In this paper

we present the learning-based data analytics moving towards transparent power grids and provide some possible extensions including machine learning

big data analytics

and knowledge transferring. The closed loops of data and knowledge are illustrated and the challenges for establishing the closed loops are discussed. General ideas and recent developments in supervised learning

unsupervised learning

and reinforcement learning are presented together with extensions for power system applications. Furthermore

much emphasis is placed on privacy-preserving data analysis

transfer of knowledge

machine learning for causal inference

scalability and flexibility of data analytics

and efficiency and reliability of computation. Existing integrated solutions in the industry featuring the Industrial Internet and the digital grid are also introduced.