An interference prediction scheme is proposed for power control in packet-switched TDMA wireless networks. The prediction scheme is based on the interacting multiple model (IMM) estimator, and it is effective to a wide range of nonstationary dynamic characteristics of the interference power. Numerical experiments show that, compared with a scheme based on a Kalman filter, the IMM estimator-based scheme predicts the interference power more accurately and allows us to adjust the transmit power more efficiently in achieving a desired level of signal-to-interference-plus-noise ratio (SINR).