Abstract: With the rapid development of renewable energy and distributed resources, centralized dispatching faces such challenges as high dimensionality, strong coupling, and real-time requirements. To address the issues of heavy computation and privacy leakage in traditional centralized optimization, this paper proposes a method for characterizing the external characteristics of distributed source-grid-load-storage clusters based on cluster partitioning and multi-parametric programming. Firstly, an index system for cluster partitioning is constructed using electrical connection strength and energy supply-demand matching capability. This enables structural dimensionality reduction for the system. Then, a multi-parametric programming approach is used to convert the internal highly coupled optimization problem into a set of piecewise affine mapping functions between exchanged power and control boundaries, thus enabling establishment of an explicit response function set for external dispatching. Finally, real-time dispatching is achieved by quickly matching the response functions from the pre-constructed set. Case studies demonstrate that in the IEEE 33-bus system, the total dispatching time for 1000 operations is reduced from 2791.3 seconds to 433.07 seconds; while in the IEEE 123-bus system, the average dispatching time is decreased from 475.18 seconds to 4.9 seconds. The proposed method exhibits significant advantages in model dimensionality reduction and solution efficiency improvement, providing a new path for the efficient integration and online coordinated control of distributed resources under a centralized dispatching framework.