Conventional simulation methods relying on single models or empirical formulas fail to support intelligent, full-lifecycle simulation and verification for complex systems such as marine power systems. Multi-domain modeling languages like Modelica, while powerful, remain at the tool level and cannot resolve bottlenecks in data fusion, high-confidence simulation, or closed-loop verification. To overcome these limitations, this paper proposes a system simulation methodology based on data, models, and mock-ups. The methodology integrates fulllifecycle multi-source data into simulation, employs simulation models to construct a digital mock-up, and achieves tight integration and closed-loop iteration among data, models, and the physical mock-up among data, models, and the physical mock-up, thereby establishing high-confidence system-level simulation across design, testing, service, and maintenance. A digital twin of a marine power system is developed using Modelica with real-time multi-dimensional data interaction. Experimental validation under four typical operating conditions (shutdown, 0kW, 30kW, 60kW) shows that simulated key parameters (steam pressure, turbine speed, output power) agree closely with measured values, with all relative errors controlled within 8%. These results demonstrate the feasibility, practicality, and engineering applicability of the proposed methodology, confirming that integrating data, models, and mock-ups effectively supports intelligent full-lifecycle operation and maintenance of complex systems and provides an implementable pathway for the intelligent transformation of systems engineering.