Cost-Benefit Analysis of Phase Balancing Solution for Data-Scarce LV Networks by Cluster-Wise Gaussian Process Regression

Wangwei Kong*, Kang Ma, Lurui Fang, Renjie Wei, Furong Li

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

8 Citations (Scopus)


Phase imbalance widely exists in the UK's low voltage (415 V, LV) distribution networks. The imbalances not only lead to insufficient use of LV network assets but also cause energy losses. They lead to hundreds of millions of British pounds each year in the UK. The cost-benefit analyses of phase balancing solutions remained an unresolved question for the majority of the LV networks. The main challenge is data-scarcity - these networks only have peak current and total energy consumption that are collected once a year. To perform a cost-benefit analysis of phase balancing for data-scarce LV networks, this paper develops a customized cluster-wise Gaussian process regression (CGPR) approach. The approach estimates the total cost of phase imbalance for any data-scarce LV network by extracting knowledge from a set of representative data-rich LV networks and extrapolating the knowledge to any data-scarce network. The imbalance-induced cost is then translated into the benefit from phase balancing and this is compared against the costs of phase balancing solutions, e.g., deploying phase balancers. The developed CGPR approach assists distribution network operators (DNOs) to evaluate the cost-benefit of phase balancing solutions for data-scarce networks without the need to invest in additional monitoring devices.

Original languageEnglish
Article number8959309
Pages (from-to)3170-3180
Number of pages11
JournalIEEE Transactions on Power Systems
Issue number4
Publication statusPublished - Jul 2020
Externally publishedYes


  • Cost-benefit analysis
  • Gaussian process regression
  • low voltage
  • phase balancing
  • phase imbalance
  • power distribution
  • three-phase system


Dive into the research topics of 'Cost-Benefit Analysis of Phase Balancing Solution for Data-Scarce LV Networks by Cluster-Wise Gaussian Process Regression'. Together they form a unique fingerprint.

Cite this