To address the limitations of traditional lithology identification methods in accuracy and efficiency, as well as the poor generalization of existing machine learning models under complex geological environments, this study proposes a logging lithology identification method based on granular-ball space, termed LogGBS. This method maps raw well log data into the granular-ball space and constructs multiscale granular-ball representations through an Adaptive Granular-Ball Generator (AGBG). Using "granular-ball purity"as the key control parameter, LogGBS ultimately performs lithology identification by a Granular-Ball kNN (GBkNN) classifier. To evaluate the effectiveness of the proposed method, well log data from the Hangjinqi Gas Field (clastic reservoir) and Sulige Gas Field (carbonate reservoir) in the Ordos Basin, as well as the Xushen Gas Field (volcanic reservoir) in the Songliao Basin, are selected as research objects. The performance of LogGBS is compared with that of eight traditional machine learning and deep learning methods, and the influence of granular-ball purity on classification accuracy is analyzed. The results show that when the purity is not less than 0.9, the average accuracy of each reservoir exceeds 90%; when the purity reaches 0.95, some indicators even surpass 99%. LogGBS demonstrates superior identification accuracy and generalization ability in all three types of reservoirs, providing a new idea for lithology identification based on well logs under complex geological conditions and contributing to the improvement of oil and gas exploration and development efficiency.