📊 statistics

Unbalanced Multi-Phase Distribution Grid Topology Estimation and Bus Phase Identification

This paper proposes an information-theoretic approach using smart meter data and symmetrical components to accurately estimate unbalanced multi-phase distribution grid topology and identify bus phase connections, even in the presence of incorrect phase labels and distributed energy resources.

Original authors: Yizheng Liao, Yang Weng, Guangyi Liu, Zhongyang Zhao, Chin-woo Tan, Ram Rajagopal

Published 2026-06-04

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yizheng Liao, Yang Weng, Guangyi Liu, Zhongyang Zhao, Chin-woo Tan, Ram Rajagopal

Original paper licensed under CC BY 4.0 (http://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

Imagine a city's power grid as a massive, complex tree. The trunk is the main substation, and the branches are the wires carrying electricity to your home. In a perfect world, we would have a perfect map of every branch and exactly which "side" of the tree (Phase A, B, or C) connects to your house.

But in reality, especially in older neighborhoods, we often don't have that map. The records are outdated, or someone made a mistake when writing them down. Furthermore, with the rise of rooftop solar panels and electric cars, the power flow is no longer steady and balanced; it's messy and unpredictable. This makes it very hard for utility companies to know what's happening in their grid, which is dangerous and inefficient.

This paper proposes a clever, data-driven way to redraw the map and fix the labels using only the information coming from smart meters. Here is how it works, broken down into simple concepts:

1. The Problem: A Messy Map and Wrong Labels

Think of the power grid like a giant puzzle.

The Missing Map: Utility companies often don't know exactly how the wires are connected in low-voltage areas (like neighborhoods).
The Wrong Labels: Even if they have a map, the labels might be wrong. For example, a house might be recorded as connected to "Phase A," but it's actually on "Phase B." In the US, about 10% of these labels are wrong or missing.
The Chaos: New energy sources (like solar) make the power flow unbalanced, like a seesaw that is constantly tipping to one side. Traditional methods for mapping the grid assume the seesaw is perfectly balanced, so they fail when things get messy.

2. The Solution: Listening to the "Whispers" of the Grid

Instead of trying to measure the grid directly (which requires expensive, high-tech equipment that most neighborhoods don't have), the authors use a "detective" approach based on smart meter data.

They treat the voltage readings from smart meters like a conversation between neighbors.

The Analogy: Imagine you are in a large house with many rooms. You don't know the floor plan, but you can hear people talking. If you hear two people talking loudly and clearly to each other, they are likely in the same room or right next door. If they are whispering or silent, they are far apart.
The Math: The paper uses a mathematical concept called Mutual Information. This measures how much "information" one meter shares with another. If two meters are connected by a wire, their voltage changes will be highly correlated (they "talk" to each other). If they are far apart, the connection is weak.

3. The Magic Trick: Symmetry in Chaos

The biggest hurdle was that the grid is "unbalanced" (messy). The authors found a way to turn this mess into order.

The Metaphor: Imagine a chaotic dance floor where everyone is moving randomly. The authors realized that if you look at the dance from a specific angle (using something called Symmetrical Components), the chaos organizes itself into three neat, balanced lines of dancers.
By translating the messy, unbalanced voltage data into these three neat "lines" (Positive, Negative, and Zero sequences), they could use a standard, efficient algorithm (called the Chow-Liu algorithm) to reconstruct the tree structure. It's like taking a tangled ball of yarn and finding a way to untangle it into three straight strands to see the pattern.

4. Fixing the Wrong Labels

Once the map of the tree is drawn, the authors tackle the wrong labels.

The Clue: They use a physics rule (Carson's equation) which acts like a "fingerprint" for the wires.
The Process: They check the voltage correlation between neighbors. If two houses are connected by a wire, their voltages should move in a very specific, synchronized way. If the records say House X is on Phase A and House Y is on Phase B, but their voltages move perfectly in sync as if they were on the same phase, the algorithm knows the label is wrong and fixes it.

5. The Results: A Super-accurate Map

The authors tested this method on real-world data from California and Europe, simulating everything from quiet neighborhoods to areas with heavy solar power usage.

Accuracy: The method was incredibly accurate, finding the correct map even when the data was noisy or the labels were wrong.
No Extra Hardware: It works using the data smart meters already send every hour. It doesn't need expensive new sensors.
Speed: It can figure out the map in seconds for small neighborhoods and a few minutes for large cities.

In Summary:
This paper teaches us how to rebuild a lost map of a power grid and correct its wrong labels just by listening to the "conversation" between smart meters. By using a mathematical trick to turn a messy, unbalanced system into a clean, balanced one, the algorithm can draw the correct tree structure and fix the labels, ensuring the grid is safe and efficient even with all the new solar panels and electric cars.

