Abstract: The prediction of remaining oil in complex fault block reservoirs has always been the key to guide well deployment and remaining oil tapping in the later stage of oil reservoir development. To improve the prediction efficiency and accuracy of remaining oil saturation field in complex fault block reservoirs, the paper constructs 2×10⁴ numerical simulation forward models reflecting different structure depth, reservoir thickness, permeability, and porosity, and acquires the distribution of remaining oil saturation field of each model, so as to build the saturation field data sample pool. Then deep convolutional adversarial neural network model is used for the training of data in sample pool, among which 70% of the data are randomly selected as the training set and 30% as the test set. Finally, the saturation field prediction method that can be applied to an actual block is established. The results show that for the new method, it does not need to carry out numerical simulation research on the actual block any more, and only requires to enter information such as physical reservoir parameters, well location coordinates and injection-production volume of the actual block, in which case the distribution of remaining oil saturation of the block at different moments can be obtained through the deep convolutional adversarial neural network model, with a prediction accuracy of up to 90%. After test, it is found that the deep convolutional adversarial neural network model has good generalization ability, and the model can be widely used for predicting the remaining oil saturation field in complex fault block reservoirs, which can greatly improve the research efficiency of reservoir exploitation.