This research introduces a novel framework for the modeling and optimization of a morphing wing bio-inspired by fish skeleton structures. The framework employs a hybrid machine learning model that synergistically combines Deep Q-Learning (DQL) with the Gray Wolf Optimizer (GWO) algorithm. The GWO metaheuristic efficiently searches for the optimal hyperparameters of the DQL neural network, which learns to predict aerodynamic coefficients. This integration enhances convergence speed, prevents entrapment in local minima, and significantly improves prediction accuracy. Once trained, the model provides real-time predictions of optimal wing geometry under variable flight conditions. Validation against experimental and numerical reference data demonstrates the superior accuracy of the proposed GWO-DQL approach compared to a standard DQL model. This work represents a significant step forward in integrating metaheuristic optimization with deep reinforcement learning for the design of intelligent, adaptive aerospace systems. The practical significance of this research is multifaceted regarding artificial intelligence. In this way, the trained model provides a powerful tool to reduce the reliance on costly (CFD) simulations and wind tunnel experiments by estimating aerodynamic responses in real-time. This capability is best visualized by the model's ability to generate comprehensive lift-drag polar curves for the entire operational envelope of the morphing wing.