Data-driven probabilistic approach to improve hosting capacity in PV-Rich LV networks via phase balancing

Abstract

With the proliferation of photovoltaic (PV) systems in residential low-voltage (LV) networks, voltage rise has become a prevalent issue. This issue is further exacerbated by phase imbalances, significantly limiting the hosting capacity of LV feeders for additional PV integration. Traditionally, the identification of the phase imbalance is a labor-intensive task, prompting distribution network operators to explore alternative, straightforward, and effective solutions to deal with the phase imbalance issue. To this end, this paper proposes a novel framework that leverages only smart meter data to identify and rectify phase imbalances, consequently enhancing hosting capacity. In the proposed method, measurements from the secondary side of the transformer are compared with customer voltage data in the feeders using regression-driven analysis, which allows the estimation of the hosting phase for each customer. This process generates a connectivity matrix comprising three coefficients of determination (CoD) per customer; the highest CoD corresponds to the hosting phase of the respective customer. For the phase balancing process, customers are partitioned into subgroups in a such way that each updated subgroup contributes to the formation of an identity matrix. In order to realistically quantify the hosting capacity improvement resulting from phase balancing, a comparative hosting capacity analysis is carried out using the Monte-Carlo method. The effectiveness of the framework is validated on different networks, demonstrating that phase imbalances can be identified with 100% accuracy. Consequently, a 30% increase in PV penetration hosting capacity is achievable through phase balancing, without the need for alternative solutions.