Aiming at the problems of the current Digital Substation Power Industry Control system’s complex number of cyberspace assets, different specialties, difficult identification and management, low asset safety detection efficiency, and poor accuracy of adaptive blocking control of security risks, research on the security protection of cyberspace assets of Digital Substation Power Industry Control System is carried out. This paper studies the systematic asset portrait technology based on multi-source network security features, establishes a deep autoencoder of multi-source data, extracts and compresses the dimensionality reduction depth features. Then, this paper studies the risk detection of network asset intelligence, and studies the risk identification model training method based on semi-supervised learning reconstruction error, and constructs an integrated learning framework for accurate and stable network asset risk intelligent detection, so as to realize the network asset intelligence risk detection for high-risk ports, vulnerabilities, abnormal communication behaviors and other risks.