水产饲料配方优化的人工智能技术

Artificial intelligence technique for optimizing aquatic fodder recipe

本文面对水产饲料配方优化中假设目标函数和约束条件均为线性而采用单纯型法求解方法的不足，指出水产饲料配方的优化应当是多项式目标函数、线性和非线性多种约束条件的非线性规划。传统的非线性规划的求解是一个迭代过程，存在着收敛性问题，且约束条件的数学模型要求十分精确。通过引入人工智能理论与技术，为复杂的水产饲料配方代化求解提供了新途径。

This paper analyzes the method available for optimizing aquatic fodder recipe,based on linear programming of linear objective function and linear subjected bind. The results indicate that the effectlve optimization of aquatic fodder recipe is a non-linear programming with polynomial objective function and many linear and non-linear subjected binds. The solution is a repeated process and needs highly accurate mathematical model of subjected binds as there exists convergence,if an optimizing method of non-linear programming is adopted. The artificial intelligence technique,expert system and arificial neural network,is introduced. The new optimization method overcomes the shortage occurred when non-linear programming is applied in practice.