Research On The Online Stability Analysis In Power Grid Based On Machine Learning

The secure and stable operation of power grid has important and long-term significance for the power industry which is developing more and more rapidly.How to evaluate the security and stability of power grid quickly,accurately and efficiently is a frontier subject in the development of electric field.With the gradual popularization of big data and machine learning algorithms,the analysis and application of power grid operation data is also a difficult problem facing the power grid.In view of the above three points,the online stability of power grid is analyzed and studied in this paper based on machine learning.This work mainly explores the static voltage stability and interval oscillation stability,and completes the following two schemes for the secure and stable assessment of the power grid:Aiming at the static voltage stability problem of the power grid,an online static voltage stability assessment scheme based on correlation detection and iterative random forest(IRF)is proposed.Part mutual information(PMI)and Pearson’s correlation coefficient(PCC)are used to detect the correlation between power grid operation variables,and the key features with strong correlation can be screened out.By using the key features and the corresponding voltage stability margin(VSM),the training of the prediction model based on IRF can be realized.Once the trained model receives the real-time operation information of the system,it will quickly give the corresponding evaluation results.In addition,a model updating measure is designed to deal with the possible changes in the operating conditions of the power grid.The proposed scheme is validated on the IEEE 39-bus system and the 1648-bus system,and its excellent prediction performance and fast online assessment efficiency are verified.Aiming at the problem of inter-area oscillatory stability for the power grid,this paper proposes an online inter-area oscillatory stability assessment scheme which takes into account the feature redundancy.Taking into account the redundant data in a large scale of operation data in a large power grid,a feature redundancy processing method is designed to eliminate the redundant and irrelevant features in the data set.Then PMI is used to further reduce the dimensionality of the data set,and key variables strongly related to oscillatory stability margin(OSM)are screened out to train IRF,and mapping relationship is constructed.Finally,the OSM prediction is completed according to the real-time PMU measurement of selected features received.The proposed scheme is validated on the IEEE 39-bus system and the1648-bus system,the test results show a good prediction performance.In addition,in order to further highlight the innovation of the research scheme in this paper,the following work is also completed: Taking into account the possible influence factors of the power grid operation to analyze and test;the assessment model based on IRF was compared with the conventional methods.