蚁群优化的改进型神经网络控制器设计

Design of Improved Neural Network Controller Optimized with Ant Colony Algorithm

利用可调激活函数改进神经网络,提高神经网络的泛化能力,并推导出激活函数参数的调节公式;对蚁群算法加以改进,以加快算法收敛速度、避免局部最优解,利用改进的蚁群算法对神经网络参数加以优化,并用优化后的改进型神经网络在线整定控制器参数。仿真表明,对于较复杂的系统该方法比传统的神经网络PID控制有更好的控制效果。

Adjustable activation function was used to improve the generalization ability of neural network and the adjustment formula for activation function parameter was derived. Improving the ant colony algorithm to accelerate the convergence rate and to avoid the local optimal solution was implemented,including making use of the improved ant colony algorithm to optimize the neural network and setting the controller parameters on line.The simulation results show that this method has better control effect on complex system than the traditional neural network PID controller.