With the continuous advancement of power grid construction, massive data demands higher processing capabilities from collection terminals. The efficiency of multi APP task scheduling for existing power collection terminals is not high, which restricts the development of the power grid in areas such as precise perception, unified IoT, and edge intelligence. This article investigates an efficient task scheduling technique based on the Storm distributed real-time computing architecture. The optimal system parameters are selected using the control variable method, and simulation results demonstrate that this technique can effectively improve throughput and reduce latency. The promotion and application of this technology will help the collection terminal to more efficiently support full data collection, station area perception, various events, data reporting, and other work.