⚡ electrical engineering

Empirical Representation of Transformer Branch Impedance Frequency Dependency

This paper proposes an empirical method to approximate the frequency-dependent branch impedance matrix of transformers using a representative eigenvalue, thereby avoiding the computationally expensive FEM calculations required for high-frequency transient studies while offering greater accuracy than classical damping factor approaches.

Original authors: B. Gustavsen, L. Oliveira, A. Portillo, F. Portillo

Published 2026-07-20

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: B. Gustavsen, L. Oliveira, A. Portillo, F. Portillo

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Invisible Dance of Electricity

Imagine the electrical grid as a massive, humming city of energy, where electricity flows like water through pipes to power our lights, phones, and dreams. But sometimes, nature throws a curveball: a lightning strike or a sudden switch flip sends a violent shockwave racing through these wires. To keep the city safe, engineers need to predict exactly how these shockwaves will behave. This is the world of electromagnetic transients—the study of those split-second, high-speed electrical storms.

To predict these storms, scientists build digital twins of the equipment, like giant transformers (the heavy-duty boxes that change voltage levels). These twins are called models. Some models are "black boxes," where you just plug in numbers and get an answer without knowing the inside. But for new transformers that haven't been built yet, or for understanding dangerous voltages inside the machine, engineers need "white box" models. These are like transparent blueprints where every coil and wire is mapped out. The tricky part? Electricity doesn't behave the same way at all speeds. When a lightning strike hits, the electricity moves so fast that the transformer's resistance and inductance change, acting like a shape-shifter. Calculating these changes for every single wire in a complex machine is like trying to count every grain of sand on a beach while the tide is coming in—it takes forever and requires supercomputers. This paper tackles the problem of how to make these digital twins accurate enough to be safe, but fast enough to actually use.

The Shortcut That Keeps the Rhythm

The authors of this paper, a team of researchers from Norway, Brazil, and Uruguay, are trying to solve a very specific headache: how to model the "branch impedance" (a fancy way of saying how much the transformer fights the flow of electricity) without spending hours or days on calculations.

In the past, engineers had two main choices. The first was the "Gold Standard": using a powerful computer simulation called Finite Element Method (FEM) to calculate the resistance of every single wire at dozens of different speeds. It was incredibly accurate, but it was also a time-sink. The authors note that for one specific transformer, this method took about 37 minutes just to gather the data for 14 different frequency samples. That's fine for a lab, but too slow for real-world planning.

The second choice was a faster, older trick called the "damping factor method." This approach was like using a generic, one-size-fits-all rule of thumb. It was super fast, but it often got the details wrong, specifically the "rhythm" or frequency of the electrical vibrations. It was like trying to play a song on a piano by guessing the notes; you might get the general tune, but the specific melody would sound off.

The Paper's Big Idea
The authors propose a clever middle ground, a "smart shortcut." Instead of calculating the resistance for every wire from scratch, they suggest borrowing a "musical fingerprint" from a representative transformer.

Here is how the analogy works: Imagine you have a complex orchestra (the new transformer you are studying). You want to know how it will sound when a lightning bolt hits it. Instead of testing every single instrument in the orchestra (which takes forever), you take a recording of the "smallest, quietest note" (a specific mathematical value called an eigenvalue) from a similar, well-studied orchestra. You then use that recording as a template. You take the physical blueprint of your new orchestra (its inductance matrix, calculated just once at high speed) and apply that borrowed "musical fingerprint" to it.

What They Found
When the team tested this method on a three-phase regulating transformer, the results were promising. They compared their "smart shortcut" against the slow "Gold Standard" and the fast "generic rule."

The Result: Their new method was significantly more accurate than the old "damping factor" trick. Specifically, it got the oscillation frequencies right. In the simulations, the old method made the electrical waves vibrate too fast, like a drum being hit too hard. The new method kept the rhythm correct, matching the slow, detailed Gold Standard almost perfectly.
The Catch: The paper explicitly rules out using a "low-frequency" blueprint for this shortcut. When they tried calculating the transformer's physical blueprint using slow electricity (60 Hz) instead of fast electricity, the results were full of errors. The "smart shortcut" only works if you build the physical foundation using high-frequency data.
The Speed Boost: The most exciting part is the time saved. By using this method, the calculation time dropped from 37 minutes to less than 3 minutes. That's a massive leap forward.

How Sure Are They?
The authors are careful not to claim they have solved the problem for every transformer in the world. They tested this on one specific type of transformer (a three-phase regulating transformer) and compared it to a similar one they borrowed data from. They admit that while the initial test was successful, the method needs to be tried on many more different types of transformers to be sure it works everywhere. They also noted that the two transformers they used had some differences in their internal wiring, yet the method still worked well, which is a good sign.

In short, this paper suggests a way to make electrical simulations both fast and accurate. It's like giving an engineer a cheat sheet that lets them skip the boring, slow math while still getting the right answer for how a transformer will survive a lightning storm. It's not a magic wand that fixes everything, but it's a very strong tool that saves a lot of time and keeps the electrical grid safer.

