The “Stuxnet” event and the Ukraine blockout event caused by cyberattacks have shown that it is very important for power systems to implement proactive cyber defense. With the construction of energy internet, the terminal-edge accessing proactive defense technology of energy internet has drawn lots of attentions. This paper constructed the proactive defense model of energy internet terminals edge-access using the network topology autoassociation, and established the multi-path and multi-parameter stochastic communication path construction algorithm in energy internet terminals edge-access network. To solve the problem of the transmission reliability and defense revenue decline caused by the dynamic adjustment of the association cycle, we presented a novel algorithm based on non-extensive entropy and Renyi cross entropy, which measures the abnormal of real-time network with multi-feature. On this basis, the proactive adjustment strategy of network topology autoassociation communication paths and the dynamic threshold value judgment method were designed to improve the accuracy of network anomaly measuring. Moreover, this paper also proposed a cyber-security reliability prediction algorithm based on Hidden Markov prediction model, and presented the guide paths space proactive adjustment strategy and associated cycle stretching strategy based on the network topology autoasssociation diagram of reliability prediction, guaranteeing the service quality of terminal edge access network. Finally, the model was applied to carry out experimental analysis on DoS attack, following attack, semi-blind attack defense, and network communication service quality. Theoretical analysis and simulation results show that the applicability and accuracy of the proposed model and the adjustment strategy are valid, which can support the proactive defense of energy internet terminal edge-access.