基于神经网络和遗传算法的PID参数自整定算法

PID parameter self‐tuning algorithm based on neural network and genetic algorithm

针对常规PID(proportional-integral-derivative)参数整定过程依赖被控对象的数学模型、不能灵活得到期望的性能指标等问题,提出了一种基于神经网络和遗传算法的PID参数自整定算法。该算法首先通过训练神经网络得到PID控制器参数和控制性能指标之间的映射关系;再利用遗传算法进行最优解搜索,得到全局范围内一组最优的控制器参数,使得给定的基于时域性能指标的目标函数值最优;最后,以水箱液位系统为例对所提算法进行实验验证。实验结果表明,与传统的ZN(Ziegler Nichols)参数整定法相比,所提算法整定的参数具有了更好的控制效果,同时可以通过改变目标函数中的性能指标权重系数,灵活得到不同期望的控制效果。

A PID(proportion-integra-derivative)controller parameter self-tuning algorithm based on neural network and genetic algorithm is proposed to solve the problems of relying on the mathematical model of the controlled plant and difficulties to acquire the expected performance index in the conventional PID parameter tuning process.First of all,the algorithm obtains the mapping relationships between PID controller parameters and control performance indicators through training the neural network.Then,the genetic algorithm is used to search for a group of global optimal PID controller parameters,so that the value of given function based on the time domain performance indexes is optimal.Finally,the proposed algorithm is verified by experiments with the water tank level system.The results show that,compared with the traditional ZN(Ziegler Nichols)parameter tuning algorithm,the parameters set by proposed algorithm get better control effects.Meanwhile,different desired control effect can be obtained flexibly by changing the weight coefficient of performance index in the objective function.