Open AccessRETRACTED: Stability analysis of distributed smart grid based on machine learning
Author(s) -
Cheng-Han Li
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/692/2/022125
Subject(s) - support vector machine , computer science , stability (learning theory) , metric (unit) , smart grid , decision tree , radial basis function kernel , kernel (algebra) , electric power system , data mining , transformation (genetics) , artificial intelligence , machine learning , performance metric , power (physics) , kernel method , engineering , mathematics , operations management , biochemistry , quantum mechanics , physics , combinatorics , electrical engineering , gene , management , economics , chemistry
Decentralized Smart Grid (DSG) is a new technology proposed to power networks with elastic nodes. It can realize dynamic electricity price demand response without large-scale transformation of the infrastructure.In order to analyze the system stability of DSG, six representative machine learning classification models were applied to analyze the stability data of 10,000 samples of the 4-node system. Combined with the requirements of power system security, stability and economic performance, the effect of each classification model on the stability prediction of DSG system was tested.The test results showed that the model with a Gaussian kernel basis function kernel support Vector machine (RBF SVM) was suitable for data analysis, with the accuracy up to 97.10% and the F metric up to 0.977.CART decision tree model is suitable for real-time forecasting of power system. Under the requirement of ensuring real-time forecasting, its accuracy can reach 84.90% and F metric can reach 0.882. Its modeling and prediction calculation requirements are only 0.98% and 1.59% of that of RBF SVM model respectively.