Digital Twin-Driven Dynamic Reactive Power and Voltage Optimization for Large Grid-Connected PV Stations

With the increasing penetration of inverter-based photovoltaic (PV) generation, utility-scale grid-connected PV plants are frequently exposed to voltage regulation and voltage stability challenges driven by intermittent irradiance and limited reactive power flexibility under operating constraints. Conventional static Volt/VAR control schemes are typically designed for quasi-steady conditions and therefore struggle to respond to fast variations in PV output and network states. This paper presents a digital twin (DT)-enabled framework for dynamic Volt/VAR optimization in large PV plants. A four-layer DT architecture is developed to achieve real-time cyber-physical synchronization through multi-source data acquisition, secure transmission, fusion, and quality control. To balance model fidelity and computational efficiency, a hybrid physics–data-driven model is constructed, and a local voltage stability L-index is incorporated as an explicit security constraint. A multi-objective optimization problem is formulated to minimize node voltage deviations and reactive power losses while maximizing the static voltage stability margin. The problem is solved using an adaptive parameter particle swarm optimization (AP-PSO) algorithm with dynamic inertia and learning coefficients. Case studies on modified IEEE 33-bus and 53-bus systems demonstrate that the proposed method reduces the voltage profile index by up to 68.9%, improves the static voltage stability margin by 76.5%, and shortens optimization time by up to 30.3% compared with conventional control and representative meta-heuristic or learning-based baselines. The framework further shows good scalability and robustness under practical uncertainties, including irradiance forecast errors and measurement noise. Overall, the proposed approach provides a feasible pathway to enhance operational security and efficiency of grid-connected PV plants under high-penetration scenarios.