There are differences in the discharge data of batteries at different life stages. Considering these differences, this paper proposes a conditional selection neural network model for estimating the SOE. It can flexibly select the appropriate network for prediction based on a set selection mechanism under different conditions, thereby improving the model’s applicability and generalization ability. Firstly, the concept of Squeeze-and-Excitation is incorporated into the CNN and the resulting SE-CNN model is pretrained. Next, a selection mechanism is designed according to the degree of battery aging, thus the training data is divided into four parts based on the number of discharge cycles to further train the pre-trained network to form a “four-in-one” CS-SE-CNN model. During SOE estimation, the CS-SE-CNN model selects the prediction network based on the number of discharge cycles in the input data. The results demonstrate that the proposed method exhibits high accuracy and robustness in estimating the SOE of lithium-ion batteries.