Abstract: Aiming at the low accuracy of single feature quantity and the difficulty of obtaining a large number of fault sample datasets in the process of line selection, an Improved Bat Algorithm(IBA) based on feature fusion is proposed to optimize Support Vector Machine(SVM) fault line selection method for small current grounding system. First, the zero-sequence current signal of the line is processed by combining S-transform threshold filter with time-frequency filter based on time-spectrum distribution. Then, the S-transform energy relative entropy and singular entropy are selected by using feature fusion technology as double-feature quantities. Then the support vector machine is optimized by using improved Bat algorithm. The IBA-SVM optimization classifier was constructed to classify and process the fault feature data set, and the result of line selection was obtained. Finally, a simulation circuit built in PSCAD/EMTDC is used to verify the accuracy and effectiveness of the proposed method.