In response to the challenges associated with feature extraction and the low accuracy in identifying partial discharge (PD) types in transformers, this paper proposes a novel transformer PD pattern recognition method that integrates multiple feature extraction techniques with ensemble learning. Experimental results demonstrate that this method achieves an overall accuracy of 99.81% across four distinct PD modes in transformers—namely, tip discharge, air gap discharge, surface discharge, and floating discharge—with individual category accuracies exceeding 96%. These results outperform those of conventional single-model approaches, thereby confirming the effectiveness and robustness of the proposed method. By enabling rapid and precise identification of PD types, this approach offers significant engineering value for enhancing the reliability of online insulation condition monitoring in power systems.