复杂环境下绝缘子故障状态自适应免疫检测方法

Adaptive Immune Detection Method for Insulator Fault Conditions in Complex Environments

中文关键词:梯度向量流蛇模型  负选择算法  绝缘子  故障状态检测  退化等级划分

keywords:gradient vector flow snake model  negative selection algorithm  insulator  fault state detection  degradation level classification

中文摘要:ation level classification 摘要：现有方法因固定处理模式或单一优化策略，难以提升无人机航拍复杂环境下绝缘子故障检测效率与准确性。针对这一问题，提出一种融合自适应视觉增强与免疫机制的绝缘子故障状态检测及退化等级划分一体化方法。首先，构建集成卷积块注意力模块和坐标注意力机制的轻量级SqueezeNet网络作为背景分类器，智能识别图像背景类别并动态优化处理参数，有效克服光照变化与复杂环境纹理干扰。其次，采用改进梯度向量流蛇模型精确提取绝缘子轮廓，并提取轮廓的关键特征构建抗原向量，通过特征阈值初步判定故障状态。再次，利用克隆选择机制改进负选择算法构建轻量级抗体库，通过抗原与抗体的特异性亲和度匹配，最终精准判定故障状态。最后，基于故障状态调用相关模型计算核心电气参数，评估绝缘子的退化等级并支撑运维决策。经实验分析验证，该方法综合性能优于对比方法，为电力设备智能运维提供了有效解决方案。

Abstract:Existing methods adopt fixed processing modes or single optimization strategies，which limit the efficiency and accuracy of insulator fault detection based on UAV aerial images in complex environments. To solve this problem，this paper proposes an integrated method combining adaptive visual enhancement and immune mechanism for insulator fault state detection and degradation level classification. First，a lightweight SqueezeNet embedded with convolutional block attention module and coordinate attention mechanism is constructed as a background classifier. It can intelligently identify image background types and dynamically optimize processing parameters to suppress interference caused by illumination changes and complex background textures. Second，an improved gradient vector flow snake model is adopted to accurately extract insulator contours. Key contour features are extracted to construct antigen vectors，and the fault state is preliminarily determined according to feature thresholds. Third，the negative selection algorithm is improved by the clonal selection mechanism to construct a lightweight antibody library. The final fault state is judged through specific affinity matching between antigens and antibodies. Finally，relevant models are invoked according to fault states to calculate core electrical parameters，so as to evaluate insulator degradation levels and support operation and maintenance decisions. Experimental results verify that the proposed method has better comprehensive performance than comparative methods，and provides an effective solution for intelligent operation and maintenance of power equipment.