Abstract: The traditional method for locating fault sections in new energy distribution networks relies on single or limited electrical quantities for threshold judgment, which makes it difficult to capture complex fault characteristics. There are logical conflicts and insufficient accuracy in the combination of multi-source information for judgment. Therefore, this article proposes an online positioning method for fault sections in new energy distribution networks based on improved naive Bayes. This method first constructs a zero sequence equivalent circuit for faults, selects three complementary electrical quantities of measuring point voltage, measuring point current, and transient harmonics, calculates multidimensional fault characteristic information, and forms a combined criterion system to comprehensively describe the fault state. Then, an improved Naive Bayes classifier is introduced to measure the degree of feature correlation based on mutual information. A maximum weight spanning tree is constructed to relax the assumption of feature independence, optimize the joint probability calculation of multiple electrical quantity features, and achieve high-precision combination judgment of fault sections. The experimental results show that the average F1 score of this method in multiple tests reaches 0.96, and the positioning accuracy is significantly better than traditional methods. This verifies its effectiveness and practicality in accurate combination judgment in complex new energy distribution network scenarios, and can achieve fast and accurate online positioning of new energy distribution network faults.