采煤机拖拽电缆在运行中常受到外部挤压力作用，致使电缆绝缘发生局部放电，影响电缆使用寿命。现有研究侧重于局部放电规律和严重程度的分析，无法评估乙丙橡胶绝缘电缆所承受应力的大小，导致无法掌握矿用乙丙橡胶绝缘电缆的运行状态。针对该问题，提出一种基于改进深度森林（S−DF）的采煤机拖拽电缆挤压力识别方法。通过实验测量了不同挤压力下采煤机拖拽电缆的局部放电，分析了局部放电谱图、平均放电电流、最大放电量和击穿场强随所施挤压力和电压的变化规律，计算了局部放电的统计特征参量。基于统计特征参量，采用S−DF模型对挤压力大小进行识别。S−DF模型在深度森林（DF）中引入Stacking集成算法，以提升识别准确率。研究结果表明：不同电压下，最大放电量和平均放电电流均随着挤压力的增大而减小；击穿场强随着挤压力的增大呈先增大后减小的趋势，挤压力大于2 000 N时的击穿场强小于未挤压时的击穿场强；不同挤压力下的局部放电统计特征参量可作为放电指纹，S−DF模型能准确地识别电缆所受挤压力的大小，且识别率高于其他传统分类算法。

The dragging cable of shearer is often subjected to external squeezing pressure during operation, which causes partial discharge of the cable insulation and affects the service life of the cable. The existing research focuses on the analysis of partial discharge law and severity, and cannot evaluate the magnitude of stress borne by ethylene propylene rubber insulated cables. This results in the inability to grasp the operating status of mining ethylene propylene rubber insulated cables. In order to solve the above problems, a method based on improved Stacking-deep forest (S-DF) is proposed for recognizing the squeezing force of shearer dragging cables. The partial discharge of shearers dragging cables under different squeezing pressures is measured through experiments. The variation law of partial discharge spectra, average discharge current, maximum discharge amount, and breakdown field strength with the applied squeezing pressure and voltage are analyzed. The statistical feature parameters of partial discharge are calculated. Based on statistical feature parameters, the S-DF model is used to recognize the magnitude of squeezing pressures. The S-DF model introduces Stacking ensemble algorithm in deep forest (DF) to improve recognition accuracy. The research results indicate that under different voltages, the maximum discharge capacity and average discharge current decrease with the increase of   extrusion pressure. The breakdown field strength shows a trend of first increasing and then decreasing with the increase of squeezing pressure. When the squeezing pressure is greater than 2 000 N, the breakdown field strength is lower than that of the non squeezing one. The statistical feature parameters of partial discharge under different squeezing pressures can be used as discharge fingerprints. The S-DF model can accurately recognize the magnitude of squeezing pressure on cables, and the recognition rate is higher than other traditional classification algorithms.