Now, the use of deep learning technology to solve the problems of the low multiclassification task detection accuracy and complex feature engineering existing in traditional intrusion detection technology has become a research hotspot. In all kinds of deep learning, recurrent neural networks (RNN) are very important. The RNN processes 41 feature attributes and maps them to a 122-dimensional high-dimensional feature space. To detect multiclassification tasks, this study proposes an intrusion detection method based on fully connected recurrent neural networks and compares its performance with previous machine learning methods on benchmark datasets. The research results show that the intrusion detection system (IDS) model based on fully connected recurrent neural network is very suitable for classification of intrusion detection. Classification methods, especially in multiclassification tasks, have high detection accuracy, significantly improve the detection performance of detection attacks and DoS attacks, and it provides a new research direction for the future attempts of intrusion detection methods for industrial control systems.

The smart grid integrates the computer network with the traditional power system and realizes the intelligentization of the power grid. The Advanced Measurement System (AMS) interconnects the power system with the user, realizes the two-way interaction of data and information between the power supplier and the user, and promotes the development of the smart grid. Therefore, the safe operation of AMS is the key to the development of the smart grid. As smart grids and computer networks become more and more closely connected, the number of cyberattacks on AMS continues to increase. Currently, AMS intrusion detection algorithms based on machine learning are constantly being proposed. Machine learning algorithms have better learning and classification capabilities for small sample data, but when faced with a large amount of high-dimensional data information, the learning ability of machine learning algorithms is reduced, and the generalization ability is reduced. To enhance the AMS intrusion detection algorithm, this paper uses a Generalized Regression Neural Network (GRNN) to identify attack behaviors. GRNN has strong non-linear mapping ability, is suitable for unstable data processing with small data characteristics, has good classification and prediction ability, and has been widely used in power grid systems. Aiming at the existing problems, this paper proposes an upgraded generalized regression neural network AMS intrusion detection method DBN-DOA-GRNN. Based on the feature extraction and dimensionality reduction of the data by DBN, GRNN is used for data with less feature information in learning classification. In addition, to improve the detection effect of the method, the Drosophila Optimization Algorithm (DOA) is used to optimize the parameters of GRNN to reduce the influence of random parameters on the detection results, improve the detection accuracy of this method on small-scale sample data, and thereby improve the detection performance of the AMS intrusion detection algorithm. The proposed method archives an accuracy of 87.61%, 3.10% false alarm rate, and 96.9 precision rate.