Abstract：The application proportion of overhead insulated conductors in distribution networks has been increasingly growing year by year, leading to a rising probability of single-phase line break faults. This study first establishes a single-phase line break fault model for distribution networks, analyzing voltage variations and waveform change patterns at both terminals of the breakpoint. A suitable combination of steady-state characteristics is selected as the input feature vector. Subsequently, an improved deep forest algorithm is proposed for fault classification. The proposed method employs mean pooling operations on class distribution vectors output by homogeneous forests to achieve linear reduction of feature dimensions. It concatenates the initial feature vector with averaged outputs from previous forest levels, thereby ensuring the continuous propagation of deep semantic correlations. Finally, PSCAD simulations are conducted to validate the effectiveness and accuracy of the proposed method through comparative analyses with various algorithms and ablation experiments.
聂洪城, 杨耿杰, 高伟. 基于改进深度森林算法的配电网不接地系统单相断线故障诊断方法研究[J]. 电气技术, 2025, 26(8): 61-69.
NIE Hongcheng, YANG Gengjie, GAO Wei. Research on a single-phase line break fault diagnosis method for ungrounded distribution systems based on improved deep forest. Electrical Engineering, 2025, 26(8): 61-69.
2025, Vol. 26 
