为了能够准确地获取变电站敞开式设备的表面温度信息,准确定位热异常区域并以此确定设备运行状态,提出了一种变电站设备红外温度预测方法。采用量子遗传(QGA)-正交最小二乘算法(OLS)优化了径向基神经网络(RBFNN),通过将红外热像图像素与温度作为该神经网络的输入和输出量,建立了从红外图像获取设备温度的理论模型。结合同一场景的红外图像与可见光图像配准技术,即可直接从可见光图像上获取对应位置的红外温度 In order to accurately obtain the surface temperature information of open equipment in substation, accurately locate the thermal anomaly area and determine the operating status of the equipment, a method of infrared temperature prediction for substation equipment is proposed. Radial basis neural network (RBFNN) is optimized by QGA-OLS algorithm. By using the pixels and temperature of infrared thermal image as input and output of the neural network, The theoretical model of image acquisition equipment temperature. With the same scene of infrared image and visible light image registration technology, you can directly from the visible light image to obtain the corresponding position of the infrared temperature