To solve the communication bottleneck problem of low-voltage distribution secondary systems, a flexible self-organizing network communication architecture based on the internet of things is proposed, which achieves efficient communication through a three-stage technology chain. Firstly, during the network initialization phase, a communication parameter database covering topology, channel quality, and latency characteristics is constructed using layered scanning and time series analysis methods. This database accurately identifies network blind spots and optimizes the spatial distribution of relay nodes. On this basis, an improved ant colony optimization algorithm is proposed, introducing a local search strategy to avoid local optimal traps, designing a dynamic pheromone update mechanism, and combining multi-objective constraints such as delay, energy consumption, and bandwidth to achieve fast search for the optimal communication path. At the same time, in order to improve the anti-interference ability, edge computing and channel state awareness technology are integrated to establish an adaptive routing strategy in noisy environments. Through real-time feedback, path weight is adjusted to maintain a high effective data transmission rate under strong noise interference. In the end, experimental verification shows that the proposed technology has low network complexity, short optimization time, and strong anti-interference ability. It has been successfully applied to the state awareness and intelligent control of distribution stations, providing a low latency and high reliability communication solution for the massive terminal access to new power systems.