To face the challenges of the management of massive storage resources, low reliability and high cost of maintenance for distributed data of the future smart grid, we first built HDFS (Hadoop Distributed File System) based on a Hadoop cloud computing platform, demonstrated its powerful performance of data storage, and found many bottlenecks by an experiment in further improving storage performance for HDFS - the loss of node status information, the unevenness of system load and the low efficiency of storage, etc. Then we considered the drawbacks of the original program and gave a newly improved program for HDFS architecture of improving the state of data nodes information, setting data transcription coefficient and assessing the selected data node randomly. Finally, we balanced the load between nodes effectively, improved the storage efficiency significantly, and enhanced the users’ experience through the experiment on the simulation platform. We proved the feasibility that this strategy can enhance the storage performance of HDFS system.