TY  - EJOU
AU  - Fu, Xieli 
AU  - Wu, Guoxing 
AU  - Shi, Yujie 
AU  - Jiang, Xinming 
AU  - Yang, Wenfeng 

TI  - Research on Intelligent Network Formation and Flexible Interconnection Mechanisms for Low-Voltage Distribution Networks in Cyber-Physical Systems
T2  - Energy Engineering

PY  - 
VL  - 
IS  - 
SN  - 1546-0118

AB  - Low-voltage distribution networks face critical challenges from large-scale distributed renewable energy integration, including bidirectional power flow control, voltage stability, and multi-microgrid coordination. Existing approaches are limited by single-scenario optimization without cyber-physical coupling considerations, device-level control lacking system-level coordination, and absence of unified frameworks bridging topology reconfiguration with power exchange. This study proposes a collaborative framework integrating intelligent network formation and flexible interconnection within a cyber-physical system environment. A four-layer architecture featuring edge-cloud collaborative computing and fault-tolerant hybrid communication was constructed. A multi-agent coordination algorithm based on TD3 deep reinforcement learning was developed for distributed decision-making with consensus-guaranteed convergence. A flexible interconnection control strategy based on virtual synchronous generator technology was designed for multi-microgrid power exchange with inherent inertia support. Validation was conducted through offline simulation (IEEE 33/69-node systems), hardware-in-the-loop testing (OPAL-RT ePHASORSIM), and field deployment across residential (150 households, 450 kW PV), industrial (5 MW peak load, 2 MWh storage), and commercial (3 MW peak demand) scenarios in Shenzhen. The multi-agent algorithm achieved over 60% faster convergence than conventional metaheuristics, with a load balancing index of 0.098 and computation time of 52 s. The flexible interconnection strategy reduced voltage regulation time by 73% (from 12 s to 3.2 s) and frequency deviation by 81% (from ±0.8 Hz to ±0.15 Hz). System performance reached 99.97% power supply reliability, 91.2% energy efficiency, 41% network loss reduction, and 21% annual operating cost reduction with 6.5–8.5 year payback periods. This framework provides a validated technical solution for intelligent upgrading of low-voltage distribution networks under high renewable penetration.
KW  - Deep reinforcement learning; virtual synchronous generator; cyber-physical system; low-voltage distribution networks; multi-agent coordination

DO  - 10.32604/ee.2026.074212