The paper presents a Stockwell transform (ST) and artificial neural network-based approach for the detection and localization of faults in distribution systems considering the complexities of network architecture and the distributed generation (DG) integration. Firstly, a faulty-line detection technique is developed based upon the total harmonic distortion of the fault current signal captured from the line ends. Then, the ST coefficients of the faulty signal are used as an attribute to classify the fault. Finally, the root-mean-square values calculated from the ST coefficients are used for the fault localization. The algorithm is tested on an IEEE 13-bus unbalanced and a 52-bus practical Indian balanced distribution system with DG. The results are compared with some existing algorithms. The comparisons and the test results show that the proposed technique can be a useful tool in detecting the faults in complicated fault scenarios. © 2018 John Wiley & Sons, Ltd.