With the development of State Grid's information technology, new challenges are presented to the data processing, concurrent requests and responsiveness of conventional power information systems. According to the characteristics of power information system's data processing, a distributed cache based on MongoDB is proposed in this paper. The operational mechanism, server-side architecture and the design of client's function modules in distributed cache are analyzed and discussed. The distributed cache based on MongoDB can effectively reduce the payload on access to the database layer of power information system, and then improve the performance of power system as a whole. It adopts the distributed files to store the cache data with good reliability and scalability to support the function of data redundancy backup and failure recovery. Now the distributed cache based on MongoDB has been successfully applied to the Project Management System of one power corporation.