In the online and tracking simulation context, precise model initialization is highly important to ensure the quality of predictions made based on simulation. This paper presents a hybrid approach for the initialization of tracking simulation systems that combines different techniques and utilizes the persistent communication between model and process in order to initialize the simulation model. In the proposed approach, the model initialization is performed throughout different stages. A model is first fed with initial conditions that are close to the physical plant's state. Then, the model is run to the same state as the physical system using a process control system. Possible mass or energy errors in the state of the simulation model are corrected before the actual tracking process is started. This paper first introduces previously studied initialization techniques for online simulation systems. Then the proposed hybrid initialization method is elaborated. Finally, its implementation is demonstrated using a simple but representative process system.