As the core hub of the modern power system, the dynamic disturbance suppression and connection stability of the virtual circuit in the process layer of the intelligent substation are the key challenges to ensure reliable grid operation. Aiming at the problems such as the limited adaptability of traditional feedback-based strategies under nonlinear communication disturbances such as communication delay and packet loss, and the lack of real-time adaptability of static configuration mechanisms, this paper proposes an LADRC -based disturbance-estimation and supervisory scheduling method for process-layer virtual circuit connection and verification. By constructing the Linear Extended State Observer (LESO), multi-source disturbances (such as delay jitter, packet loss, and queue congestion) manifested across the network layer, protocol layer, and underlying physical infrastructure are estimated in real time as a filtered total-disturbance signal. The experimental results show that: Compared with the traditional baseline strategies, the proposed method improves communication delay-related performance, reducing the observed average end-to-end latency from 2.47 ms to 1.28 ms under typical disturbance scenarios, decreases the packet loss rate by 88.5% (0.06% vs. 0.52%), and increases the jitter suppression rate by 127%. The research results provide theoretical and technical support for improving the communication robustness and connection stability of smart substations.