This paper presents a linear rolling predictive con- trol method applied to eliminate the potential overshoots in superconducting magnetic energy storage (SMES) applications. A bridge-type chopper is introduced to form a four-quadrant coil operation condition. The linear extrapolation method is adopted to predict the upcoming data of the controlled variable after an inevitable system time delay in a practical digital control system based on the current sampled data and previous sampled data. When the predicted data reaches or exceeds the reference range, the digital control system will output the corresponding state switching signals immediately. Two- and three-point linear rolling predictive control methods are discussed in two cases of constant- speed energy charge and constant-voltage energy discharge processes, and compared with the conventional digital control methods without prediction. The results show that the proposed method saves signal processing and computing time effectively, and also provides a very high control accuracy.