In order to accurately identify faults in multi-layer complex distribution network, an on-line fault node identification and location method is proposed. This method injects the high frequency test signal into the power grid, obtains the impedance parameters from the grid information provided by smart meters to the central unit, and detects whether there are fault nodes according to the impedance characteristics and the response results under the high frequency test signal of the power grid. The method is used to simulate 45-node 3-layer distribution network under various conditions, then the 8-node low-voltage network model is tested. The influence of analog-to-digital conversion resolution and measurement error tolerance on detection accuracy is discussed. The simulation and test results show that the difference between the received signal voltage of high frequency and the estimated voltage at the fault node is the largest, indicating that the method can accurately identify the fault node in the distribution network. This work is supported by National Natural Science Foundation of China (No. 51767008), Science and Technology Project of State Grid Corporation of China (No. PDB17201800040), and Fund of China Electric Power Research Institute (No. PD83-17-006).