Data-Driven Approach for Distribution Network Topology Detection

Abstract: This paper proposes a data-driven approach to detect the switching actions and topology transitions in distribution networks. It is based on the real time analysis of time-series voltages measurements. The analysis approach draws on data from high-precision phasor measurement units ($\mu$PMUs or synchrophasors) for distribution networks. The key fact is that time-series measurement data taken from the distribution network has specific patterns representing state transitions such as topology changes. The proposed algorithm is based on comparison of actual voltage measurements with a library of signatures derived from the possible topologies simulation. The IEEE 33-bus model is used for the algorithm validation.

• The paper demonstrates that topology transitions yield distinct, rank-one voltage signatures that enable robust and accurate detection.
• It employs a signature library comparing real-time μPMU voltage trends to simulated models on the IEEE 33-bus system, ensuring sensitivity to switching events.
• The study confirms that strategic PMU placements can provide performance comparable to full instrumentation, promoting cost-effective network management.

Data-Driven Approach for Distribution Network Topology Detection

Introduction

The paper presents a novel data-driven algorithm designed to detect switching actions and topology transitions in distribution networks using real-time analysis of time-series voltage measurements. The approach is based on data from high-precision phasor measurement units ($\mu$PMUs), which are crucial for distribution network operations. Conventional topology detection methods rely heavily on state estimator results, which can be compromised by topology errors and sensitivity to measurement placement. This paper proposes an algorithm that compares actual voltage measurements with predefined signatures from simulations of possible topologies, validated through the IEEE 33-bus model.

Distribution Network Model and Observability

The distribution network is modeled as a graph, with buses and electrical lines represented as nodes and edges, respectively. The algorithm leverages sinusoidal signals and complex numbers to express voltages and currents, ensuring compatibility with real-time PMU data. The topology is defined by a bus admittance matrix, whose eigenstructure is manipulated to infer topology changes through a unique symmetric, positive semidefinite matrix $\green_\sigma$. This enables the extraction of specific signatures associated with switching actions, providing a crucial mechanism for topology detection.

Identification of Switching Actions

The core of the algorithm is driven by the notion that each switch status change leaves a distinct signature on the voltage profile. Under specific assumptions about line characteristics and switch actions, the paper derives that topology transitions yield rank-one matrices, where signatures are independent of variable states like voltages and loads, simplifying detection. The algorithm uses these properties to construct a library of signatures for detection.

Topology Detection Algorithm

The algorithm effectively compares trend vectors from voltage measurements to the signature library to identify topology changes. Using a heuristic threshold, the algorithm ensures robustness by handling measurement noise and load dynamics. The theoretical framework supports accurate detection even with limited PMU instrumentation, although optimal placement remains a challenge.

Performance Evaluation

The algorithm is tested on the IEEE 33-bus distribution test feeder, confirming its applicability and robustness across different scenarios, including noise and load variations. With few PMUs (seven strategically placed based on Monte Carlo simulations), the algorithm's performance is comparable to a setup with PMUs at every node, indicating cost-effective implementation possibilities.

Conclusion

The proposed data-driven approach significantly enhances distribution network topology detection using $\mu$PMU data, offering a robust solution compatible with real-time operations. Future work should focus on refining PMU placement strategies, handling diverse load models, and optimizing detection thresholds. The algorithm's potential to adapt to real-world conditions could pave the way for advanced network management applications, reducing the dependence on traditional state estimators and expanding observability with minimal instrumentation.