Power grid bad-data detection and identification based on online kernel learning method

The accuracy of remote measurements and network parameters is the basis for a variety of power system analysis in the smart grid technology support system. Considering multi-period RTU/PMU information and complex power balance equations,an online kernel learning-based bad-data detection and parameter identification method is proposed. First,the information of complex power and node voltage amplitude is used to define the indices of complex power balance. Second,the Bayesian data analysis-based extended Kalman filter method is described to detect the bad-data dynamically. Third,locally weighted projection regression method is adopted to identify the network parameters. The calculation results of different scale power system show that this algorithm is feasible and has a high computing accuracy. It provides a viable approach for the advanced application research of smart grid technology support system.