机构地区: 哈尔滨工业大学
出 处: 《科学技术与工程》 2008年第8期2112-2118,共7页
摘 要: 针对复杂情况下模糊控制器难以获取经验规则的缺点,利用遗传算法来优化模糊控制器的规则,并采用一种权重和方法以实现多目标的优化控制。为了减少时滞对控制效果的影响,应用BP神经网络以预测模糊控制器的输入。基于76层风振Benchmark模型对提出的控制算法进行了计算仿真分析。结果表明神经-模糊控制(NN-FLC)方法在理想情况下与传统LQG控制算法控制效果相当;但在结构刚度不确定时,该方法具有较强的稳定性和鲁棒性,远优于LQG算法。 Using genetic algorithm to optimize the control rules to overcome the difficulty of acquisition of empirical rules for fuzzy logic controller, a weight-sum-approach method is employed here to realize the multi-objective optimal control. In order to reduce time delay, BP neural network is employed here to predict the inputs of fuzzy logic controller also. Trough the analysis of 76-story wind-excited Benchmark model, the result shows that the neu- ral network-fuzzy control (NN-FLC) method performs closely as the LQG under ideal condition; but in the case of uncertainty in stiffness of structures, the stability and robustness of this control method have been demonstrated, it performs much better than LQG.