The scheduling of decoking operations in a naphtha cracking furnace system is an important issue to ethylene producers, because excessive coke deposits inside the furnace coils can negatively impact plant safety and productivity. For optimal scheduling, accurate online estimation of the thickness of the deposited coke is essential. In practice, the coke thickness can be estimated from various furnace operating Variables, but measurement errors and unexpected changes in the coke growth rate cause significant uncertainties in the estimation. Errors in the coke thickness estimate manifest themselves as gaps between the model prediction and actual measured values of key operating variables such as the pressure drop and the tube temperature. To handle the potential conflicts in an established schedule caused by the uncertainties, we propose to use a "proactive" scheduling strategy. In "reactive" scheduling, rescheduling is performed whenever an unexpected operational problem causes an unscheduled decoking operation, thus making a standing schedule no longer viable. On the other hand, in the "proactive" scheduling strategy, model information, as well as measurement information, are used to determine appropriate rescheduling points before actual operational problems arise. Under the proposed proactive scheduling strategy, the model predictions of the pressure drop and the tube temperature are compared against their measurements while the plant is operating according to a current decoking schedule. Whenever the gap between the model prediction and the measurement is larger than a given threshold value, the model is updated based on the measurements and the scheduling problem is solved again with the updated model information. The new scheduling solution is applied to the operation until the next scheduling point is found. This proactive scheduling procedure is applied to a simulated system of multiple furnaces. The advantages of the proactive scheduling strategy, in terms of productivity and risk management, are shown by comparing it with a reactive scheduling strategy and a heuristic decoking strategy over a large number of scenarios.