In order to improve the accuracy of spatial power system fault detection,Convolutional Neural Networks (CNN) are combined with Long Short Term Memory (LSTM).A fault detection model of space power system based on adaptive selection is proposed to realize the fault detection of space power system.By establishing digital twin virtual model,introducing typical faults,increasing the type and quantity of fault data,it is used as the data set of training model.The CNNLSTM algorithm is used for machine learning and training of sample data set to build fault detection model.The experiment results show that the fault detection accuracy of CNNLSTM model can reach 98% and the loss function is less.For each type of fault detection,the balance F score is as low as 96% and as high as 100%.The results demonstrate the effectiveness and feasibility of the proposed scheme,which has certain practical value.