Transient Stability Assessment Of Power System Considering Category Imbalance

The increasing interconnection scale of the modern power grid,the continuous access to various types of power electronic equipment and large-scale renewable energy sources have caused many uncertain factors to affect the operating state of the power system,and its safety,complexity and The contradiction between the three of the rods also makes its dynamic characteristics more complicated,so a higher standard of system safety and stability analysis and control is required.In terms of the stability and safety of system operation,it is very important to put forward effective preventive control measures and make accurate and rapid predictions of the transient stability of the system.This is because the important reason for the large-scale power outage of the system is that the transient instability of the power system.Recent years,with the large-scale application of wide-area measurement system and the rapid development of computer technology,the prediction of transient stability of power systems based on artificial intelligence technology has attracted widespread attention of scholars from various countries.Generally,the prediction of transient stability based on artificial intelligence is treated as a two-mode classification problem,that is,it is divided into two types: stable and unstable.The traditional power system transient stability assessment method does not consider the unbalance of classification results,so it will seriously affect the final classification results.From the perspective of considering the unbalance of classification results,in this paper,a transient stability evaluation method based on the integrated limit learning machine is proposed.The main content of the thesis research is as follows:(1)Firstly,the unbalance of classification results of power system transient stability assessment is described,and the impact of category imbalance on the transient stability evaluation results is analyzed based on loss function.According to confusion matrix,the performance indexes for evaluating the prediction results are listed.In view of the imbalance of classification results,this paper puts forward the criteria that the evaluation indexes should meet.In addition,the traditional evaluation indexes and the G-mean evaluation indexes used in this paper are compared and analyzed.(2)Take New England 10 machines and 39 nodes as the test system,adopt the transient stability time domain simulation software Power System Tool 3.0 package,by setting different initial conditions and different fault conditions to generate different sample data.And obtain the required classified initial sample set according to the construction of the original feature set.In order to eliminate redundant features and irrelevant features to reduce classification difficulty,on the basis of analyzing and comparing the advantages and disadvantages of different feature selection methods,this paper proposes a hybrid feature selection method combining Relief and sequence forward search(SFS).(3)In order to solve the imbalance of the prediction results of power system transient stability,considering from the sampling point of view,and based on the extreme learning machine,an integrated learning method for transient stability evaluation of power system is proposed.The validity and accuracy of the proposed method are verified on a test system.