Multi-Agent Volt/Var Control in PV-Rich Active Distribution Networks: A Deployment-Oriented Review of Architectures, Safety, and Readiness

The rapid growth of photovoltaic (PV) generation in active distribution networks has intensified the need for Volt/Var control with the ability to sustain voltage compliance under fast variability, limited observability, and heterogeneous device interaction. This review focuses on how architecture class, safety mechanisms, hosting-capacity treatment, and validation realism jointly affect deployment readiness. In this review, 127 peer-reviewed papers from 2020 to 2026 that focus on coordinated Volt/Var control for PV-rich active distribution networks have been evaluated. These research works were categorized based on architecture class, coordination family, communication structure, observability regime, safety mechanism, constraint coverage, hosting-capacity treatment, storage integration, reproducibility signal, and evaluation context. The analysis shows that many approaches in the literature are either hybrid or learning-based, with multi-agent reinforcement learning (MARL) dominating, and stronger safety mechanisms appearing mainly in the smaller set of optimization-based methods. Hosting capacity is often presented as a potential benefit of the proposed method rather than a control problem addressed during runtime. Real-feeder studies, hardware-in-the-loop (HIL) testing, delay-aware evaluation, and open implementation artifacts are less common. The synthesis is organized through a coordination, assurance, integration, and realism (CAIR) framework.