In the deep space exploration, the signal is delayed by the troposphere and a certain delay will occur at the receiver end, which affects the detection accuracy. The existing methods mainly realize the delay prediction by the grid model and the space model, but the accuracy of the model is limited due to the regional difference, and the prediction accuracy still has room for improvement. In this paper, a deep-space tropospheric delay prediction model based on adaptive multiple-input multiple-output (MIMO) signals was proposed. The satellite signal MIMO transmission mode was simulated based on a single transceiver antenna, and then an adaptive Kalman filter was constructed. The optimal transmission path was selected by adaptively adjusting the weight coefficient of the MIMO signal component to predict the tropospheric delay. The number of satellites participating in the measurement was four. Experiments were carried out under different signal-to-noise ratios and changing the number of MIMO channels to study the accuracy and actual measurement error of the adaptive MIMO model. The experimental results show that the prediction accuracy of the new method is much higher than that of the GPT2 model, GPT2w model and the commonly used UNB3 model and EGNOS model in real-time navigation and positioning.