Abstract: Enhancing drilling efficiency is crucial for reducing costs and accelerating energy extraction. This study proposes a multi-objective coupling optimization model based on such objective functions as mechanical specific energy (MSE) and rate of penetration (ROP), aiming to improve the drilling efficiency through deep learning prediction and intelligent algorithm optimization. Initially, various deep learning architectures were evaluated, thereby, CNN-BiGRU-Attention model was selected to predict MSE and ROP, which combines Convolutional Neural Network(CNN), Bidirectional Gated Recurrent Unit(BigRU), and Attention mechanism. Subsequently, a multi-objective coupling optimization model was established, and three optimization algorithms including NSGA-II (Non-dominated Sorting Genetic Algorithm II), SPEA2 (Strength Pareto Evolutionary Algorithm 2), and RVEA (Reference Vector-guided Evolutionary Algorithm) were employed to solve the model. The optimization performance of the three algorithms was comparatively analyzed under different conditions of minimum ROP limits, corresponding to 50%, 70%, 90%, and initial ROP value at the current depth. The results indicated that the RVEA algorithm performed the best in the multi-objective coupling optimization of MSE and ROP. To align more closely with practical applications, torque constraint was introduced, and a corresponding deep learning model was established to investigate the impact of adjustments in rotational speed and drilling pressure on torque. The experimental results demonstrated that the RVEA algorithm could still effectively optimize MSE and ROP even after incorporating torque. This study does not only identify optimal strategies for reducing MSE and increasing ROP under various ROP constraints, but also provides practical theoretical guidance and optimization solutions for drilling engineering parameters.