Probabilistic Assessment of Unbalance in Distribution Networks Based on Limited Monitoring

  • Zhixuan Liu

Student thesis: Phd


This thesis assesses the voltage unbalance in distribution networks due to load asymmetry or line asymmetry, based on measurement data from a limited number of monitors. The main outcomes of this research are a probabilistic methodology for estimating both momentary and long term unbalance and an optimal monitor placement providing the highest accuracy for the monitored level of unbalance. With increasing numbers of large single-phase loads and distributed generation integrated into the power system, the future distribution network is expected to be more flexible, robust and "smart". This results in the requirement for high quality of electricity supply to be delivered to customers and is a challenge for the operation of the system. As the unbalance results in excessive heating, accelerated thermal ageing, reduction of efficiency and financial losses, the unbalance should be regulated to be below the statutory limit. Given the fact that unbalance is a long term phenomenon that may not cause any triggering of protection or faulty response of equipment, it can be only determined from available data such as loading levels of a network and the incomplete monitored voltage of a network. Due to limited monitoring in the network and therefore insufficient data, unbalance may be unobservable. This thesis therefore aims to develop a methodology to increase the observability of unbalance in the network in spite of limited monitoring.This research uses Voltage Unbalance Factor (VUF) to quantify the level of unbalance. The first major part investigates the unbalance caused by asymmetrical loadings. By properly identifying the source of unbalance, the basic patterns of propagation of unbalance under possible scenarios are revealed and a methodology of probabilistic estimation of unbalance can be developed accordingly. Seen from the MV level of distribution networks, the loads are usually in the constant power form. Therefore, the variation in the load can be modelled by changing either active power or reactive power or both of them, depending on the data availability. The combinations of daily loading curve at buses and the normally distributed power factors in three phases of loads are used to create an unbalanced condition at the sources. Realistic assumptions of power factors and reasonable categories of types of loads result in realistic modelling of the unbalanced load. The probabilistic VUFs at different buses in the network are calculated and the weak areas in the network are identified using heat maps. The simulation results match the real VUF levels measured in the distribution network. The second part of the thesis explores the influence of asymmetrical lines in addition to the asymmetrical loading on propagation of unbalance. The last part provides a guideline for optimal monitor placement for unbalance. Two methods, manual ranking of buses and automatic optimization using Genetic Algorithm, are proposed. The two methods indicate the same optimal locations for monitor placement in the network. The developed methodologies enable the assessment of unbalance in the network when monitoring is limited and can be applied to real networks to assess the level of unbalance at non-monitored buses.
Date of Award31 Dec 2014
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorJovica Milanovic (Supervisor)


  • Power Quality
  • Monte Carlo
  • Voltage Unbalance

Cite this