Research On Rapid Assessment Of Transient Stability Of AC/DC Hybrid System Based On Artificial Intelligence

Due to the needs of national development and the improvement of people’s living standards,the traditional power system is undergoing great changes,the feature and characteristic of modern power system is becoming increasingly complex and flexible,the AC-DC system is highly coupled,mutual influence,and the system running state is changeable.In the future,transient stability of power grid has become one of the key research problems in power system direction.Under the background of the new era,The transient stability of the power system is becoming very important,modern power system is developing and improving,the shortcomings of traditional power system transient stability evaluation method are gradually exposed.When solving differential equations of large systems,the time domain simulation method has too much calculation,which directly affects the speed of solution and is not suitable for the rapid safety evaluation of power system in the future;Although the direct method is not limited by the mathematical model of the system,how to construct the energy function of the power system is still a difficult problem.With the rapid development of artificial intelligence technology,all kinds of new methods,such as data mining,machine learning,deep learning and so on,have gradually entered the people’s vision,which has brought a new research direction to the problem of power system transient stability evaluation under the background of the new era.In this paper,data mining technology is combined with machine learning and deep learning methods to be applied to the future power system transient stability evaluation.The main contents of this paper are as follows:(1)A set of characteristic data sets for transient stability assessment of AC / DC power systems is constructed.PSD-BPA simulation software is used to build AC /DC power system model and time domain simulation,change the fault location of the power system,the duration of the fault,and the load level of the system,obtain the actual operating data of the system in different operating states,and combine the existing Relevant literature,calculate the transient characteristic quantity,construct the transient data sample set.(2)A method for transient stability assessment of AC and DC systems based on the combination of data mining technology and support vector machines is proposed.First,the T-distributed stochastic neighbor embedding algorithm is used to perform feature reduction processing on the transient sample data set.The feature visualization effect of the TSNE algorithm is better,which can effectively improve the evaluation accuracy of the system,Then use cross-validation and grid parameter optimization to optimize the parameters of kernel function of the SVM,and input the sample features processed by the TSNE algorithm into the SVM evaluation model for training and testing evaluation model.Finally,the research results show that this method has high accuracy and robustness than others.(3)Aiming at the problem of insufficient mapping capabilities of the machine learning model to large samples of complex data,a transient stability evaluation method based on depth neural network(DNN)for AC / DC hybrid power system is proposed.DNN model contains multi-layer hidden layers,which has more powerful feature learning and feature expression ability,and describes the feature of sample data in more detail.Compared with other machine learning evaluation models,the DNN model can has high evaluation accuracy and good robustness.