A Machine Learning based Knowledge Graph Framework for Heterogeneous Power Grid Systems

Knowledge graph system is becoming one of the acknowledged approaches for executing the transactional works of power grid applications. However, due to the heterogeneous data structure of recent power grid systems in the Big data environments, traditional machine learning (ML) based knowledge management approaches show limited adoptability. Hence, the paper proposed a comprehensive ML based knowledge graph framework for complex and heterogeneous power grid applications. The framework mainly involves two ML approaches which are a concept drift detection based incremental learning for homogeneous data and a reverse verification based transfer learning among heterogeneous data. The proposed framework is tested by multiple heterogeneous databases collected from power grid systems of state grid Anhui electric power research institute (China), and the results show improved practicability compare to traditional ML methods.