Abstract: The existing regional energy Internet (REI) is characterized as multi-dimension, and multi-agent operation. When the conventional centralized cloud service architecture is applied, the system is prone to high latency, cloud explosion, and other problems. To solve this issue, this paper proposes an edge-cloud collaborative architecture and its optimization strategy. First, the mathematical model of REI is established. Secondly, taking REI as the distributed edge control unit, the edge-cloud collaborative architecture is proposed, including the cloud service, edge service, and equipment layers. In each layer, the computation tasks are redistributed according to their practical workload. Next, the optimization strategies of the cloud service and edge service layers are designed individually, considering the game relationship of multi-stakeholders in the energy Internet. Finally, to improve the system restoration speed under emergencies, an emergency management method is proposed. Two multi-REI case studies verify that the proposed architecture has significant advantages in computing and emergency speed, compared with the conventional centralized cloud service architecture. In addition, the proposed optimization strategy can effectively improve the operation revenue of REIs.