<p>Shale lithofacies prediction is a fundamental component in shale oil and gas exploration. However, the strong heterogeneity of marine shale leads to insufficient identification accuracy of the traditional logging curve intersection graph method. Therefore, this paper takes the shales of Wujiaping Formation in Kaijiang–Liangping Trough as an example, aiming to establish a high-precision lithofacies identification model. The dream optimization algorithm (DOA) and support vector machine (SVM) are used to improve the classification performance by optimizing key parameters. The main lithofacies types of the Wujiaping Formation shale were determined based on X-ray diffraction, and six logging curves, including gamma ray (GR) and acoustic (AC) data, were selected as inputs. The data were preprocessed through balancing and outlier handling. Subsequent model validation was carried out using a combination of cross-validation and an independent test set. The results show that the main lithofacies shale types in the Wujiaping Formation are siliceous shale, mixed shale, and clay shale. The accuracy rate of predicting lithofacies using the DOA-SVM model reached 92.95%. Moreover, multiple evaluation indicators are superior to those of convolutional neural networks, extreme learning machine models, and the optimized SVM model. This study presents a highly efficient new method for identifying shale lithofacies. Its superior performance indicates that this method has good geological application potential in the identification of strongly heterogeneous shale lithofacies. It can provide reliable technical support for the fine evaluation of rock facies in unconventional oil and gas exploration.</p>