为了提高变电站设备故障的识别精度、图像清晰度、细节完整度，提出了基于时间序列算法的故障红外识别方法。首先，设计变电站设备红外遥视与故障识别框架，利用混合探测器-云台组合采集其红外、可见光视频图像，经光通信设备及光纤传输视频图像至监控中心，通过滑动窗口结合时间序列筛选并标记可疑故障图像；然后，经小波阈值变换的中值去噪方法去除噪声后，利用改进分水岭算法完成图像目标分割，采用Zernike不变矩提取其特征值，将其作为支持向量机的输入，完成变电站设备分类；最后，结合设备的温度信息，通过划分故障类型、故障识别规则，实现不同设备的故障识别。结果表明，采用该方法去噪后，图像清晰度、对比度大幅提升，分割后目标设备细节完整度高，可实现不同设备的故障识别。

In order to improve the accuracy of defect recognition, image clarity and detail integrity, a fault recognition method of substation equipment based on time series algorithm was proposed. The framework of infrared remote viewing and fault identification of substation equipment was designed. The hybrid detector PTZ combination was used to collect its infrared and visible video images, transmit the video images to the monitoring center through optical communication equipment and optical fiber. The suspicious fault images were screened and marked through sliding window combined with time series, and the noise was removed by the median denoising method of wavelet threshold transform. The improved watershed algorithm was used to complete the image target segmentation, and the Zernike invariant moment was used to extract its eigenvalue, which was used as the input of support vector machine to complete the substation equipment classification. Combined with the temperature information of the equipment, the defect identification of different equipment was realized by dividing the defect types and designing the defect identification rules of substation equipment. The experimental results showed that after denoising, the image clarity and contrast were greatly improved, the detail integrity of the target equipment was high, and different equipment faults could be identified.