基于神经元网络的多回路控制系统的全局优化算法

A New Global Optimization Algorithm Based on Multi-loop Control System with Neural Networks Controller

提出了一种新的基于多回路控制系统的全局优化算法。该算法把目标函数设定为每个子系统的控制对象。通过运用神经元控制器和填充函数法,随着迭代的进行,目标函数输出值逐渐趋向于每个子系统的输入值,直至得到全局最优解。为了能从有许多局部极小点的控制对象中得到全局最优解,我们采用转换函数来改善控制对象,这可使得从多回路控制系统中找到全局最优解变得简单。同时我们把填充函数作为辅助函数,因为正是由于辅助函数的应用可以使得极小值点移动到下一个比当前目标函数值更小的极小值点。为了验证所提出方法的有效性,我们对5个基准测试用例进行了仿真,由后面的结果可得上述方法非常有效。

In this paper,a new optimization algorithm based on multi-loop control system with neural networks is presented,where the object function is used as the control plant of each sub-control system,the value of the object function is close gradually to the input of a sub-control system in the iterative process by using the controller of a special neural networks and filled function method,until its global optimization solution is found.For getting the global optimization solution from the control plant with many local minimum points,a transformation function is presented.It changes the complex object function into a simple function as much as possible on the condition of unchanged global optimal solution.On the other hand,multiple transformation functions also consist of special neural networks where the node function can simply be positioned locally.At the same time,a new filled function method is presented.