Consensus based phase connectivity identification for distribution network with limited observability

TL;DR

This paper introduces a consensus algorithm-based method for accurately identifying phase connectivity in large distribution networks with limited observability, using synthetic data and zone clustering to improve scalability and robustness.

Contribution

It proposes a novel consensus algorithm approach utilizing multiple reference points and zone clustering to enhance phase identification accuracy in large, limited-observability distribution networks.

Findings

Achieves 98.4% accuracy in phase identification with 1% measurement error.

Consensus algorithms outperform single-reference methods in robustness.

Method scalable to large networks with minimal measurement infrastructure.

Abstract

The mitigation of distribution network (DN) unbalance and the use of single-phase flexibility for congestion mitigation requires accurate phase connection information, which is often not available. For a large DN, the naive phase identification proposed in the majority of the prior works using a single voltage reference does not scale well for a multi-feeder DN. We present a consensus algorithm-based phase identification mechanism which uses multiple three-phase reference points to improve the prediction of phases. Due to the absence of real measurements for a real-suburban German DN, the algorithms are developed and evaluated over synthetic data using a digital twin. To utilize strongly correlated measurements, the DN is clustered into zones. We observe those reference measurements located in the same zone as the single-phase consumer leads to accurate prediction of DN phases. Four consensus algorithms are developed and compared. Using numerical results, we recommend the most robust phase identification mechanism. In our evaluation, measurement error, and the impact of the neutral conductor are also assessed. We assume limited DN observability and apply our findings to a German DN without smart meters, but only less than 8% of nodes have measurement boxes along with single-phase consumers with a home energy management system. Voltage time series for 1 month (hourly sampled) is utilized. The numerical results indicate that for 1% accuracy class measurement, the phase connectivity of 308 out of 313 single-phase consumers in a German DN can be identified. Further, we also propose metrics quantifying the goodness of the phase identification. The phase identification framework based on consensus algorithms for DN zones is scalable for large DN and robust towards measurement errors as the estimation is not dependent on a single measurement point.