As it was manifested, the implementation of accurate SE and making robust decision for power systems was a worthwhile and in-depth investigation. Leveraging the advancements in artificial intelligence technology, this monograph introduced an innovative framework for monitoring and controlling power grids. This framework adeptly integrated, in its initial stage, unique NNs specifically designed for interpreting power grid topology and predicting future states. Additionally, the principles of composite optimization theory were incorporated; it simplified the complexity inherent in the estimation of power system states, facilitating rapid and accurate assessments of system conditions. Moreover, this framework took a step further by intricately interweaving a data-driven approach with physics-based optimization, thereby unleashing the full potential of system flexibility. Ultimately, by employing NNs to parameterize the relationship between system states and control policies, that framework excelled in providing precise and targeted robust control strategies when confronted with newly estimated system states. The numerical experiments in each chapter validated our theoretical findings and demonstrated the effectiveness, robustness, and computational efficiency of our approach.