基于GMDH网络的电厂地下水取水自适应控制方法研究  

Research on adaptive control Method in taking ground water of power plant based on GMDH ANN

作　　者：皮玉珍[1,2] 苑全德[1,2] 孟祥萍[1,2] 谭万禹[1,2] 舒英利[1,2] 谷春苗[1] 

机构地区：[1]长春工程学院电气与信息工程学院 [2]吉林省高校配电自动化工程研究中心,长春130012

出　　处：《长春工程学院学报（自然科学版）》2011年第4期35-37,共3页Journal of Changchun Institute of Technology：Natural Sciences Edition

基　　金：吉林省教育厅"十一五"科学技术研究项目(2010420;2009241;2010181;2010190)

摘　　要：对某火电厂的取水系统中的新水取用控制进行了研究。提出了使用GMDH网络控制深井泵的方法。设计了一个多层GMDH网络,该网络通过遗传算法原理产生新的下一层神经元。通过对该电厂一年的深水井泵运行记录的转化提取获得学习样本和测试集,将该训练数据陆续输入到神经网络,进而获得对深水井是否进行启动或者停止的信息。给出了网络决策与实际运行结果对比,结果表明本方法对于深井泵的控制具有很好的效果。The control methods in a power plant ground water intaking system have been studied in this paper.A new method to control the deep well pump by GMDH ANN has been put forward.We design a multi-layer CMDH ANN,which generate new neurons of next layer by the theory of GA.We extract learning samples and test set from the one-year running record of ground water intaking system in this power plant.Moreover,the information on start and stop to this deep well has been achieved by inputting these training data into GMDH ANN gradually.The comparison between network predictions and actual running results has been made in the paper.The results show that this method is benefit to control the pumps in deep wells.

关 键 词：GMDH网络 遗传算法 自适应控制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]