Drought was an important factor restricting agricultural production and economic development. It was of great significance for promoting economic development and ensuring food security to study the law of occurrence and development of drought and effectively predict the local future drought situation. The purpose was to verify the applicability of vegetation temperature condition index (VTCI) in the drought prediction during summer maize growing season. Taking the central plain of Hebei as the research area and the time series of drought monitoring results of vegetation temperature condition index as the data source, and autoregressive integrated moving average (ARIMA) model and seasonal autoregressive integrated moving average (SARIMA) model were used to forecast agricultural drought. First of all, based on the time series of vegetation temperature condition index of 49 meteorological stations, the VTCI data of different lengths were used to build ARIMA prediction models, and the variation characteristics of ARIMA model prediction accuracy with the increase of VTCI time series length were analyzed. The results showed that there existed no clear dependence between the performance of the model and the training lengths corresponding to the historical datasets of VTCI, but the prediction accuracy of the model tended to be stable with the increase of time series length. Then, the VTCI time series data from early July 2010 to late August 2017 was used as modeling data, the ARIMA model and SARIMA model were applied to predict VTCI in September 2017, and the prediction accuracy of the two models was evaluated. The results showed that the prediction accuracy of the ARIMA model was higher than that of the SARIMA model. The root mean square error of the 1-step VTCI prediction of the ARIMA model was 0.06 lower than that of the SARIMA model, and the 2-step prediction was 0.07 lower, and the 3-step prediction was 0.09 lower. Therefore, the ARIMA model was more suitable for the drought prediction during the summer maize growing season in the study area. Finally, the ARIMA model with better performance was modeled pixel by pixel to obtain the VTCI prediction results from early September to late September, 2016—2018. The results showed that the ARIMA model had a good prediction accuracy for 1-step, 2-step and 3-step of VTCI during summer maize growth season in different years. The average percentage of pixels with absolute error larger than 0.20 in 1-step, 2-step and 3-step in 2016—2018 was only 5.84%, 6.38% and 8.72%, respectively.