With the rapid development of the intelligent power grid, the scale of the intelligent energy meters in Eleco Energy Data Acquire System is increasing day by day. Real-time analysis of massive data and maintenance of electric energy meter is a big challenge for Eleco Energy Data Acquire System. Recently, the highly data-parallel computing and rapidly training large-scale neural network of Graphics Processing Unit(GPU) has been a new research hotspot in the field of high-performance parallel computing. However, until now, GPU has not been applied to Eleco Energy Data Acquire System. In order to improve the performance of statistical analysis and realize the prediction of meter health status, this paper do research on how to realize GPU in Eleco Energy Data Acquire System, to support the lean operation and maintenance of electric energy meter.