The water resource optimization allocation model is established with the minimum regional water shortage rate and the minimum pollutant emission as the objective function． Taking the valley in the lower reaches of the Fenhe Ｒiver as an example，this paper predicts the water supply and demand in 2025 on condition of 20%，50%，75% and 95% rainfall frequency． Aiming at the problem that the calculation of Moth－Flame Optimization algorithm takes a long time and is easy to fall into local optimum，the adaptive formula for the number of flames and the logarithmic spiral formula are improved． Through the simulation experiment of the test function commonly used in swarm intelligence algorithm，the optimization results of Moth－Flame Optimization algorithm before and after the improvement are compared and analyzed． The improved Moth－Flame Optimization algorithm is used to solve the model，and the results show that the water resource optimization allocation model is reasonable and effective，and the improved Moth－Flame Optimization algorithm has fast convergence rate and optimal performance， which can be used for an analysis of optimal allocation of water resources．