【作者】 徐辰华；

【导师】 孔峰；

【作者基本信息】 广西大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 神经网络PID控制属于一种先进的智能控制策略，其中的神经网络通常采用较成熟的BP网络。针对BP网络的一些优缺点，本论文提出在神经网络PID控制中使用具有学习算法简单、收敛速度快等优点的CMAC神经网络。CMAC是Cerebellar Model Articulation Controller的缩写，它是一种模拟小脑功能的神经网络模型，其学习方法采用δ算法。CMAC神经网络和常规PID控制器相结合，共同作用构成一个复合控制方案。对比、分析上述两种神经网络PID控制的控制效果。为了获得更好的控制效果，在上述智能控制算法的基础上，采用二进制编码的遗传算法(GA)，对CMAC神经网络内部结构进行优化，以便达到提高控制系统精度的要求。将优化后的CMAC神经网络用于复合控制方案，实现GA、CMAC、PID三者的结合。 针对无刷直流电动机进行了仿真试验，仿真结果验证了该算法的有效性，表明这种控制方案能够有效的提高系统的实时性能，并且具有很好的适应性和鲁棒性。更多还原

【Abstract】 The NN-PID control is an advanced intelligence control strategy. Back Proragation (BP) is often applied in NN-PID control. By comparison, it is opposed that CMAC is used in this paper, because of its simple algorithm and high convergence speed. CMAC is abbreviation of Cerebellar Model Articulation Controller and it is a kind of neural network model simulating cerebellar function. CMAC together with PID form an concurrent control scheme. CMAC uses δ algorithm in learning. An analysis and a comparison are given the effect of two algorithms on controlling. In order to get better control, based on the above intelligence control algorithm, a binary genetic algorithm (GA) is used to optimize CMAC internal structure, so that the control system can gain high accuracy. Optimized CMAC is used in this concurrent control scheme so as to perform the combination of GA, CMAC and PID.This concurrent control scheme is used for the brushless DC motor (BLDC) control system, and meanwhile some simulation tests are finished. The results of simulation show this new algorithm is effective, and demonstrate that this control method can effectively improved real time performance of control system, and it is also adaptive and robust.更多还原