一类过程控制对象的神经网络建模及仿真

针对一类典型过程控制系统中存在的非线性和参数不确定问题，提出了神经网络的建模方法。辨识结构采用串并联形式，并分别使用改进BP算法和浮点式遗传算法进行了网络的训练。仿真结果表明遗传算法的全局搜索能力及高效率对神经网络的权值优化具有相当明显的效果，它不仅学习速度快而且稳定性好，可以作为一种良好的优化方法运用到神经网络建模和控制当中。

Neural Networks Modeling of a Class of Process Control Object and Its Simulation Research

A modeling method based on Neural Networks (NN) is proposed in this paper for the sake of tackling problems such as nonlinearity and parameter uncertainty in a typical process control system. The series-parallel identification structure is adopted to reflect characteristics of the control object well and truly. Furthermore, the improved BP (Back Propagation) algorithm and floating-point genetic algorithm are respectively employed to train the weights of networks. The simulation results show that the global searching ability and high efficiency of genetic algorithms have a marked effect not only on fast learning of the multi-layer forward neural networks, but also on stability of training. So, this method can be used in NN modeling and controlling as a good way of optimization.