An Edge Computing-Enabled Optimal Control Method for Low-Voltage Distribution Networks

Abstract

High penetration of distributed photovoltaic (PV) generation in low-voltage distribution networks can cause voltage violations, limited PV accommodation, and uneven loading among neighboring distribution transformers (DTs). To address these issues, an edge computing-enabled optimal control method is proposed for low-voltage distribution networks organized as distribution transformer clusters (DTCs). An intelligent fusion terminal serves as the edge computing platform and locally performs measurement acquisition, day-ahead schedule generation, intraday rolling correction, safety verification, and control command issuance. In the day-ahead stage, PV generation scenarios are identified from forecast profiles, and scenario-dependent energy storage state-of-charge (SOC) reservation rules and soft open point (SOP) active power balancing schedules are generated. In the intraday stage, PV inverter reactive power, energy storage charging and discharging, necessary PV active power curtailment, and SOP power transfer are coordinated according to real-time measurements. A safety scaling and boundary verification mechanism is further introduced to ensure voltage security and device feasibility without repeatedly solving complex optimization models. Case studies on a three-DT system show that the proposed method reduces overvoltage occurrences in TQ1 by 81.0% and eliminates undervoltage occurrences under four PV scenarios. With SOP-based balancing, total overvoltage and undervoltage occurrences in the DTC are reduced by 4.35% and 88.99%, respectively, while PV utilization and loading ratio balance are improved. The intraday control time is approximately 20 ms, demonstrating the feasibility of real-time rolling control on resource-constrained edge computing platforms.