Data-Driven Surrogate Modeling for Voltage Security Assessment and EV Hosting Capacity in Solar-Integrated Distribution Networks

Today’s distribution networks face challenging voltage stability problems from the increasing deployment of electric vehicles (EVs) and solar photovoltaic (PV) systems. The current methods of voltage stability assessment rely on computation-intensive power flow analysis, thereby limiting their applicability for real-time decision-making. This paper introduces a data-driven machine-learning-based approach to rapidly and interpretatively assess voltage security when considering EV and solar variability. The Voltage Security Score (VSS) is a continuous measure of security based on the power flow results of the Newton–Raphson calculation. The surrogate machine learning model is then trained only with actionable input variables including EV demand, solar generation and temporal variables to avoid leaking information from internal system states. The proposed model has good predictive accuracy, which can well reflect the nonlinear relationship between the operating conditions and the voltage stability. The framework supports VSS-based security-region screening, visualization of the operating envelope, and preliminary estimation of EV loading levels under different solar-generation conditions. The results show that the higher the level of solar generation, the greater the voltage stability and the higher the EV penetration. It provides a novel and efficient solution that is relevant from a practical perspective to plan and manage modern power system networks that include significant amounts of EVs and renewables.