta_a,\ \frac{1+c}{1+bc}\theta_a\right),$$9, and a saddle-node of periodic orbits curve $u_a/v_a=(1-b)/(1+c)$0; the paper emphasizes that the abrupt change in amplitude and period is a canard-like explosion rather than a smooth deformation of one periodic orbit into another (García-Selfa et al., 2021).

The same work derives a phase approximation model from the weak-coupling reduction

$u_a/v_a=(1-b)/(1+c)$1

with interaction functions obtained by averaging over the limit cycle using the phase response curve. In the synchronized state, phase locking occurs at $u_a/v_a=(1-b)/(1+c)$2, and the collective period is approximated by

$u_a/v_a=(1-b)/(1+c)$3

The phase model reproduces the period well across parameter variation and captures the discontinuity in period at the synchronized-to-super-synchronized transition. In the synchronized regime, $u_a/v_a=(1-b)/(1+c)$4 is dominated by first Fourier modes, whereas in the super-synchronized regime higher harmonics up to third order become important, reflecting a change in the underlying limit cycle rather than a mere shift in phase locking (García-Selfa et al., 2021).

A different approximation problem arises in the finite-size Kuramoto-Sakaguchi model. There, the synchronized cluster $u_a/v_a=(1-b)/(1+c)$5 is continuously forced by non-entrained rogue oscillators $u_a/v_a=(1-b)/(1+c)$6, and the order parameter $u_a/v_a=(1-b)/(1+c)$7 does not settle to a constant but exhibits persistent fluctuations. The cumulative rogue forcing is expressed through collective terms $u_a/v_a=(1-b)/(1+c)$8 and $u_a/v_a=(1-b)/(1+c)$9, and numerical results suggest that the centered fluctuations are approximately Gaussian with variance scaling like $x$0. This motivates a stochastic closure in which the collective rogue forcing is approximated by a two-dimensional Ornstein-Uhlenbeck process, leading to a closed stochastic evolution equation for the synchronized oscillators and the order parameter. The reduced model reproduces the distribution of $x$1, the time evolution of $x$2, and the mean and variance of entrained phases, while remaining less accurate near the cluster edge or near the synchronization threshold (Yue et al., 2023).

Hydrodynamic active-matter theory provides a third variant. In oriented active fluids with internal duty-cycle phase, synchronization is represented by Fourier modes $x$3, with $x$4 serving as a complex Kuramoto order parameter, or by a coarse-grained phase field $x$5. The synchronized state is a phase-coherent hydrodynamic state in which the active stress becomes time-periodic, and the continuum approximation

$x$6

describes the phase-locked regime. The paper argues that this state spontaneously breaks time-translation invariance and is generically unstable to flow-pattern formation distinct from the instabilities of phase-incoherent active matter (Fürthauer et al., 2013).

3. Stability theory, invariant manifolds, and synchronized geometry

In several papers, synchronized state approximation is justified by explicit stability analysis. For the reaction-diffusion Lotka-Volterra predator-prey model with Dirichlet boundary condition, equal diffusion rates, and equal growth rates, the synchronized steady state

$x$7

exists when $x$8, $x$9, and $(u,v)\approx (p\theta_a,q\theta_a)$0, and it is locally asymptotically stable. The linearization about $(u,v)\approx (p\theta_a,q\theta_a)$1 has spectrum

$(u,v)\approx (p\theta_a,q\theta_a)$2

with all eigenvalues positive, so perturbations decay exponentially at the linear level. In this setting, synchronized state approximation is therefore a local asymptotic statement: nearby solutions retain the same spatial shape $(u,v)\approx (p\theta_a,q\theta_a)$3 up to fixed component-wise scaling, but the paper does not prove global stability in general (Huang et al., 2020).

Abstract mean field models give a more geometric formulation. For systems

$(u,v)\approx (p\theta_a,q\theta_a)$4

and their perturbations $(u,v)\approx (p\theta_a,q\theta_a)$5, the synchronization hypothesis

$(u,v)\approx (p\theta_a,q\theta_a)$6

yields a positively invariant synchronized region and, in the periodic setting, a periodic locked solution of the form

$(u,v)\approx (p\theta_a,q\theta_a)$7

A later development proves that, under $(u,v)\approx (p\theta_a,q\theta_a)$8 and $(u,v)\approx (p\theta_a,q\theta_a)$9, the linearized dynamics admit a decomposition into a one-dimensional non-decaying synchronized direction and an $|x_i(t)-x_j(t)|$0-dimensional exponentially stable complement. For each $|x_i(t)-x_j(t)|$1 in a synchronized neighborhood there exists a codimension-one submanifold $|x_i(t)-x_j(t)|$2 such that

$|x_i(t)-x_j(t)|$3

for $|x_i(t)-x_j(t)|$4, and the nonlinear dynamics approach a stable limit cycle in the form

$|x_i(t)-x_j(t)|$5

Here synchronized state approximation is literally a decomposition into a collective orbit plus an exponentially decaying transverse remainder (Oukil, 2017, Oukil, 2024).

For coupled map lattices, synchronized geometry can also be reproduced by altered temporal updating. Complete synchronization is characterized by $|x_i(t)-x_j(t)|$6, and synchronous updating admits chaotic complete synchronization only in a finite region of the $|x_i(t)-x_j(t)|$7 plane. Under asynchronous random sequential updating, the complete-synchronization domain is enlarged and the synchronized attractor becomes the fixed point $|x_i(t)-x_j(t)|$8. The paper then shows that pseudo-continuous delayed dynamics with interpolation

$|x_i(t)-x_j(t)|$9

reproduces exactly the complete-synchronization domain of the asynchronous system when $$\max_{1\le i\le n}|x_i(t;X_0,Z_0)-z_i(t;X_0,Z_0)|\le \frac{1}{k-1},$$0. This is an explicit equivalence result: an adequate delayed dynamics yields the same synchronized states and the same stability frontier as asynchronous updating, with lower threshold $$\max_{1\le i\le n}|x_i(t;X_0,Z_0)-z_i(t;X_0,Z_0)|\le \frac{1}{k-1},$$1 (González-Avella et al., 2015).

4. Time-synchronized state estimation in power systems

In power-system literature, synchronized state approximation has a distinct operational meaning: reconstructing the full electrical state from synchronized PMU or uPMU snapshots. The basic Bayesian target is the MMSE estimator

$$\max_{1\le i\le n}|x_i(t;X_0,Z_0)-z_i(t;X_0,Z_0)|\le \frac{1}{k-1},$$2

where $$\max_{1\le i\le n}|x_i(t;X_0,Z_0)-z_i(t;X_0,Z_0)|\le \frac{1}{k-1},$$3 is the synchronized measurement vector and $$\max_{1\le i\le n}|x_i(t;X_0,Z_0)-z_i(t;X_0,Z_0)|\le \frac{1}{k-1},$$4 is the full system state. In incompletely observed distribution systems, this motivates deep neural networks that learn the direct mapping $$\max_{1\le i\le n}|x_i(t;X_0,Z_0)-z_i(t;X_0,Z_0)|\le \frac{1}{k-1},$$5 offline from historical data and Monte Carlo power-flow simulations, thereby bypassing the explicit observability constraints of classical least-squares estimators (Azimian et al., 2020, Azimian et al., 2023). A related transmission-system framework, DeNSE, uses SCADA data only offline through kernel density estimation and Monte Carlo scenario generation, while using PMU data as the synchronized online input; this is designed to achieve sub-second situational awareness without direct online SCADA-PMU fusion (Varghese et al., 2022). A