Research On Intrusion Detection Model Of Smart Grid Network Based On Machine Learning Algorithms

The smart grid architecture is the deep integration of information technology and power system.While bringing many convenience to the people,the highly open communication network and complex information interaction environment also make the smart grid architecture face more security risks.As an active defense technology after firewall technology,network intrusion detection technology can not only actively defend against attacks from outside the system,but also identify abnormal data inside the network compared with other security measures of power system.With the increasing trend of automation and scale of network attacks against the smart grid,the traditional network intrusion detection technology is gradually unable to meet the needs of intrusion detection in the smart grid.Applying machine learning technology to the intrusion detection field can greatly improve the detection efficiency and save computing resources.Therefore,this paper focuses on the application of machine learning technology in smart grid network intrusion detection.The main research contents are as follows:(1)Aiming at the problems of low detection efficiency and poor accuracy of traditional network intrusion detection models,this paper first uses genetic algorithm to optimize the random forest model and obtain the optimal model parameters.After that,Smote oversampling algorithm is used to fill the data of categories with less training samples in the data set,and K-means clustering algorithm is used to delete outliers from the filled data and then the processed data is used for model training.The optimized random forest network intrusion detection model combined with oversampling algorithm is simulated on CIC-IDS-2017 data set,and the experimental results prove the superiority of the model.(2)In order to detect unknown attacks in smart grid networks,a network intrusion detection model based on pearson product-moment correlation coefficient and open convolution neural networks is applied in smart grid network intrusion detection in this paper.This model eliminates redundant features by calculating Pearson correlation coefficient between feature columns,and adds Open Max layer to CNN model to identify unknown attacks.Experiments on KDDCUP99 and CIC-IDS-2017 data sets show that the model can ensure the overall detection accuracy and detect unknown attacks.(3)A false data injection attack detection model based on generative adversarial network and convolutional neural networks is used to detect the system operation state and determine whether the system is under false data injection attack by identifying the abnormal state of the system.The model builds CNN model according to the power grid structure,and uses GAN model to generate part of abnormal data,which can solve the problem of data imbalance and optimize the training effect of CNN model.The experimental results on IEEE-14 bus standard test system and IEEE-39 bus standard test system verify the effectiveness of this model.