Prediction Methods for Energy Internet Security Situation Based on Hybrid Neural Network

Abstract

Under the duplex framework of the rapid transformation of the conventional power grid to the energy Internet and the swift development of network attack technologies, the security and protection of the energy Internet are becoming increasingly precarious. The energy Internet attack prediction method has become a current research hotspot. The current research in this area is predominantly focused on security to protect and reduce losses, it is impossible to counter the attacks faced by the Energy Internet promptly, and there is a lack of online predictive security analysis methods and tools. Aiming at the shortcomings of traditional research and the current stage, this paper proposes an online prediction and analysis method of energy Internet security situation based on the combination of temporal convolutional neural network (TCN) and gated recurrent neural network (LSTM). This method which based on the security status data set can predict and analyse the region Energy Internet security, and the mean absolute error and mean squared error are 0.352 and 0.2714 respectively. The results show that the prediction performance of the algorithm has been improved, the stability is robust, and it can be applied to the energy Internet security situation prediction.