The control rod drive plugin is a critical component for implementing control rod actions. To enhance the stability and safety of nuclear reactor operations, this study focuses on the prediction methods for the operational status of control rod drive plugins. Addressing the issue that traditional deep learning methods have low accuracy in predicting the operational status of control rod drive plugins, a hybrid model based on an attention mechanism, convolutional neural network, and bidirectional long short-term memory network (CNN-BiLSTM-AM) is proposed. First, operational data of the control rod drive plugin during lift/insertion is collected, and the acquired data is normalized and preprocessed. Then, the processed data is input into the CNN-BiLSTM-AM hybrid model for training to determine the optimal parameters of the hybrid model. Finally, the CNN-BiLSTM-AM hybrid model is compared with other models in terms of accuracy, loss value, root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). The results demonstrate that the CNN-BiLSTM-AM hybrid model is more effective for predicting the status of control rod drive plugins.