According to the invention, real-time monitoring and behavior analysis are carried out on an operator, non-standard operation and behaviors are intervened, video monitoring and error prevention are integrated, and post-event tracing of an accident can be realized; and an intelligent analysis function is achieved, specific events and user behaviors are monitored, and intelligent early warning of accidents is achieved. A method comprises the following steps: erecting a plurality of cameras on a transformer substation equipment site, acquiring video signals of the cameras in a transformer substation, and transmitting the video signals to a computer; building an artificial intelligence OpenCV visual library, and utilizing a GPU parallel processing technology; and constructing a deep neural network to recognize the image, and realizing behavior analysis, judgment and monitoring of an operator. Video monitoring, intelligent analysis, behavior monitoring and anti-misoperation locking technologies are deeply fused, and when a specific behavior is found, an alarm is automatically generated to inform monitoring personnel of timely early warning, linkage and processing.