Wind power provokes increasing concerns on its impacts on a power system as many wind power generators have been integrated to a power system. One of the concerns is a challenge to keep supply-and-demand balance in a power system.Battery energy storage systems (BESS) installed at a wind farm are considered as a promising option for the challenge. Two types of operation strategies were proposed for its operation: fluctuation mitigation and scheduled output types. On the other hand, in order to constitute an economical BESS, selecting battery size as small as possible is vital. However, for the scheduled output operation, no study has been published both on the required battery size and on a methodology to examine the size.This paper proposes a methodology to evaluate the required kWh capacity of battery for the scheduled output operation using accumulated prediction error of a wind farm. The methodology evaluates a standard deviation of accumulated prediction error with autocovariance functions of prediction error. Based upon the statistical characteristics obtained for actual prediction error of wind power, the paper presents results of a sample study to show the effectiveness of the methodology.