Abstract: Aiming at the problem that the seawater quality assessment model based on traditional BP neural network falls easily into local minimum, a new seawater quality assessment model (BSO-BP) used BP neural network optimized by brainstorm optimization algorithm (BSO) is proposed. The model introduces a brainstorming optimization algorithm with global optimization features, which is used to simulate the process of human beings to solve problems with creative thinking. It has powerful global and local search capabilities of BP neural network with good nonlinear mapping ability. Taken advantage of learning adaptability and fault tolerance, and taken into account the nonlinear and nonstationary relationship of marine water quality assessment factors to the greatest extent, the optimal solution of the weight and threshold of each layer of BP neural network is obtained for much accurate and reasonable result in seawater quality assessment. The monitoring data from 12 stations in the Jiaozhou Bay is used for water quality evaluation. The experimental results show that the evaluation model can overcome local minimum problem, and the evaluation results are highly accurate and have certain practicability.