Abstract: In the process of flood control and operation calculation for estuaries, it is often encountered that under the determined water level at a certain section in the research section of the estuary, the double boundaries(upstream flow and downstream tide level) conditions are required to derive in reverse.Aiming at the problems of low computational accuracy and search efficiency of the traditional trial algorithm based on the MIKE 11 model for inverse deducing of estuarine double boundaries conditions, this paper proposed an optimization method using Back Propagation(BP) neural network and Particle Swarm Optimizer(PSO) algorithm to improve the traditional trial algorithm.Specifically, the offline database of the double boundary of the estuary and the water level of acertain section were established by the MIKE 11 model.Subsequently, the BP neural network was used to establish the high-precision non-linear mapping relationship between the double boundaries and the water level of the section.Finally, taking the determined water level of the section as the optimization objective, PSO algorithm was used to reversely calculate and determine the mapping relationship of the double boundaries.A case study of Jinjiang River Estuary showed that compared with the traditional trial algorithm, the optimization method reduced the calculation time to about one-tenth of the original one, and greatly improved the calculation efficiency.The research results can provide reference basis for flood control and operation of Jinjiang River Estuary.