In order to achieve high accuracy of ionospheric total electron content (TEC) short-term prediction for Europe, a novel hybrid deep learning model was established by applying the dung beetle optimizer (DBO) algorithm to optimize a bidirectional long short-term memory (BiLSTM) neural network, named DBO-BiLSTM. TEC predicted by this model was compared with TEC computed using GPS observations released by the European Permanent Global Navigation Satellite System network and with predictions from SSA-BiLSTM, BiLSTM, and LSTM models. Test results indicate that DBO-BiLSTM predictions have the closest agreement with GPS-derived values and the highest prediction accuracy, with root mean square errors for 1-h and 2-h predictions reaching 0.57 TECU and 0.92 TECU, respectively. The optimized model effectively captures spatial-temporal ionospheric changes under quiet and moderately disturbed geomagnetic conditions during a moderate solar activity period, providing a useful model for high-accuracy short-term TEC prediction in Europe.