The invention discloses a real-time detection method, server and system for an infrared image power equipment component. The method comprises the following steps that: (1) obtaining an infrared imagewhich contains a known power equipment component to form a sample set, wherein each infrared image in the sample set is labeled with a target frame and is provided with a component-level tag, and thetarget frame is an image area which contains the single known power equipment component; (2) constructing a neural network detected on the basis of the power equipment component of a YOLO target detection frame, and inputting the infrared image in the sample set and the corresponding component-level tag into the constructed neural network, and training the infrared image; and (3) adopting the neural network detected by the trained power equipment component to process a to-be-measured infrared image which is provided with the known component-level tag, and outputting the detection result of thepower equipment component.