The invention discloses a surface potential pilot frequency comparison-based ground grid defect identification method. Reference frequency point current and testing frequency point current are respectively inputted into any ground grid downlead of a ground grid to be identified, moreover, the surface potentials of the two frequency point currents respectively corresponding to a plurality of observed points are detected, and the surface potential difference between the two frequency point currents corresponding to each observed point, which is pilot frequency surface potential difference, is calculated; a pilot frequency surface potential difference-observed point variation curve is drawn, and according to the curve, the gray correlation algorithm or the fuzzy clustering algorithm is utilized to obtain the observed points with over-small or over-large pilot frequency surface potential differences, which are defective points. The invention also provides a system for implementing the method, which comprises a sinusoidal variable-frequency alternating-current power supply, a measurement circuit and a control center. By comparing the differences between surface potential responses when the ground grid is excited by currents with differential frequencies, the invention can identify the defects or faults of the ground grid, and compared with conventional methods, the method has the characteristics of convenient detection and high anti-interference capability, and is easy to operate.