This paper presents an approach to single line-to-ground fault location in a power distribution system based on the time lag of different modal components . It is a single-ended method applicable to unsynchronized measurements. The fault location formulation is derived by analyzing the traveling wave propagation characteristic in a distributed parameter model. To tackle the key problem of ground-mode velocity estimation, an iterative velocity method based on prior knowledge is first proposed. Since the practical measurements are not ideal because of sensor accuracy and nonequivalent transient response, fuzzy neural network is utilized to process traveling wave data acquired at multiple measurements to get an error-tolerant effect. Furthermore, electromagnetic transient simulations for a 34-bus test feeder system under various conditions of fault resistances, inception angles, and distributed generator access reveal the robust estimation of fault location. The performance of the proposed method is also validated by simulations with noisy and noise-free measurements.