Water and aqueous solutions are the most important materials on Earth. Their liquid structures are defined by a hydrogen(H)-bond network whose organization gives rise to the unique properties. The arrangement of water molecules can be probed by scattering experiments, such as neutron diffraction. But most currently available experimental instruments typically yield spatially and temporally averaged structural information. Complementary to the scattering experiments, the optical spectroscopy experiments such as X-ray absorption spectroscopy (XAS) provide important electronic structural information of the H-bond network, based on which its interaction with solvated ions can be inferred indirectly. In order to clearly understand these experiments, theoretical modeling is required both at electronic ground state and excitation. However, in water, highly accurate theories are often demanded in order to model the delicate H-bonding. To address the above challenges, we have developed advanced ab initio molecular dynamics method, machine learning techniques, and GW-based Bethe-Salpeter method.  By employing these approaches, we have theoretically studied the excitonic effect and the molecular structure in both water and aqueous solutions. Our results resolved the controversy concerning the fundamental structure of water and salt water, which have been under intense debate over the last twenty years.