Smart grid intrusion detection based on machine learning

• Main Author: Sun, Yihan
• Other Authors: Xiao Gaoxi
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering::Wireless communication systems
• Online Access: https://hdl.handle.net/10356/155663
• Institution: Nanyang Technological University

In recent years, with the development of smart grid technology and the increas- ing awareness of network security, the application and security of smart grid have become more important. In order to solve the security problems of smart grid in Home Area Network, Neighborhood Area Network and Wide Area net- work, methods based on intrusion prediction and detection, using different cal- culation methods and networks for different topologies have been gradually pro- posed. This report uses an approach based on network intrusion detection to address the security problems. This report combines intrusion detection methods with deep learning and pro- poses an Adaptive Deep Learning algorithm based on the size of the network. The algorithm obtains the number of layers for deep learning and the number of neurons per layer by determining the characteristic dimension of the network traffic. By adjusting the parameters, the migration capability of the network can be improved so that it can extract the original data dimensions and obtain new abstract features. By learning the newly generated abstract features, this report sets different parameters depending on the scope of the smart grid. By abstract- ing the network features, the algorithm can generate new features with a higher dimension. By combining deep learning models with traditional machine learn- ing models, the classification of network traffic data is significantly improved. The experiments use the KDD99 dataset to evaluate the usability of the intrusion detection model. From the experimental results, it can be seen that the algorithm used in this report improves the effectiveness of intrusion detection and reduces the training time.