Aimed at the challenges of low fault location accuracy in power cables

scarcity of fault samples

and the difficulty of jointly utilizing newly acquired and historical data when faults occur randomly

a small sample power cable fault location method based on multi-feature fusion and incremental learning is proposed. First

a multi-dimensional feature evaluation system is established

and an adaptive optimal waveform selection strategy is designed for high-quality data screening. Then

an incremental particle swarm optimization-support vector machine (PSO-SVM) waveform recognition model is introduced. By incorporating a historical data replay strategy that balances data importance and diversity

the model enables efficient use of both existing and new fault data

thereby overcoming the challenges of limited historical fault samples and catastrophic forgetting. Finally

a precise waveform bifurcation point localization method combining gradient difference analysis and differential search is proposed to improve location accuracy. Experimental results show that the proposed method achieves fault-location errors below 4% for power cable faults at different distances

demonstrating its effectiveness for power cable fault location under small-sample conditions.