Research On Key Technologies Of Intrusion Detection For Smart Grid

In recent years,the development of economy and technology has promoted the innovation of the power system,and the traditional grid architecture has been unable to meet the needs of users and society.Based on this,the concept of smart grid was proposed.The combination of computer technology and the power grid has brought some conveniences,but at the same time,the openness of the communication network and the complex information interaction environment also make the smart grid face more risks.Among them,intrusion detection technology,as a new generation of security defense technology,can not only resist external attacks,but also identify abnormal data inside the network.With the continuous progress of artificial intelligence algorithms,many scholars have applied artificial intelligence algorithms to intrusion detection technology and achieved good results.Among them,the gene expression programming algorithm detects the attack behavior by mining the relationship between the characteristic attributes of the power grid data and the behavior attributes,but the algorithm is easy to fall into the local optimum;the convolutional neural network can obtain the learning model by training the historical data of the power grid,and,judges whether the data is abnormal data,but the training process is easy to overfit and is easily affected by parameters.In order to solve the above problems,this paper proposes a smart grid intrusion detection algorithm based on feature selection and mixed parameter optimization GEP(ID-MGEPS)and an intrusion detection algorithm based on GEP optimized convolutional neural network(IDSBGEP-CNN).Experiments have verified the effectiveness of the model,and the model proposed in this paper is superior to several comparative detection algorithms in detection accuracy.The details of this article are as follows:(1)This paper proposes an intrusion detection technology based on Pearson’s correlation coefficient feature selection.Using the Pearson coefficient-based feature selection method(SVIFSPCC)can effectively remove redundant feature attributes in the intrusion data and reduce the dimensionality of the intrusion data set.At the same time,the support vector machine(SVM)classifier is used to detect and classify the intrusion data after feature selection.Through experimental comparison,it can be seen that the SVIFS-PCC method compares mutual information,random forest,variance filtering and other feature selection methods have higher approximations.The simplicity,less time loss and higher intrusion detection accuracy verify the effectiveness of the method.(2)This paper proposes an intrusion detection technology based on mixed parameter optimization GEP,which mainly includes a new decoding method to improve the decoding efficiency of the algorithm for the low decoding efficiency of the algorithm,and the problem of easy to fall into the local optimum in the process of algorithm genetic evolution.A population optimization strategy is proposed to ensure the diversity of the population and improve the local convergence defects of the algorithm.And use the improved GEP algorithm to construct an intrusion detection model to detect and analyze the intrusion data.Finally,the experiment compared with other intrusion detection algorithms shows that the algorithm proposed in this paper has higher detection accuracy and less time loss.(3)This paper proposes a combined intrusion detection technology based on GEP optimized convolutional neural network,which uses the global optimization capability of the GEP algorithm to optimize the network structure and parameters of CNN to improve the learning and generalization capabilities of CNN.And the optimized convolutional neural network is applied to intrusion detection.Through experimental comparison,it can be known that the optimized CNN has a higher accuracy rate than other networks.