Process of hydrological cycle is the basis of water resources reasonable allocation. With the characteristic of multi-time transformation and the pattern of the reutilization of water resources in the inland river basin, this paper attempts to establish an optimal allocation model based on the simulation of water resources transformation. The inland river basin is divided into three independent subsystems according to hydraulic connection. An independent subsystem is divided into three primary calculation units of canal irrigation district, well irrigation district and mixed canal and well irrigation district according to the characteristic of water resources transformation and pattern of water utilization. A primary calculation unit includes several irrigation districts (secondary calculated units). Two models of water balance and BP artificial neural network were established to simulate inflow of well irrigation district and mixed canal and well irrigation district respectively. Two-layer optimal models of primary and secondary were established to allocate water resources between primary and secondary calculation units respectively, with the objectives of maximum irrigation benefit and equity respectively. The optimal results of water resources allocation were obtained by the solution of simulation optimizing and genetic algorithms through combining the two simulation models. And the water resources of irrigation district and administrative district were allocated with Liangzhou 10.68×10⁸m³ and Jinchuan 2.63×10⁸m³, irrigation water is 16.61×10⁸m³ which is 75% of the whole basin water and reduced greatly when compared with present water use. The unity of allocation of basin and administrative district will be realized.