FS-LSSVM Based Fake Measurement Detection in Substation Automation System Using IEC 61850

With the prevalence of IEC 61850 protocol, Substation Automation System migrates to implement data acquisition with Electronic Instrument Transformer (EIT) and transmit measurement via Ethernet. Mal-function of EIT, cybernetic attack, as well as unspecified defects may lead to fake measurements deviated from system state. Thereafter, protection system may mis-operate due to these fake measurements. Pattern identification based approach has been utilized to detect fake measurement. In order to achieve satisfactory performance, Support Vector Machine (SVM) is investigated in the paper. Fixed Size Least Squares Support Vector Machine (FS-LSSVM) is utilized to optimize the network structure to mitigate contradictory requirements of high pattern classification precision and low time delay imposed by protection systems. Comparative study of FS-LSSVM and Othogonal Least Squared (OLS) algorithm Radial Basis Function Neural Network (RBFNN) show both approaches can achieve satisfactory performance within limited time while FS-LSSVM is more promising for its better generalization abilities.