A method for building a water demand model that considered both water use efficiency (WUE) and photosynthetic rate for greenhouse crops was proposed. Firstly, a nested experiment was performed to measure the photosynthetic rate and WUE of tomato under different combinations of temperatures, photosynthetic photon flux densities (PPFD), CO2 concentrations and soil moisture. Secondly, the photosynthetic rate prediction model and WUE prediction model were established by using the radial basis function (RBF) algorithm. On this basis, the response curve of photosynthetic rate to soil moisture was obtained. Then, the optimal soil moisture ranges under certain environmental conditions were found by applying the Uchord discrete curvature algorithm and particle swarm optimization (PSO) algorithm. At last, the water demand model was established based on the support vector machine regression (SVR) algorithm. The results showed that the model was of high accuracy, with determination coefficient of 0.9969, and mean square error of 0.23%. Compared with the water demand model that only considered photosynthetic rate, this model increased the WUE by 15.22% on average, while the soil moisture and the photosynthetic rate were decreased by 12.76% and 4.05% on average, respectively. These results proved that the crop water demand model proposed can take good account of both crop demand and agricultural water consumption, and provide a theoretical basis for the dynamic and efficient soil moisture regulation of greenhouse crops.