Abstract: To address the complexity of electricity trading in new-type power systems-characterized by tightly coupled generation–grid–load–storage and multiple market mechanisms-this paper proposes an upgraded decision support system for multi-dimensional uncertainty. The system adopts a cloud–edge–terminal architecture and forms a closed loop of “perception– decision–execution.” It integrates uncertainty optimization with reinforcement learning to support rolling strategy generation and adaptive control. A digital cockpit enables real-time data monitoring, strategy simulation, and access management. Data security is enhanced through classification, federated learning, and trusted execution. Field deployment on a provincial trading platform shows notable improvements in cost control, response time, clearing efficiency, and economic returns, validating the system’s adaptability and practical value.