Abstract: Power supply monitoring system provides important real-time data for machine research, auto beam adjustment and deep learning on HIRFL (Heavy Ion Research Facility Lanzhou). During the EPICS upgrade of HIRFL, the power supply monitoring system was redesigned and a distributed structure was adopted to form an extended layout of the upper layer/middleware and the bottom layer. The middleware processes the upper layer data and sends data to the lower layer, avoiding invalid parameters applied to the lower devices and simplifying the handling of exception. To facilitate beam adjustment, the display interface is integrate, displaying large amounts of data on different OPI is optimized and the display is improved. In order to carry on the off-line analysis of the data, a NoSQL database cluster was built to realize multiple copies of high priority data and to store multi-node real-time data, also to improve data security and database read/write speed, in addition to avoid designing complicated table structure and to expand database cluster optimization and nodes increasing.