Marine forecasting is essential for human activities at sea. Marine forecasting system technology supports modern marine meteorological services. Advances in oceanic observation, data assimilation, numerical simulation and high-performance computing also drive the development of marine operational forecasting. The present study uses the Weather Research and Forecasting (WRF) atmospheric model, the Coastal and Regional Ocean Community (CROCO) model and the Simulating Wave Nearshore (SWAN) model to develop a multi-model high-resolution offline coupled forecasting system over China's marginal seas, for offshore wind power forecast and marine disaster warning. This forecasting system integrates a series of ocean model parameterization schemes, including wave mixing parameterization, mesoscale parameterization, and seamount- and eddy-induced parameterization developed by the Oceanic Modeling and Observation Laboratory of the Nanjing University of Information Science and Technology. Furthermore, observation data are assimilated using data assimilation and artificial intelligence technology in this system. The comparison and analysis of forecast and observation results show that the forecast system can accurately simulate the marine meteorological elements such as surface wind, ocean surface current, sea surface temperature, wave and tide. At the same time, on-demand real-time visual panorama display is realized.