Small current grouding fault frequently occurs in distribution network.In this paper,the design and installation of a new small current grouding line selection device is used to troubleshoot such fault.Device uses the nonlinear fitting ability of BP neural network,combined with wavelet energy,active power and fifth harmonic to determine the fault line.A design of hardware and software platform on ARM meeting to the requisition of reliability and rapidity in fault line detection is presented.ARM processor S3C2440 makes the controlling core of diagnostic mathine and C8051F060 makes the controlling core of the data acquisitionunit.The advantages of ARM,excallent real time interrupt response and high caculating speed,make the synthetical intellective algorithm to be fast processed in time.The operation test show that the device has good running stablity and fault detection is fast and reliable. It is a practical line selection device.