Technical Summary: Unbalanced Multi-Phase Distribution Grid Topology Estimation and Bus Phase Identification

Problem Statement
The integration of Distributed Energy Resources (DERs) such as rooftop solar and electric vehicles introduces significant uncertainty and unbalance into distribution grids. Effective monitoring and control of these resources require accurate knowledge of the grid topology and bus phase labeling. However, this information is often unavailable or unreliable in medium- and low-voltage distribution networks due to limited investment, outdated records, and human errors. Existing topology estimation methods predominantly focus on balanced or single-phase systems and often assume correct phase labels. Furthermore, many approaches rely on Phasor Measurement Units (PMUs), which are not widely deployed in distribution grids. The specific challenge addressed in this paper is the estimation of multi-phase grid topology and the identification of correct bus phase connections using only smart meter voltage data, in the presence of unbalanced loads, incorrect phase labels, and the absence of PMUs.

Methodology
The proposed approach is a data-driven method utilizing smart meter time-series voltage measurements. The methodology proceeds through the following stages:

Probabilistic Graphical Modeling: The distribution grid is modeled as a probabilistic graphical model where nodes represent bus voltage vectors and edges represent statistical dependencies. The problem is formulated as recovering the radial structure of this graph.
Symmetrical Component Transformation: To handle unbalanced multi-phase systems, the paper converts the three-phase unbalanced phasors into symmetrical components (positive, negative, and zero sequences). This transformation decouples the phases, allowing the system to be treated as a balanced multi-phase system within the sequence component frame.
Mutual Information Maximization (Chow-Liu Algorithm): The paper proves that under the assumption of independent current injection increments, the conditional independence relationships in a radial grid imply that the topology estimation problem is equivalent to finding a Maximum Weight Spanning Tree (MWST). The weights are determined by the mutual information between voltage increments of bus pairs. The Chow-Liu algorithm is employed to efficiently solve this optimization problem.
- Robustness to Phase Labels: A key theoretical finding is that the mutual information calculation is invariant to the permutation of phase labels. Therefore, the algorithm can estimate the correct topology even if the smart meter phase labels are incorrect or unknown.
- Voltage Magnitude Only: Recognizing that phase angles are often unavailable, the paper demonstrates that using only voltage magnitudes ( $|\Delta V|$ ) yields sufficient information for accurate topology estimation, as voltage magnitude changes contain the dominant statistical dependency.
Bus Phase Identification: Once the topology is estimated, the paper proposes a method to identify and correct bus phase labels. By utilizing Carson's equations, the method calculates the self and mutual impedances of lines. It leverages the fact that voltage correlations are highest between buses connected on the same phase. Starting from the substation (where labels are assumed reliable), the algorithm traverses the estimated topology to identify the true phase connection of each bus.

Key Contributions

Unbalanced System Handling: The paper extends topology estimation from balanced single-phase systems to unbalanced multi-phase systems by utilizing symmetrical component transformations.
Label Invariance: The proposed algorithm does not require prior knowledge of correct bus phase labels. It is mathematically proven to be invariant to incorrect phase labeling, addressing a critical practical issue where up to 10% of labels may be erroneous.
PMU-Free Estimation: The method functions effectively using only smart meter data (voltage magnitudes), eliminating the need for expensive PMU installations.
Phase Correction Mechanism: A novel data-driven approach is introduced to identify true bus phase connections using voltage correlations and Carson's equations, correcting the labeling errors after topology estimation.
Scalability: The computational complexity is analyzed, and a parallel processing strategy is suggested for large-scale systems (e.g., 8500-bus) by decomposing the grid into independent low-voltage sub-grids.

Results
The algorithm was validated using IEEE 37-bus, 123-bus, and 8500-bus test systems, simulated with real-world data sets from PG&E (California), the ADRES project (Austria), and Pecan Street (Texas).

Accuracy: Under noiseless conditions, the algorithm achieved 0% topology estimation error rate for both 37-bus and 123-bus systems, regardless of whether phase angles or only voltage magnitudes were used.
DER Integration: The method maintained high accuracy (0% error) even with 20% to 100% DER penetration levels, demonstrating robustness against the unbalance caused by high renewable integration.
Incorrect Labels: The algorithm maintained 0% error rates even when up to 20% of bus phase labels were incorrect, whereas modified single-phase methods showed significant performance degradation (up to 11.7% error).
Noise Sensitivity: With realistic measurement noise (up to 0.2%), the error rate remained low (approx. 3-4%), primarily affecting feeder connections, which can be corrected via post-processing.
Data Requirements: The algorithm requires significantly less data than single-phase counterparts, achieving perfect topology recovery with as little as 16–20 hours of measurements depending on the sampling frequency.

Significance
The paper claims that this work provides a critical solution for the "blind" operation of modern distribution grids. By enabling accurate topology estimation and phase identification without relying on expensive hardware (PMUs) or accurate historical records, the method facilitates the safe and efficient integration of DERs. The ability to operate under unbalanced conditions and incorrect labeling makes the approach particularly suitable for real-world low-voltage distribution networks where such imperfections are common. This capability is presented as a prerequisite for the advanced monitoring and control tools necessary to manage the increasing complexity of future distribution grids.

1. The Problem: A Messy Map and Wrong Labels

2. The Solution: Listening to the "Whispers" of the Grid

3. The Magic Trick: Symmetry in Chaos

4. Fixing the Wrong Labels

5. The Results: A Super-accurate Map

More like this