Technical Summary: Empirical Representation of Transformer Branch Impedance Frequency Dependency

Problem Statement
High-frequency transformer models are essential for simulating transient overvoltages and calculating internal winding voltages. While "black-box" models can be derived from measurements, "white-box" models are required for transformers that have not yet been built or for detailed internal voltage analysis. These white-box models typically rely on lumped parameter representations involving shunt/capacitance matrices ( $C$ ) and branch impedance matrices ( $Z_b$ ).

The primary challenge identified is the computational cost of calculating the frequency-dependent branch impedance matrix $Z_b$ using the Finite Element Method (FEM). Because $Z_b$ can be large (e.g., $159 \times 159$ in the study) and requires calculation at multiple frequency samples (e.g., 5 samples per decade), the column-wise FEM calculation becomes CPU-intensive. For instance, calculating $Z_b$ at 14 frequency samples for a 159-branch system can take approximately 37 minutes.

Existing alternatives, such as Fergestad's damping factor method, utilize a constant branch inductance matrix ( $L_0$ ) and introduce frequency dependency via an empirical damping factor applied to state-space eigenvalues. While computationally efficient, this approach sacrifices accuracy, particularly regarding the reproduction of oscillation frequencies in transient simulations.

Methodology
The authors propose a new empirical method to approximate the frequency dependency of $Z_b$ without performing full FEM calculations at every frequency sample. The core of the method involves:

Selection of an Empirical Eigenvalue: A representative eigenvalue, $\lambda_{emp}(\omega)$ , is extracted from the $Z_b(\omega)$ of a reference transformer (specifically, the smallest eigenvalue of a 1-phase WEG transformer studied by CIGRE JWG A2/C4.52). This eigenvalue is treated as an empirical quantity representing the frequency behavior of the system.
High-Frequency Inductance Matrix: For the target transformer, the high-frequency inductance matrix, $L_0$ , is calculated at a single high-frequency sample using FEM (via $L_0 = \text{Im}\{Z_b(\omega_0)\}/\omega_0$ ).
Eigenvalue Decomposition: The matrix $L_0$ is decomposed into eigenvectors ( $T$ ) and eigenvalues ( $\Lambda_L$ ).
Reconstruction of $Z_b(\omega)$ : The frequency dependency of the target transformer's $Z_b$ $Z_{b}$ is reconstructed by assuming the empirical eigenvalue $\lambda_{emp}(\omega)$ $λ_{e m p} (ω)$ applies to all eigenvalues of the target system, scaled by factors $k_m$ $k_{m}$ . These scaling factors are determined such that the imaginary part of the scaled empirical eigenvalue matches the high-frequency inductance eigenvalues at the reference frequency $\omega_0$ $ω_{0}$ .
- The reconstructed matrix is formed as $Z_b(\omega) \approx T \Lambda_Z(\omega) T^{-1}$ , where $\Lambda_Z$ contains the scaled empirical eigenvalues.
Rational Approximation: The resulting frequency-dependent matrix is approximated using a passive rational function (via Vector Fitting) to facilitate state-space modeling.

This approach is compared against two other models:

Model #1 (Reference): A full FEM calculation of $Z_b$ at all frequencies, fitted with a 7th-order passive rational model.
Model #3 (Baseline): The classical Fergestad damping factor method, which modifies the real part of state matrix eigenvalues using an empirical damping function derived from 25 transformers.

Key Results
The proposed method was tested on a three-phase regulating transformer (modeled as a single phase) with a $159 \times 159$ branch impedance matrix. Simulations involved applying a 1.2/50 $\mu$ s lightning impulse to High Voltage (H1) and Low Voltage (X1) terminals.

Accuracy: Both the proposed method (Model #2) and the damping factor method (Model #3) provided accurate results for the first few microseconds of the transient response. However, over longer durations (up to 30 $\mu$ s), the proposed method demonstrated superior accuracy. Specifically, the damping factor method resulted in oscillation frequencies that were too high, whereas the proposed method successfully reproduced the oscillation frequencies of the reference model.
Damping: The damping characteristics of the proposed method were found to be similar to the damping factor method.
Frequency Dependency of $L_0$ : The study highlighted that using a low-frequency (60 Hz) inductance matrix $L_0$ resulted in substantial errors for both simplified methods. The high-frequency $L_0$ is essential for the validity of the approach.
Computational Efficiency: The proposed method reduced the calculation time from approximately 37 minutes (full FEM at 14 frequencies) to less than 3 minutes (single FEM calculation for $L_0$ ).

Significance and Claims
The paper claims that the proposed empirical method offers a significant improvement in accuracy over the classical damping factor method, primarily by better capturing the frequency-dependent oscillation frequencies of the transformer.

The significance of the work lies in its ability to bypass the lengthy FEM calculations required for multi-frequency $Z_b$ extraction while maintaining high fidelity in transient simulations. By leveraging a single empirical eigenvalue from a representative transformer and combining it with the high-frequency inductance matrix of the specific transformer under study, the method provides a computationally efficient alternative for network studies involving electromagnetic transients.

The authors note that while the initial testing was successful, the method's generality requires further validation across a broader range of transformer designs, particularly regarding how similar the reference transformer must be to the target transformer in terms of winding sequence, conductor type, and geometry.

The Invisible Dance of Electricity

The Shortcut That Keeps the Rhythm

More like this