Abstract: In order to accurately predict downhole weight on bit (WOB), improve drilling efficiency, and reduce drilling cost, a hybrid model combining bidirectional long short-term memory network (Bi-LSTM) and convolutional neural network (CNN) was established. The model used the trigonometric function-driven particle swarm optimization (TDCSO) method for hyperparameter optimization, so as to improve the accuracy of WOB prediction. The public data sets of Well FORGE 58−32 and Well FORGE 58−62 in Utah were used to verify the established model, and the model performance was evaluated by the mean absolute error (MAE), root mean square error (RMSE), coefficient of determination, and mean square error (MSE). The results show that the proposed TDCSO-CNN-Bi-LSTM model achieves excellent results in three key indicators of MAE, MSE, and RMSE, and the coefficient of determination was higher than 0.98, which is significantly better than the existing methods such as LSTM, GRU, CNN-LSTM and CNN-Bi-LSTM, etc. The study shows that the proposed TDCSO-CNN-Bi-LSTM model has excellent accuracy in downhole WOB prediction and enables real-time monitoring. It can be integrated with an automated drilling system to achieve precise control of WOB. This not only improves drilling efficiency but also reduces drilling cost, which has important practical application value for future drilling operations.