<p>Substation equipment defects are critical factors affecting the safe operation of power grids. Many non-rigid defects exhibit low detection accuracy and poor robustness due to fuzzy boundaries, irregular shapes, and small sizes. Additionally, naturally collected substation equipment defect images often present a long-tailed distribution, complicating the detection of tail categories and further reducing model performance. To address these challenges, we propose a causal-aware equipment defect detection framework based on the receptance weighted key value (RWKV) architecture. Firstly, to effectively detect equipment defects and enhance the distinguishability of image features, we integrate the RWKV architecture with a global receptive field. Secondly, we apply a causal-aware detection head to mitigate the long-tailed effect using causal intervention methods. Finally, we establish a comprehensive substation equipment defect detection dataset, providing a benchmark for non-rigid defect detection in substation power equipment. Extensive experimental results validate the effectiveness of our proposed framework.</p>