This paper focuses on the day-ahead allocation of operation reserve considering wind power prediction error and network transmission constraints in a composite power system. A two-level model that solves the allocation problem is presented. The upper model allocates operation reserve among subsystems from the economic point of view. In the upper model, transmission constraints of tielines are formulated to represent limited reserve support from the neighboring system due to wind power fluctuation. The lower model evaluates the system on the reserve schedule from the reliability point of view. In the lower model, the reliability evaluation of composite power system is performed by using Monte Carlo simulation in a multi-area system. Wind power prediction errors and tieline constraints are incorporated. The reserve requirements in the upper model are iteratively adjusted by the resulting reliability indices from the lower model. Thus, the reserve allocation is gradually optimized until the system achieves the balance between reliability and economy. A modified two-area reliability test system (RTS) is analyzed to demonstrate the validity of the method.