Abstract: [Objective] This study aims to enhance the accuracy and efficiency of lithology identification from well logs during oil and gas exploration and development. It addresses the limitations of traditional methods that heavily rely on expert judgment, are labor-intensive and time-consuming, and fail to reveal the complex relationships among logging parameters. The objective is to provide a more scientifically robust basis for reservoir evaluation and modeling. [Methods] To tackle the issue of incomplete well logging data, a multi-layer perceptron neural network optimized via a particle swarm optimization algorithm is first employed to fill in missing curves. Subsequently, a Generative Adversarial Network (GAN) is utilized to generate well log data that conforms to the actual distribution based on the number of lithology categories. Building upon this, a hybrid model integrating a Convolutional Neural Network (CNN) with a Transformer architecture is constructed. The Optuna hyperparameter optimization framework is then applied to rapidly identify the optimal set of model parameters. This integrated model is designed to effectively fuse local and global features, thereby addressing the challenges inherent in the identification of complex carbonate lithologies. [Results] Experimental validation using blind well log data from a specific oil field demonstrates that the hybrid model achieves a lithology identification accuracy of approximately 93%. In addition, it exhibits enhanced robustness and generalizability in discriminating carbonate lithologies. Comparative experiments further indicate that the model consistently captures key features across different depth intervals and complex geological settings, underscoring its strong applicability and scalability. [Conclusion] The proposed GAN-Optuna-CNN-Transformer hybrid model successfully mitigates the shortcomings of traditional methods related to data dependency and feature extraction. It enables efficient and accurate lithology identification, particularly for carbonate formations.