A Transient Stability Assessment Method For Power Systems Based On Maximum Mutual Information Feature Identification

In order to achieve the goal of carbon neutrality and carbon peaking,the development direction of the power system structure began to change from traditional energy to clean and sustainable energy.With the large-scale integration of new energy into the power grid,the operation of the power system becomes more and more complex,and once the power system is subject to large-scale disturbance,it will lead to equipment and line failures,and even cause the power system to be paralyzed,affecting people’s production and life,and the transient stability assessment of the power system can effectively predict the operation of the power system after the disturbance,and take timely measures to ensure the stable operation of the power system and avoid the adverse consequences caused by the power system failure.Therefore,how to improve the accuracy and computational speed of transient stability assessment has become a popular problem in the field of transient stability assessment.However,the existing research mainly focuses on the optimization of transient stability assessment classifiers,ignoring the role of feature identification in improving the accuracy and computational speed of transient stability assessment.To this end,this paper shifts the focus of research to the features affecting transient stability,and proposes a transient stability assessment method based on maximum mutual information feature identification by identifying the features affecting transient stability,screening out the features with high impact on transient stability and eliminating redundant features,which provides a new tool for rapid and accurate assessment of transient stability of power systems in the context of new energy grid connection.The key technical issues in the transient stability assessment problem are studied in terms of stability criterion and model evaluation index feature selection.For the characteristics of voltage stability and power angle stability,the transient stability criterion based on power angle is proposed;for the problem of unbalanced samples in the transient stability sample set,the model evaluation index applicable to the unbalanced classification problem is proposed;for the transient stability sample characteristics.a three-stage fault information representation method is proposed to filter out 24-dimensional features.which lays a theoretical foundation for the transient stability assessment later.Secondly.for the problem of high dimensionality and large scale of raw data in transient stability assessment,the raw data are clustered,and a clustering method based on mutual information is proposed to improve the traditional K-means clustering algorithm by using K’dist curve plots and DK analysis plots to determine K values and introducing mutual information into clustering,and the simulation data of artificial data set,WSCC 9-node system and IEEE 39-node standard system are selected to verify the clustering advantages of the proposed algorithm and provide a data basis for feature identification and transient stability assessment.To address the problem that the previous transient stability assessment mostly focuses on improving the classifier performance and ignores the impact of features,a transient stability assessment method based on maximum mutual information feature identification is proposed,introducing the entropy weighting method to objectively assign the maximum information coefficient and the correlation-based fast feature selection algorithm to maximize the advantages of both algorithms,and modifying the IEEE 39-node system to include wind farms to simulate The new energy grid connection is verified that the algorithm proposed in this paper is also applicable to the new energy grid connection scenario.