Solving scalability issues in calculating PV hosting capacity in low voltage distribution networks

Abstract: The share of end-users with installed rooftop photovoltaic (PV) systems is continuously growing. Since most end-users are located at the low voltage (LV) level and due to technical limitations of LV networks, it is necessary to calculate PV hosting capacity. Most approaches in calculating a network's hosting capacity are based on three-phase optimal power flow (OPF) formulations. Linearized and relaxed three-phase OPF formulations respectively lose their accuracy and exactness when applied to solve the hosting capacity problem, and only non-linear programming (NLP) models guarantee the exact solution. Compared to linearized or relaxed models, NLP models require a higher computational time for finding an optimal solution. The binary variables uplift the problem to mixed-integer (MI)NLP and increase the computational burden. To resolve the scalability issues in calculating the hosting capacity of single-phase connected PVs, we propose a method that does not entail binary variables but still ensures that PVs are not connected to more than one phase at a time. Due to a risk of a sub-optimal solution, the proposed approach is compared to the results obtained by the MINLP formulation. The comparison includes values of the solution time and technical quantities such as network losses, voltage deviations, and voltage unbalance factor.