Abstract: The highly dynamic operation environment of the distribution network without power failure leads to the multi-modal characteristics of the data in the high-risk areas. It is difficult to extract the multi-mode key features from the non-power failure operation scenarios of the distribution network, which reduces the effect of risk identification. Therefore, the AI technology is introduced, and the high-risk identification method of the distribution network operation area without power failure is proposed. The Gaussian kernel function is used to quantify the correlation between the data points, the random motion state transition probability and diffusion distance are calculated, and the distribution network operation area of the potential risk points is displayed through the three-dimensional visualization interface. Through the multi-layer restricted boltzmann machine(RBM) stacking structure, the potential multimodal key features are automatically extracted from the visual area to realize the high-risk identification of the operation area of the distribution network. The experimental results show that the proposed method verifies its high efficiency in complex environments by clearly showing the risk distribution in the operation area and significantly shortening the time-consuming of high-risk identification.