基于一类GA-RBF神经网络的转炉炼钢静态模型控制

讨论了具有非线性、大时滞、不确定特性的工况复杂的转炉炼钢过程建模与控制问题.针对传统的控制方法控制效果差、精度不高,难以达到期望结果的问题,结合RBF神经网络的特点,提出用基于混合编码方式的混合遗传算法训练的RBF神经网络,同时优化网络的结构和参数,并利用RBF神经网络建立转炉炼钢静态模型.仿真结果表明,该模型具有在线调整和学习的功能,比传统模型具有更好的计算精度和适应能力,为提高转炉冶炼过程的控制精度给出了一个有效的方法.

A class of GA-RBF neural network control for the BOF steelmaking static model

The modeling and control problems of basic oxygen furnace(BOF) steelmaking process with nonlinearity,large time-delay,uncertainty and complexity are discussed.Considering the problem of poor effect and precision of the BOF steelmaking control,a radial base function(RBF) neural network based on hybrid coding genetic algorithm(GA) is presented combining with characteristic of RBF neural network.The structure and parameter of neural network is optimized by the proposed method,which is used to control the BOF steelmaking static model.The simulation results show that the new method can improve the on-line adjustment and self-learning capacity compared with the traditional methods.The GA-RBF neural network control is an effective method to improve the adaptability and practicability of BOF steelmaking control.