【摘要】 提出了一种具有模糊分割的动态神经网络（ＤＮＮＦＰ）．给出了网络的拓扑结构，得到了该网络在监督学习与再激励学习方式下的学习算法，并将其成功地应用于ｐＨ中和过程这一典型的连续时间非线性动态系统的控制．该文的研究表明，这种动态模糊神经网络综合了模糊逻辑、ＣＭＡＣ网络以及再激励学习的有关结果，不仅具有明确的物理意义，而且无需以任何显式方式建立被控对象的数学模型．由此获得的动态学习控制系统结构简单、鲁棒性强并具有广泛的适用性．在相同控制限幅下，利用该方法获得的结果较之传统方法具有更为优越的性能更多还原

【Abstract】 This paper proposes a novel dynamic neural network with fuzzy partitioning ——DNNFP. After the network structure of DNNFP is introduced, the learning algorithms of this network, i.e. the supervised learning algorithm and the reinforcement learning one, are described respectively. The control of the pH neutralization process, which is known as a highly nonlinear continuous system,is illustrated by using the proposed network (together with the two learning algorithms). The simulation results show a superior control performance compared to the conventional PID methods.更多还原