微型电脑应用

为了解决变电站万用接地线在进行导通状态检测中学习模型鲁棒性较差、检测结果偏差较大的问题，设计一种基于单发多盒检测器(SSD)深度学习模型的变电站万用接地线导通状态检测方法。针对变电站万用接地线的检测需求设计接地线导通状态检测设备结构，将采样模块、通信模块进行重新设计，分析SSD深度学习模型结构，利用区域候选框检测多尺度特征图，获取特征信息实现图像判断，确定接地线导通状态检测流程，完成变电站万用接地线的导通状态检测。测试结果表明，所设计方法得到的精确率和召回率均能够达到90%以上，验证了所设计方法在实际应用中的可靠性。

In order to solve the problem of poor robustness of learning model and large deviation of detection results in conduction state detection of substation universal ground wire, a conduction state detection method of substation universal ground wire based on single shot multibox detector(SSD) deep learning model is designed. According to the detection requirements of substation universal ground wire, this paper designs the structure of the ground wire conduction state detection equipment, redesigns the sampling module and communication module, analyzes the SSD deep learning model structure, uses regional candidate boxes to detect multi-scale feature maps, obtains feature information to achieve image judgment, and determine the ground wire conduction state detection process to complete the conduction state detection of substation universal ground wire. The test results show that the accuracy and recall rate of the test results obtained by the design method can reach over 90%, verifying the reliability of the design method in practical applications.