An online robust adaptive system identification method is proposed for ships under measurement noise and environmental disturbances, addressing model identification without persistent excitation. Using a double-integrator filter-based composite update law, it achieves real-time estimation of all parameters, states, and initial states conditions with adaptability to parameter variations. An extended state observer estimates total system error as an event-triggered signal to save resources and monitor system accuracy. Under initial excitation conditions, the method ensures globally uniformly ultimately bounded stability of estimation errors. Simulations show superior identification speed and and accuracy over existing algorithms.