基于SE-YOLOV4的变电站断路器分合状态识别算法

Substation Circuit Breaker Opening and Closing State Recognition Algorithm Based on SE-YOLOV4

针对高压变电站断路器人工检测速度慢、误差大等问题,提出基于YOLOv4改进的变电站断路器分合状态识别方法;针对电力系统背景复杂、断路器分合状态不易识别的问题,加入通道注意力机制,关注目标的显著性特征,忽略非目标区域,然后使用路径聚合网络有效提取目标特征;针对数据样本单一性的问题,提出SE-YOLOv4算法,在其中加入数据增强技术,提高模型的泛化能力,使算法网络具有更强的鲁棒性。实验结果表明,该算法的精确率为97%,召回率为73.45%,F1为0.84,平均精确率为79.2%,相比原算法的平均精确率提高了2.6%,表明基于深度学习的检测方法可快速、高效地检测目标,避免人工检测出现的问题。

Aiming at the problems of slow manual detection of high-voltage substation circuit breakers and large errors,this paper proposes a research method based on YOLOV4 improved substation circuit breaker switching status identification. In view of the complex background of the power system and the difficulty of identifying the opening and closing status of the circuit breaker,this paper adds the channel attention mechanism to focus on the salient features of the target and ignore the non-target areas. Then use the path aggregation network to effectively extract the target features. Aiming at the unity of data samples,this paper proposes the SE-YOLOv4 algorithm to add data enhancement technology to it to improve the generalization ability of the model,and the algorithm network has better robustness. In the data set used in this paper,the accuracy of the experimental results is 97%,the recall rate is 73.45%,the F1 is 0.84,and the average accuracy is 79.00%,which is2.45% higher than the average accuracy of the original algorithm. Therefore,the detection method based on deep learning can detect the target quickly and efficiently,and avoid the problems of manual detection.