Transient Characteristics Selection And Stability Evaluation Of New Energy Grid-Connected Systems

With the proposed dual-carbon target and the continuous development of power grid intelligence technology,machine learning is gradually applied to the transient stability evaluation of new energy power systems.Transient stability evaluation technology faces the problems of model structure selection,model online updating and evaluation performance optimization.This paper establishes a two-stage transient stability evaluation model based on feature selection and ensemble learning.Taking IEEE medium system and large system with a high proportion of wind power access as examples,this paper studies wind turbine characteristics and the influence of artificial intelligence algorithm model structure on transient stability evaluation,and proposes a transient stability evaluation method to improve the accuracy,generalization and speed of the original model.The transient stability evaluation model can be updated online by keeping the synchronization between the transient stability model and the power network topology.The main contributions of the paper are as follows:(1)The transient stability mechanism and feature analysis method of the power system with wind power are studied.First,the mathematical model of the wind turbine is analyzed,the transient stability feature set of the power system with wind power is summarized,and the statistical information coefficient which can be used for correlation analysis and redundancy removal of the grid features is proposed.The correlation between the feature and the outcome was evaluated.Improve power grid data utilization efficiency and model accuracy.(2)Study the principles of Wasserstein’s generative adversarial network,transfer learning and integrated learning,and build simulation models.In this paper,Wasserstein generated adversative network algorithm to extend the transient stability evaluation of small sample data sets or unbalanced sample data sets is proposed.Bagging integrated learning method is used to synthesize the final confidence of each machine learning algorithm to obtain the transient stability evaluation results after integrated learning.After the abrupt change of the system structure,the data set of the original topology structure is migrated and reconstructed according to the new system topology structure to realize the online update of the transient stability model.(3)According to the system data released by IEEE,the improved 39-node and 118-node models were built in the simulation software PSASP with different proportions of new energy access.The influence modes of new energy access on the transient stability of the system were studied,and the short-circuit states of various lines under different operating modes were simulated in the simulation system to form the original data set of model input.(4)A two-stage transient stability evaluation model driven jointly by data and knowledge is proposed.By finely dividing the confidence degree of transient stability assessment,the prediction results in the AI model are divided into safe domain samples and non-safe domain samples,and the accuracy,speed and generalization of the original model are improved through the second stage evaluation of the knowledge model.