To realize simulation of the two-layer progressive causal relationship of soil salt-water and crop production benefits under the influence of multiple factors, based on deep learning theory and technology, the progressive salt-water embedding neural network (PSWE) model was constructed. In PSWE model, the time serialized data encoder framed by hierarchical long short-term memory (HLSTM) and decoder framed batch-normalized multi-layer perceptron (BMLP) were coupled, and the coupling between Dropout and Adam algorithm was optimized as an improved algorithm for convergence regression. The validity of PSWE model was evaluated, and the dynamic changes of soil water-salt of different irrigation amounts under multi-factors cooperative straw deep burial were simulated, and the production benefit of summer maize was predicted. The results showed that PSWE model had multivariable overall synergy, self-learning habit and high accuracy. PSWE model could effectively describe the law of soil salt-water migration under straw deep burial in Hetao Irrigation District, the internal dependence relationship between summer maize production benefit and various variables. The root mean square error of the PSWE model was 0.031, the mean absolute error was 0.569, and the determination coefficient was 0.987. Through the model simulation, along with the summer maize growth period, the moisture content of treatment of single irrigation 60mm was reduced continuously in the tillage layer (0~40cm), and affected the summer maize for normal growth, while the change of treatment of 135mm was larger. In the mature stage, they produced salt accumulation in the straw inter-layer, and the salt accumulation rate was 49.2% and 11.2%. The water content in the tillage layer of single irrigation 90mm and 120mm were kept between 16% and 24%. At the end of the growth period, the water content in the soil layer over 40cm was kept stable. The straw inter-layer showed a trend of desalting, and the desalting rate was 6.1% and 5.9%, respectively. It was suggested that the single irrigation amount should be 89.3~96.8mm,and the theoretical salt content of cultivated layer was 1.38~1.55g/kg. In conclusion, under multi-factors cooperative straw deep burial, appropriate irrigation amount could achieve the goal of salt suppression effect and improvement of water use efficiency. The PSWE model could effectively simulate soil salt-water migration. The simulation of soil water-salt migration and crop productivity benefit by PSWE model was applicable, which provided a reference for deep learning theory and technology in soil salt-water migration.