The optimization of cyclic scheduling for an ethylene cracking furnace system (ECFS) is beneficial for ethylene plants. The scheduling problem involves multiple feeds, tanks, furnaces, and different operating periods. Moreover, the operations of ECFS are subjected to several constraints, such as tank capacity, outlet temperature of transfer line exchangers (TLEOT), and non-simultaneous decoking. These problems bring challenges to the cyclic scheduling for ECFS. By considering tank capacity constraints, this paper proposes a data-driven modeling and cyclic scheduling framework to address this issue. Fuel consumption, generation of superhigh-pressure steam (SS), and dilution steam (DS) consumption are considered in the scheduling model. Case studies from the literature and an actual ethylene plant are conducted to determine the effectiveness of the proposed method. Cracking furnace simulation software is used to generate the data of key products such as ethylene, propylene, and benzene to develop data-driven yield models. Industrial data are employed to establish the regression models of fuel consumption, SS production, and TLEOT. The results of two case studies indicate that the proposed method can manage inventory constraints in cyclic scheduling and achieve feasible solutions.