Ionospheric delay is one of the important sources of error in precision single-point positioning, time synchronization and other related fields. Accurate prediction of the total ionospheric electron content is an important prerequisite for compensating for ionospheric delay. This paper uses the long short-term memory（LSTM）network prediction algorithm to construct the TEC prediction model. The influences of different hidden layers, the number of neurons, the number of training data, the number of training times, and whether the data is preprocessed on the prediction of ionospheric data are verified. The best parameters of ionospheric prediction are obtained. The results show that the number of hidden neurons in the 2nd layer, the 1st and 2nd layers are 200 and 300, respectively, the number of input neurons is 168, the number of output neurons is 12, and the number of model iterations is 400 times, the prediction effect is the best, and the root mean square error of the prediction results of 18 single-site TEC prediction results is 43.87, which is 275.58 lower than that of the BP neural network algorithm. The LSTM network prediction algorithm effectively improves the prediction accuracy, and the clock error can be effectively compensated.